Behavioral and Brain Sciences, Volume 32, Issue 1, February 2009

An International journal of current research and theory with open peer commentary Volume 32 | Issue 1 | February 2009 | ISSN: 0140-525X

Behavioral and Brain Sciences Editors Paul Bloom Barbara L. Finlay URL: http://www.editorialmanager.com/bbs E-mail: [email protected] Behavioral and Brain Sciences Journals Department Cambridge University Press 32 Avenue of The Americas New York, NY 10013-2473, U.S.A.

Editorial Administrator Ralph DeMarco

Chief Copy Editor Sumitra Mukerji

Proofreaders Sylvia Elvin Rashidah Ismaili AbuBakr

Editorial Board Atypical Neurodevelopment Simon Baron-Cohen/Cambridge U. Behavioral Neurogenetics Wim E. Crusio/CNRS UMR Cognition and Artificial Intelligence Zenon Pylyshyn/Rutgers U. Cognitive and Decision Sciences Nick Chater/University College London Cognitive Development Annette Karmiloff-Smith/Birbeck College Computational Neuroscience Nestor A. Schmajuk/Duke U. Evolution of Brain and Cognition Dean Falk/Florida State U. Evolution of Cognition Celia Heyes/Oxford U. Experimental Analysis of Behavior A. Charles Catania/U. Maryland, Baltimore County Language and Language Disorders Max Coltheart/Macquarie U. Linguistics Robert A. Freidin/Princeton U. Perception Bruce Bridgeman/U. of California, Santa Cruz Philosophy of Science Massimo Piattelli-Palmarini/U. of Arizona Primate Cognition Laurie R. Santos/Yale U. Social Cognition Mahzarin R. Banaji/Harvard U. Social Cognitive Neuroscience Rebecca Saxe/MIT Vision, Language and Computation Shimon Edelman/Cornell U.

FOUNDING EDITOR (1978–2001)

Editorial Policy Behavioral and Brain Sciences (BBS) is an international journal providing a special service called Open Peer Commentary* to researchers in any area of psychology, neuroscience, behavioral biology, or cognitive science who wish to solicit, from fellow specialists within and across these BBS disciplines, multiple responses to a particularly significant and controversial piece of work. (See Instructions for Authors and Commentators, inside back cover and also at http://www.editorialmanager.com/bbs/ account/BBS_ifc.pdf) The purpose of this service is to contribute to the communication, criticism, stimulation, and particularly the unification of research in the behavioral and brain sciences, from molecular neurobiology to artificial intelligence and the philosophy of mind. Papers judged by the editors and referees to be appropriate for Commentary are circulated to a large number of commentators selected by the editors, referees, and author to provide substantive criticism, interpretation, elaboration, and pertinent complementary and supplementary material from a full cross-disciplinary perspective. The article, accepted commentaries, and the author’s response then appear simultaneously in BBS. Commentary on BBS articles may be provided by any qualified professional in the behavioral and brain sciences, but much of it is drawn from a large body of BBS Associates who have become formally affiliated with the project (see http://www.bbsonline.org/assoclist.html). Qualified professionals are eligible to become BBS Associates if they have (1) been nominated by a current BBS Associate, (2) refereed for BBS, or (3) had a commentary or article accepted for publication. A special subscription rate is available to Associates. Individuals interested in serving as BBS Associates are asked to view the full instructions for joining at http://www.editorialmanager.com/bbs/account/BBS_ifc.pdf. and then email [email protected]. © 2009 Cambridge University Press. All rights reserved. No part of this publication may be reproduced, in any form or by any means, electronic, photocopy, or otherwise, without permission from Cambridge University Press. General enquiries from the U.S.A., Mexico, and Canada should be addressed to the New York office of Cambridge University Press http://www.cambridge.org/us/information/rights/contacts/newyork.htm; general enquiries from elsewhere should be addressed to the Cambridge office http://www.cambridge.org/ uk/information/rights/contacts/cambridge.htm; permission enquiries from Australia and New Zealand should be addressed to the Melbourne office http://www.cambridge.org/aus/ information/contacts_melbourne.htm; enquiries regarding Spanish-language translation rights (only) should be addressed to the Madrid office http://www.cambridge.org/uk/ information/rights/contacts/madrid.htm. Copying This journal is registered with the Copyright Clearance Center (222 Rosewood Dr., Danvers, MA 01923, U.S.A.). Organizations in the U.S.A. who are also registered with the CCC may therefore copy material (beyond the limits permitted by sections 107 and 108 of U.S. Copyright Law) subject to payment to the CCC. This consent does not extend to multiple copying for promotional or commercial purposes. Notwithstanding the above, the Item-Fee Code for this publication is 0140-525X/09. ISI Genuine Article Service, 3501 Market Street, Philadelphia, PA 19104, U.S.A. is authorized to supply single copies of separate articles for private use only. For all other use, permission should be sought from the Cambridge or New York offices of the Press. Subscriptions Behavioral and Brain Sciences (ISSN 0140-525X) is published bimonthly in February, April, June, August, October, and December. The subscription price of Volume 32 (2009) for institutions is US $1020.00 for print and online, US $860.00 for online only, and US $920.00 for print only in the U.S.A., Canada, and Mexico; and UK £580.00 for print and online, UK £490.00 for online only, and UK £530.00 for print only elsewhere. The subscription price for individuals is US $290.00 for print and online and US $265.00 for print only in the U.S.A., Canada, and Mexico; and UK £170.00 for print and online and UK £155.00 for print only elsewhere. For BBS Associates, with proof of eligibility with order, US $109.00 in the U.S.A., Canada, and Mexico; and UK £68.00 elsewhere. For students, with proof of eligibility with order, $89.00 in the U.S.A., Canada, and Mexico; and UK £54.00 elsewhere. Subscription price includes surface postage. Single parts cost US $210.00 (UK £90.00) plus postage. Institutional orders may be sent to a bookseller, or, in the U.S.A., Canada, and Mexico direct to: Cambridge University Press, 32 Avenue of The Americas, New York, NY 10013-2473 email: [email protected]; in the U.K. and rest of the world to: Cambridge University Press, The Edinburgh Building, Shaftesbury Road, Cambridge CB2 8RU, England, e-mail:[email protected]. Individuals must order direct from the Press. You may also subscribe through the Cambridge Journals website, http://journals.cambridge.org/bbs. Postmaster: Send address changes in the U.S.A., Canada, and Mexico to Behavioral and Brain Sciences, Cambridge University Press, Journals Dept., 100 Brook Hill Drive, West Nyack, NY 10994-2133, U.S.A. Send address change elsewhere to Behavioral and Brain Sciences, Cambridge University Press, The Edinburgh Building, Shaftesbury Road, Cambridge CB2 8RU, England. Online availability Behavioral and Brain Sciences is part of the Cambridge Journals Online (CJO) service at http://journals.cambridge.org. Institutional subscribers: Access to full-text articles online is currently included with the cost of the print subscription. Subscription must be activated; see http://cambridge.journals.org. Advertising Inquiries about advertising should be sent to the Journals Advertising Department of the Cambridge or New York Office of Cambridge University Press.

Stevan Harnad *Modelled on the “CA Comment” service of the journal Current Anthropology.

Contents

Volume 32:1

February 2009

Del Giudice, M. Sex, attachment, and the development of reproductive strategies Open Peer Commentary Ackerman, J. M. & Kenrick, D. T. Selfishness and sex or cooperation and family values? Bakermans-Kranenburg, M. J. & van IJzendoorn, M. H. No reliable gender differences in attachment across the lifespan Beckes, L. & Simpson, J. A. Attachment, reproduction, and life history trade-offs: A broader view of human mating Campbell, A. “Fatal attraction” syndrome: Not a good way to keep your man Chen, B.-B. & Li, D. Avoidant strategy in insecure females Figueredo, A. J., Sefcek, J. A. & Olderbak, S. G. Attachment and life history strategy Flinn, M. V., Muehlenbein, M. P. & Ponzi, D. Evolution of neuroendocrine mechanisms linking attachment and life history: The social neuroendocrinology of middle childhood Goetz, C. D., Perilloux, C. & Buss, D. M. Attachment strategies across sex, ontogeny, and relationship type Goldstein Ferber, S. Co-regulation of stress in uterus and during early infancy mediates early programming of gender differences in attachment styles: Evolutionary, genetic, and endocrinal perspectives Harris, J. R. Attachment theory underestimates the child Ho¨nekopp, J. Pre-adjustment of adult attachment style to extrinsic risk levels via early attachment style is neither specific, nor reliable, nor effective, and is thus not an adaptation Jackson, J. J. & Ellis, B. J. Synthesizing life-history theory with sexual selection: Toward a comprehensive model of alternative reproductive strategies

31

Kang, M. J. & Glassman, M. Attachment patterns of homeless youth: Choices of stress and confusion Kerns, K. A. Developmental transformations in attachment in middle childhood Kruger, D. J. Life history as an integrative theoretical framework advancing the understanding of the attachment system Lewis, A. J. & Tooley, G. Disorganized attachment and reproductive strategies Li, N., He, J. & Li, T. Gender difference of insecure attachment: Universal or culture-specific? Maestripieri, D. The contribution of comparative research to the development and testing of life history models of human attachment and reproductive strategies Penke, L. Adaptive developmental plasticity might not contribute much to the adaptiveness of reproductive strategies Petters, D. & Waters, E. Modeling, simulating, and simplifying links between stress, attachment, and reproduction Quinlan, R. J. Predicting cross-cultural patterns in sex-biased parental investment and attachment Seltzer, L. J. & Pollak, S. D. Neuroendocrine features of attachment in infants and nonhuman primates Symons, D. K. & Szielasko, A. L. Attachment styles within sexual relationships are strategic Volpe, L. E. & Barton, R. A. Attachment and sexual strategies Zayas, V. & Ram, D. What love has to do with it: An attachment perspective on pair bonding and sexual behavior

31

Author’s Response Del Giudice, M. Human reproductive strategies: An emerging synthesis?

21 22 23 24 25 26

27 28

29 30

1 32 33 34 35 36

37 38 39 40 41 42 43 44

45

Oaksford, M. & Chater, N. Pre´cis of Bayesian Rationality: The Probabilistic Approach to Human Reasoning Open Peer Commentary Allott, N. & Uchida, H. Classical logic, conditionals and “nonmonotonic” reasoning Brighton, H. & Olsson, H. Identifying the optimal response is not a necessary step toward explaining function Danks, D. & Eberhardt, F. Explaining norms and norms explained De Neys, W. Beyond response output: More logical than we think Evans, J. St. B. T. Does rational analysis stand up to rational analysis? Griffiths, T. L. The strengths of – and some of the challenges for – Bayesian models of cognition Hahn, U. Explaining more by drawing on less Halford, G. S. Complexity provides a better explanation than probability for confidence in syllogistic inferences Khalil, E. L. Are stomachs rational? Liu, I.-m. Is the second-step conditionalization unnecessary? McKenzie, C. R. M. Bayes plus environment Nelson, J. D. Naı¨ve optimality: Subjects’ heuristics can be better motivated than experimenters’ optimal models Oberauer, K. Oaksford & Chater’s theory of reasoning: High prior, lower posterior plausibility

85 85 86 87 88 89 90 91 91 92 93

O’Brien, D. P. Human reasoning includes a mental logic Over, D. E. & Hadjichristidis, C. Uncertain premises and Jeffrey’s rule Pfeifer, N. & Kleiter, G. D. Mental probability logic Poletiek, F. H. Popper’s Severity of Test as an intuitive probabilistic model of hypothesis testing Politzer, G. & Bonnefon, J.-F. Let us not put the probabilistic cart before the uncertainty bull Schroyens, W. On is an ought: Levels of analysis and the descriptive versus normative analysis of human reasoning Stenning, K. & van Lambalgen, M. “Nonmonotonic” does not mean “probabilistic” Straubinger, N., Cokely, E. T. & Stevens, J. R. The dynamics of development: Challenges for Bayesian rationality Wagenmakers, E.-J. How do individuals reason in the Wason card selection task?

69 96 97 98 99 100 101 102 103 104

94 95

Authors’ Response Oaksford, M. & Chater, N. The uncertain reasoner: Bayes, logic, and rationality

105

BEHAVIORAL AND BRAIN SCIENCES (2009) 32, 1 –67 Printed in the United States of America

doi:10.1017/S0140525X09000016

Sex, attachment, and the development of reproductive strategies Marco Del Giudice Center for Cognitive Science, Department of Psychology, University of Turin, 10123 Torino, Italy [email protected] http://www.psych.unito.it/csc/pers/delgiudice/delgiudice.html

Abstract: This target article presents an integrated evolutionary model of the development of attachment and human reproductive strategies. It is argued that sex differences in attachment emerge in middle childhood, have adaptive significance in both children and adults, and are part of sex-specific life history strategies. Early psychosocial stress and insecure attachment act as cues of environmental risk, and tend to switch development towards reproductive strategies favoring current reproduction and higher mating effort. However, due to sex differences in life history trade-offs between mating and parenting, insecure males tend to adopt avoidant strategies, whereas insecure females tend to adopt anxious/ambivalent strategies, which maximize investment from kin and mates. Females are expected to shift to avoidant patterns when environmental risk is more severe. Avoidant and ambivalent attachment patterns also have different adaptive values for boys and girls, in the context of same-sex competition in the peer group: in particular, the competitive and aggressive traits related to avoidant attachment can be favored as a status-seeking strategy for males. Finally, adrenarche is proposed as the endocrine mechanism underlying the reorganization of attachment in middle childhood, and the implications for the relationship between attachment and sexual development are explored. Sex differences in the development of attachment can be fruitfully integrated within the broader framework of adaptive plasticity in life history strategies, thus contributing to a coherent evolutionary theory of human development. Keywords: adrenarche, attachment, cooperative breeding, evolution, life history theory, mating, middle childhood, phenotypic plasticity, reproductive strategies, sexual selection, stress

1. Introduction 1.1. Aim and scope

In this article, I present an integrated evolutionary model of the development of attachment and reproductive strategies in humans. The model is built on the foundations of life history theory, parental investment theory, and sexual selection; it aims to provide a significant update to current life history models of attachment formulated by Belsky and colleagues (Belsky 1997a; 1999; Belsky et al. 1991) and Chisholm (1999). In particular, the model I describe is the first to explain the development of sex differences in attachment patterns, permitting tighter integration between attachment theory, human reproductive ecology, and behavioral endocrinology. The gist of life history models of attachment (reviewed in sect. 5) is that infants and young children use their parent’s caregiving behavior as an indicator of the safeness and predictability of their local environment. Attachment security is the result of this unconscious evaluation process; the degree of security experienced in the first 5–7 years is hypothesized to set development on alternative developmental pathways, and to adaptively shape the individual’s future reproductive strategy. Secure attachment should lead to reproductive strategies based on late maturation, commitment in long-term relationships, and higher investment in parenting. Insecure attachment, on the other hand, # 2009 Cambridge University Press

0140-525X/09 $40.00

should lead to strategies based on early reproduction, short-term mating orientation, and lower parental investment in a larger number of children. My contribution extends the above-sketched theoretical framework by making a series of new points, which I briefly synthesize here. 1. Sex differences in attachment have adaptive significance. I argue that sex differences in attachment patterns arise as a result of asymmetries in parental investment and sexual selection, and that they are adaptive both in children and in adults. Previous theorists (e.g., Belsky 1999) have tried to make adaptive sense of the differences between avoidant and ambivalent attachment, but the link between attachment patterns in childhood and adult

MARCO DEL GIUDICE is an evolutionary developmental psychologist at the Center for Cognitive Science, University of Turin, Italy. Still at the beginning of his research career, he has published in leading psychology and biology journals, including Developmental Psychology, Developmental Science, and Evolution. He is especially interested in the dynamics of developmental plasticity, the evolution of human life history, and the origins of individual differences in behavioral strategies.

1

Del Giudice: Sex, attachment, and the development of reproductive strategies reproductive strategies is still poorly understood. I will show that taking sex differences into account makes it possible to reconcile individual differences in insecure attachment patterns with the concept of sex-specific reproductive strategies. 2. Sex differences in attachment arise in middle childhood. I present evidence that sex differences in attachment patterns are found not only in adults but also in children, starting from about 6 – 7 years of age (see also Del Giudice 2008). The available data suggest a phase of sex-biased reorganization of the attachment system in middle childhood, with a majority of insecure females shifting to ambivalent attachment and most insecure males shifting to avoidant attachment. 3. Sex-related endocrine mechanisms can influence the development of attachment patterns. Finally, I propose a hypothesis about the hormonal basis of the middle childhood transition in the organization of attachment. I review evidence showing that middle childhood is marked by intense, sex-related endocrine activity, and that the interplay between sexual maturation and attachment might be deeper and more bidirectional than is currently acknowledged. This view of attachment is also consistent with recent evolutionary models of the stress response system, suggesting intriguing avenues for cross-disciplinary research. 1.2. Overview of the target article

Because the idea of middle childhood as a transitional phase implies a degree of discontinuity in the development of attachment, I begin by sketching the issue of continuity versus change in attachment theory (sect. 2), and by linking it to the general biological problem of trait continuity across different life stages (sect. 3). Then, I summarize current evidence regarding sex differences in attachment, from infancy to adulthood (sect. 4). After reviewing extant life history models of attachment (sect. 5), I describe how sexual asymmetries in reproduction and sexual selection can be included in the picture to account for sex differences in reproductive strategies. I then discuss the resulting implications for attachment theory (sect. 6). Finally, I outline an updated synthesis of the development of human attachment and reproductive strategies, and explore the possible hormonal basis of the changes observed in middle childhood (sect. 7).

2. Continuity and change in attachment 2.1. Attachment as an evolved motivational system

Attachment theory, pioneered by John Bowlby (1969/ 1982; 1973; 1980), is to date the most comprehensive account of the nature and development of child– caregiver relationships. In addition, it embeds a complex theory of personality development, and has many implications for the study of social adjustment, emotion regulation, couple relationships, and psychopathology (see Cassidy & Shaver 1999, for an overview). According to attachment theory, infants are innately motivated to form selective emotional bonds with their caregivers, and organize their own behavior in order to seek and maintain proximity to them. Attachment is thus conceptualized as an innate 2

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

behavioral –motivational system, with the evolved function of protecting the child from danger while motivating the caregiver to provide for the child. Whereas the attachment system is a universal characteristic of human beings, there is much individual variation in the organization of actual attachment relationships. The systematic study of such variation started with the work of Ainsworth et al. (1978) and led to the concept of attachment patterns. Following early experience, infants adjust their care-eliciting behavior in order to maximize the caregiver’s availability; the resulting patterns – ranging from clingy, anxious care-seeking to apparently detached and distancing behaviors – are found cross-culturally (van IJzendoorn & Sagi 1999) and seem to represent the basic human ways of organizing parent – infant relationships (see Suomi 1999, for a description of similar patterns in other primates). Individual differences in attachment relationships have profound consequences for the child’s social and emotional development; hundreds of studies have been carried out to identify the causes of such differences, their developmental outcomes, and their mechanisms of change. 2.2. Patterns of attachment

A central notion in attachment theory is that relationships with caregivers become internalized as internal working models (IWMs), which are described as sets of beliefs and expectations about the self, the world, and relationships, together with rules for the direction of behavior and the appraisal of experience. IWMs guide the child’s interpersonal behavior, and are at the root of individual attachment patterns, or “styles” (see Ainsworth et al. 1978; Weinfield et al. 1999, for detailed descriptions). Children experiencing a consistently available, sensitive caregiver who is able to tune in to their states and feelings develop a secure attachment (labeled B); they use their caregiver as a “secure base” for exploration and, when distressed, turn to him or her for help and are easily comforted. In European and North American low-risk samples, the normative proportion of secure infants is about 65%, with remarkable consistency across different countries (see van IJzendoorn & Sagi 1999, for a review). Children who experience a rejecting, cold, and uninvolved caregiver establish an insecure – avoidant attachment pattern (labeled A): They treat the caregiver as unavailable, tend to avoid physical contact, and when distressed, don’t ask for help or comfort. They are adopting a “minimizing” or “deactivating” behavioral strategy, since signalling distress and need would lead to further rejection. On average, about 25% of infants in Western samples are classified as avoidantly attached, but proportions vary in different countries. If the caregiver is inconsistently available, alternating acceptance and rejection and being scarcely tuned to the child’s needs, the child is expected to develop an insecure –ambivalent/resistant attachment (labeled C). Ambivalent children are easily distressed and ask vigorously for help and comfort, but are not easily calmed and protest angrily in order to maintain closeness with the caregiver. Their attachment strategy can be described as “maximizing,” “hyperactivating,” and overdependent, since they exaggerate their signalling of need in order to control the caregiver’s behavior. The proportion of

Del Giudice: Sex, attachment, and the development of reproductive strategies ambivalent infants is about 10% on average, again with some cross-cultural variation. Sometimes, the child faces caregivers who are frightening or threatening in their parental behavior. Frightening behaviors can range from sudden, trance-like dissociative states, resulting from traumatic experiences or unresolved losses on the parent’s side, to downright physical or sexual abuse. Such caregiver’s behaviors tend to disrupt the child’s attachment strategy, leading to more or less severe forms of disorganization. Disorganized children (labeled D) may show elements of the previously described attachment strategies, but they experience abnormally high levels of motivational conflict, since the caregiver is simultaneously a source of comfort and fear. This results in conflicting approach/avoidance displays, dissociative states (e.g., “freezing”), and intrusion of sudden aggressive actions directed at the caregiver (Hesse & Main 2006; Lyons-Ruth & Jacobvitz 1999; Lyons-Ruth et al. 1999; Main & Hesse 1990). The proportion of disorganized children is highly variable across samples, and can range from 10% – 15% in low-risk families to 70% or more in extremely high-risk settings. Attachment patterns can be described either as categorical types, as I have done here, or as dimensional constructs. While many researchers rely on assessment procedures yielding categorical three-way (ABC) or fourway (ABCD) profiles, some have proposed that a better understanding of attachment dynamics is gained by assessing individual styles as combinations of underlying dimensions, such as high – low anxiety and high –low avoidance. Many researchers adopt some combination of the two methods; in particular, attachment security and disorganization are often described (and measured) as continuums rather than categories. I will not pursue the issue further here; for an overview of the ongoing debate, see Fraley and Spieker (2003) and related commentaries (Cassidy 2003; Cummings 2003; Sroufe 2003; Waters & Beauchaine 2003). 2.3. Longitudinal studies

Internal working models (and their corresponding behavioral patterns) are thought to be somewhat persistent and self-sustaining, but, at the same time, open to revision in the face of changing relational experiences (Bretherton & Munholland 1999). The question, then, is to what extent do IWMs persist (even across generations) rather than change or adjust to new conditions and life events. Of course, a detailed treatment of the issue is beyond the scope of this article; excellent reviews can be found in Grossmann et al. (1999), in the journal Child Development (2000, vol. 71), and in Grossmann et al. (2005). The growing consensus among attachment theorists is that stability in attachment is strongly tied to stability in caregiving conditions (e.g., Allen & Land 1999; Waters et al. 2000). Social stressors and negative life events (such as illness or death of relatives, changes in living arrangement, parental divorce, abuse, etc.) are associated with instability of attachment patterns from infancy to early adulthood; in particular, they lower stability by increasing the likelihood of shifting from secure to insecure attachment styles during development (see Hamilton 2000; Lewis et al. 2000; Waters et al. 2000; Weinfield et al. 2000). On the other hand, low-risk samples in relatively stable conditions

can yield high degrees of consistency between infant and adult attachment security (even in the 70% range; e.g., Waters et al. 2000). A classic three-generation study by Benoit and Parker (1994) provided an extreme example of stability, with 75% concordance between infants and their grandmothers on three-way attachment classifications. A general pattern seen in longitudinal studies is that attachment security is more stable and predictable than specific insecure strategies (e.g., avoidant or ambivalent) are. It is possible, then, that attachment security is at the “core” of lifelong IWMs, with specific A/C patterns providing a fine-tuned (and somewhat contingent) response to current caregiving style and ecological circumstances. This idea is pursued further in sections 6 and 7. Recently, Fraley (2002) performed the first meta-analysis of stability in attachment security from ages 1 to 21, and, in the same pioneering study, attempted to test two mathematical models of the underlying process of change. His results confirmed the association between psychosocial risk and stability: The overall correlations between security at age 1 year and subsequent ages were estimated at .48 for low-risk samples and .27 for highrisk samples (stability of specific attachment patterns was not assessed). Thus, this meta-analysis provided evidence of moderate stability, especially in low-risk samples; as discussed earlier, the lower stability associated with high-risk samples is not random, but reflects frequent shifts towards greater insecurity. Comparing the predictions derived from his mathematical models to the meta-analytic data, Fraley found support for a prototype model of stability, in which early security continues to influence security at later ages without being overridden; the model was tested against a so-called revisionist model, which instead assumed no persisting effect of early security. The model, of course, does not tell which factors are responsible for such prototype-like dynamics; likely candidates are early experience, strong continuity in rearing environment, and heritable genetic factors. Evidence from twin studies shows that attachment in infants and young children is mainly influenced by shared and non-shared environmental effects (note that nonshared environmental effects may also include genotype-environment interactions, and thus do not exclude broad-sense genetic influences on attachment stability), with no or little additive genetic contribution (Bakermans-Kranenburg et al. 2004; Bokhorst et al. 2003; O’Connor & Croft 2001; O’Connor et al. 2000; but see Finkel et al. 1998). In contrast, two studies with adult twins (one using the Adult Attachment Interview [AAI] and one using the Relationships Questionnaire [RQ]; see sect. 2.4) both found moderate heritability in attachment security and style (Brussoni et al. 2000; Torgersen et al. 2007). Thus, it seems that genetic factors may contribute to discontinuity rather than continuity in attachment, with additive genetic factors becoming more influent in adulthood. 2.4. The assessment of attachment from infancy to adulthood

An additional source of complexity in the study of attachment is that measures developed for a given age group typically cannot be employed at other ages. This has led to a variety of assessment methods, some based on BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

3

Del Giudice: Sex, attachment, and the development of reproductive strategies actual behavior, some on behavior representations (e.g., stories, doll play), and others still on interviews or questionnaires. All tend to retain the core construct of attachment security, but insecure styles may be categorized in different ways that are not always directly comparable. For in-depth reviews of the topic, I refer the reader to Solomon and George (1999), Crowell et al. (1999), and Kerns et al. (2005). As children grow up, the focus of measurement tends to shift from observed behavior towards behavior representations; some tasks are predominantly verbal, whereas others include “behavioral” aspects (e.g., doll-play procedures). Nevertheless, most childhood measures can be easily mapped onto the classic ABCD classification. From adolescence on, however, two distinct approaches to the assessment of attachment exist, and they differ substantially in scope and results.

end), whereas questionnaires are more predictive of mating outcomes, such as couple stability, satisfaction, and sexual behavior (e.g., Bernier & Dozier 2002; see further in the target article). The two aspects, of course, are not completely independent, and they show some overlap (especially on the security –insecurity dimension). There has been considerable debate on the relative merits of one approach over the other (e.g., Belsky 2002; George & West 1999; Shaver & Mikulincer 2002); in particular, questionnaire studies have been criticized because there was no evidence linking the attachment styles they measure to specific developmental antecedents (Belsky 2002).

2.4.1. Measures of adult attachment. The first approach is

Although seldom realized, the issue of attachment stability can be seen as an instance of a more general biological problem: that of continuity of phenotypic traits across different life stages. Attachment patterns are described as (relatively) coherent behavioral strategies, affecting not just the relationship with caregivers but also a wide range of developmental outcomes and processes, such as aggression, social competence, and emotion regulation (see Thompson 1999, for a review). As such, they are trait-like parts of the behavioral phenotype, and are clearly capable of affecting an individual’s biological fitness. As stressed by Belsky (1999), the ultimate fitness effects of a trait are to be understood in terms of reproduction, both of the individual itself and of its genetic relatives (the inclusive fitness concept; Hamilton 1964). However, early attachment theorists (e.g., Bowlby 1969/1982; Cassidy & Berlin 1994; Hinde 1982; Main 1981; 1990) have selectively emphasized the survival value of attachment (i.e., eliciting protection and parental investment from caregivers), even if survival is only an intermediate (and sometimes unnecessary) step towards evolutionary fitness. Life-history theory approaches, on the other hand, focus exactly on the reproductive consequences of attachment, and I review them in section 5. Before getting to reproduction, however, it will be useful to discuss the problem of trait stability in some detail.

based on interviews like the Adult Attachment Interview (AAI; see Main & Goldwyn 1998). These interviews do not assess present attachment behavior, rather focusing on the “mental state” with respect to past attachment experiences, inferred by discourse analysis. The AAI categories (“free,” “dismissing,” “entangled,” and “unresolved”) refer to how the person relates to his or her own past experiences with parents, not to the way he or she behaves with present attachment figures (see Hesse 1999). Adult AAI categories are reliably associated with the attachment classification of sons and daughters (e.g., entangled parents tend to have ambivalently attached children; see Belsky 2005, for a review). The second approach, often referred to as “social psychological,” is based on self-report questionnaires and is mostly employed in research on romantic (couple) attachment. Compared with interviews, most questionnaires are conceptually closer to childhood measures, because (1) they focus on present behavior and feelings towards romantic partners, and (2) their classification of insecure attachment is modeled on the avoidant and ambivalent patterns of infancy. Analysis of many selfreport attachment questionnaires reveals two robust dimensions underlying romantic attachment patterns, labeled avoidance and anxiety (Brennan et al. 1998). Secure adults (low avoidance, low anxiety) feel it easy to get emotionally close to others, feel comfortable depending on someone else, and do not worry much about rejection. Dismissing-avoidant adults (high avoidance, low anxiety) are distancing with their partners, show a low need for intimacy and closeness, and describe themselves as self-sufficient. Preoccupied adults (low avoidance, high anxiety) report intense desire for closeness, feel uncomfortable when not being involved in close relationships, and worry about partner’s rejection. Finally, fearful-avoidant adults (high avoidance, high anxiety) show a mix of desire for closeness and fear of rejection, and they report feeling uncomfortable in depending on others.1 Interviews and questionnaires show only low to moderate correlations with one another, usually below r ¼ .30 (Crowell et al. 1999; Roisman et al. 2007; Shaver et al. 2000); in addition, they seem to predict somewhat different outcomes. Roughly stated, interviews are most powerful at predicting parenting outcomes such as children’s security (and indeed have been originally devised to this 4

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

3. The general problem: Trait continuity across life stages

3.1. Discontinuity across life stages

Psychologists often assume that development is an essentially cumulative process, in which each stage builds on the preceding ones, and previous characteristics (especially in the behavioral domain) have a natural tendency to persist unless actively modified. Even if this makes intuitive sense, it is important to realize that, from the point of view of natural selection, such continuity is neither necessary nor always useful (see also Bjorklund 1997; Geary & Bjorklund 2000, for a general introduction to this topic). In many species, development involves dramatic alterations in shape and behavior, as, for example, the metamorphosis process that turns tadpoles into frogs; furthermore, many developmental transitions involve the loss or disposal of previous phenotypic characters (such as the tadpole’s tail). Although humans do not undergo such radical metamorphoses as frogs, a careful look at human development does reveal many subtler examples of the same principle,

Del Giudice: Sex, attachment, and the development of reproductive strategies both in morphology (e.g., detachment of the placenta, replacement of milk teeth, loss of brown fat in adults) and in behavior (e.g., loss of neonatal reflexes, abandonment of quadruped locomotion). The key to understanding such apparent exceptions to the cumulative nature of development is to look at developmental traits (morphological as well as behavioral) from a fitness perspective. In order to be selected for, traits need to solve two problems: being adaptive at the present time and being adaptive in the future of the organism. Sometimes, the solution of the puzzle is to build “disposable” traits, or ontogenetic adaptations (Bjorklund 1997), which are only adaptive during certain developmental stages and are replaced or modified when necessary. In this way, development becomes “modularized,” and selection can act independently on different life stages (see Wilkins 2002; West-Eberhard 2003, for the concept of modularity in development). In altricial mammals like humans (which are born immature and undergo an extended period of parental care), some infantile traits could be selected for because they are adaptive in the context of parental care; on the other hand, the same traits might become useless, or even maladaptive, when the individual becomes independent (see also Lynch [1987] for a genetic approach to the same problem). Selection is expected to act on traits such as these by rendering them transient (i.e., disposable), so that aspects of the phenotype that are no longer necessary are lost, or replaced, during maturation. 3.2. Continuity across life stages

At the same time, continuity is a major feature of development. There are many reasons for this, including continuity of environment and ecology during growth and the costs involved in switching and reshaping phenotypes (Bateson 2005; Boyce & Ellis 2005; Ellis et al. 2006). Yet another powerful source of continuity in development, even across modularized life stages, is that it is often adaptive for an organism to rely on early outcomes to make strategic decisions about the next developmental phases it will face. A classic illustrative example comes from male dung beetles, whose development involves a neat binary switch between two alternative phenotypes (or “morphs”). The nutritional condition of a beetle’s larva, determined by maternal food supply, is strongly predictive of the beetle’s adult body size; body size, in turn, determines whether the individual is to develop horns (and fighting behavior) or not. The whole process is orchestrated by hormonal mechanisms. As a result, there are two kinds of males in the population: those who can afford the metabolic expense of growing horns and fighting, and those who are better off if they “decide” in time to adopt a less costly developmental strategy, together with different reproductive behaviors (Emlen 1997; West-Eberhard 2003). In this sense, previous development provides the organism with useful information, which can be used to direct the next phases in an adaptive way. Sometimes it is possible to identify developmental “switch points” between alternative pathways (see also Hagen & Hammerstein 2005), while at other times the process looks more gradual. What is important to keep in mind is that stability in phenotypic traits is not to be taken for granted, and

must always be weighted against developmental tradeoffs between present and future contributions to reproductive success. 3.3. Parent –offspring conflict and the adaptive value of childhood traits

When the environment in which selection takes place includes genetic relatives of the developing organism, additional issues arise. Most relevant for the present discussion is the concept of parent –offspring conflict (Mock & Parker 1997; Parker et al. 2002; Trivers 1974), which is the conflict of interest between parents and offspring about the amount of investment (e.g., energy, time, food) to be provided in parental care. Parent – offspring conflict follows from the fact that, while an offspring is perfectly related to itself, its relatedness coefficient with siblings (i.e., the probability of sharing an allele by common descent) is only 0.5. Although a parent optimizes its inclusive fitness by investing the same amount of resources in each offspring (all else being equal), a single offspring maximizes its own fitness by requiring a higher amount for itself, as the benefits enjoyed by siblings must be discounted by their relatedness coefficient. The bottom line is that costs and benefits of a given amount of parental investment will not affect the fitness of parents and offspring in the same way. This concept, originally formulated to explain patterns of parental investment, can actually be extended to a much more general principle: Parents and offspring will value differently every developmental outcome (including those unrelated to parenting), provided that benefits gained by one side translate into fitness costs to the other, even indirectly. Trivers (1974), for example, suggested that parents and offspring can disagree about offspring’s degree of altruism (towards both kin and nonrelatives), mate choice, and reproductive effort. Following this line of reasoning, Trivers (1985) suggested a non-obvious implication of the theory. He suggested that offspring should not allow themselves to be permanently influenced by parental behavior, as the genetic interest of parents ultimately differs from their own. Referring to human development, he speculated that “compliance” with parental influence should last until the end of dependency, and then be erased during puberty through a sort of personality reorganization. In this view, childhood personality traits influenced by parents (and attachment patterns certainly fall into this category) are exactly the kind of disposable behavioral phenotypes described above; they are adaptive in the limited context of parental care, but need to be modified or replaced in the transition to adulthood. The idea is quite powerful, and it has been reprised by critics of “family socialization” theories of personality development such as Harris (1995; 2005) and Pinker (1997) to argue that parents should not be expected to permanently shape their children’s personality. However, there are a number of biological reasons to doubt a “black-and white” approach, and to predict a more balanced mix of continuity and discontinuity. First, it is true that genetic interests of parents and offspring differ, but there is still quite a lot of overlap, so that a certain degree of parental “shaping” can be expected. Second, the conflict hypothesis only applies to cases in which parental influence involves BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

5

Del Giudice: Sex, attachment, and the development of reproductive strategies costs (or benefits) on the parents’ side; cost-free parental influence is not expected to lead to this kind of conflict. Third, parental behavior can sometimes provide offspring with indirect information about the state of the local environment; in other words, offspring may use parental behavior as a proxy for external ecological conditions, even independently from the parent’s willingness to provide such information (see Bateson 2005; Chisholm 1993; Draper & Harpending 1982; Ellis et al. 2006). Life history models of attachment, which predict strong continuity between attachment patterns in childhood and adult behavior, are essentially based on the latter assumption: Because parental behavior carries useful information about the local ecology, children can be responsive to their rearing environment to the point of basing their adult reproductive strategy on early attachment experience. However, it doesn’t follow that a child becoming an adult should employ the same behavioral strategy which proved useful with parents in his or her early childhood. One reason is that, in humans, the attachment system is not just a care-eliciting mechanism for the young – it has also been recruited by evolution to serve as a powerful pair-bonding device in the mating couple. 3.4. The double life of human attachment

The “double life” of the attachment system, as a care-eliciting and pair-bonding device, is a central theme of the present discussion. Attachment theorists have realized from the start that infant – caregiver bonding and couple relationships share many key features, so that adult love can be properly characterized as involving an attachment dimension (in fact, intimate friendship may also be characterized as attachment relationships, so that human attachment can be said to have “multiple lives”; see, e.g., Sibley & Overall 2008). Research has shown that the dynamics of bond formation, separation, and loss in adults are strikingly similar to those observed in infants (for reviews, see Feeney 1999; Hazan & Zeifman 1999). Neurobiological studies also suggest that the neurochemical/neuroanatomical substrates involved are largely overlapping (see Carter 1998; Insel 2000; Insel & Young 2001; Leckman et al. 2005; Panksepp 1998; Pedersen et al. 2005). Similarities notwithstanding, the two processes are not identical (see Simpson 1994), and they are subject to different selective pressures. In particular, as I discuss in depth in section 6, attachment-related traits are expected to show sex differences in adults but not in young children, so that (for example) a detached, uncommitted, lowinvestment relationship style would usually be more advantageous to men than to women (e.g., Kirkpatrick 1998). Behavioral correlates of attachment patterns (e.g., dependency, aggression) would also have quite different fitness consequences in infancy and in adulthood if they happened to be involved in mate choice or sexual competition. For example, fearfulness and overdependency (related to ambivalent attachment) are likely to be equally adaptive for males and females in infancy, when they only affect the regulation of parental care. In adults, however, the balance could shift dramatically: If, for example, fearful/overdependent males (but not females) were less desirable as partners, and/or less able to compete with other males for status, the fitness consequences of attachment would no longer be neutral with 6

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

respect to sex. It is wise, then, to ask whether sex differences in attachment have been found, what they are, and when they appear in the course of development. 4. Sex differences in attachment 4.1. Infancy and early childhood

The first decades of attachment research were characterized by the almost complete absence of reported sex differences in attachment security and style. This was due to a prevailing focus on infants and preschoolers, who usually do not show sex differences in attachment (e.g., van IJzendoorn 2000). Studies with children as old as 6 years usually find a comparable proportion of avoidant and ambivalent children in both sexes (e.g., Moss et al. 1998). The only exceptions were a few studies with high-risk samples, in which boys were found to be more frequently and/or more severely disorganized than girls (Carlson et al. 1989; Lyons-Ruth et al. 1999). Moreover, Turner (1991) found some behavioral differences in peer relationships between insecurely attached 4-year-old boys and girls: Insecure males were more aggressive and attentionseeking, whereas insecure females were more compliant, dependent, and affiliative. Another report of small sex-related effects came from the meta-analysis by van IJzendoorn (2000), who found that siblings of the same sex were more likely to be both secure or both insecure, compared with mixed-sex pairs. More recently, David and Lyons-Ruth (2005) reconsidered sex differences in disorganization in the light of sexually dimorphic responses to stress and threat (“tend-andbefriend” versus “fight-or-flight”; Taylor et al. 2000; more on this in sect. 7.2.2). In a low-income infant sample, the researchers found different behavioral patterns in males and females consistent with the “tend-and-befriend” hypothesis: specifically, females responded with more approach displays than males when faced with frightening or threatening maternal behaviors. 4.2. Middle and late childhood

The picture changes dramatically when one considers middle childhood. To my knowledge, nearly all of the studies in which sex was examined have revealed significant biases in the avoidance-ambivalence dimension. This result holds across nations (to date: USA, Canada, Italy, Israel, and perhaps Hungary) and across assessment methods (questionnaires vs. doll-play procedures). Granot and Mayseless (2001) performed a study on 113 Israeli children aged 9 to 11 years, with a doll-play task (an adapted version of the Doll Story Completion Task by Bretherton et al. 1990). The study focused on the relationship between attachment and school adjustment. Results showed an unanticipated sex difference: boys and girls differed significantly in their prevailing insecure patterns, with girls more often ambivalent than avoidant (18% C to 7% A on the total number of girls), and all of the insecure-organized boys classified as avoidant (27% A to 0% C on the total number of boys). I found the same effect in an Italian sample of 122 7-year-old children (Del Giudice 2008), using a different doll-play task, the Manchester Child Attachment Story Task (MCAST; Green et al. 2000). Almost all insecure

Del Giudice: Sex, attachment, and the development of reproductive strategies boys were classified as avoidant (27% A to 2% C), while insecure girls were mostly ambivalent (25% C to 4% A). The similarity of attachment distributions between Italy and Israel is even more striking since infant studies in Israel, but not in Italy, usually find a high proportion of ambivalent (17–37%) and a very low proportion of avoidant (0–7%) patterns (Harel & Scher 2003; Sagi et al. 1985; 1994; van IJzendoorn & Sagi 1999). Sex effects were also apparent in the distribution of secure subtypes; boys were more often classified as secure/avoidant than secure/ambivalent (46% B/A vs. 21% B/C), and girls showed the opposite pattern (32% B/C vs. 16% B/A). However, this effect was much weaker than that observed in insecure children. In the same study, I found that boys tended to get higher disorganization scores than girls, thus confirming the findings obtained with younger children by Carlson et al. (1989) and Lyons-Ruth et al. (1999). Toth et al. (2006) used the MCAST in a Hungarian sample of 84 six-year-olds. Although the sample was somewhat younger than the others cited here, their results seem to show a smaller effect in the same direction (I. Toth, personal communication, October 19, 2007). The proportion of ambivalent and avoidant girls was the same (6%), whereas in males, the proportion of avoidant children (14%) was higher than that of ambivalent ones (2%). Unfortunately, the very low frequency of non-D insecure patterns in this sample (14% overall) makes statistical comparisons uninformative (in contrast, disorganization was significantly more frequent in boys than in girls: 47% vs. 20%). Marked sex differences in middle childhood were also found in three studies using a self-report questionnaire on attachment behaviors, the Coping Strategies Questionnaire (CSQ). The first study was performed in the United States by Finnegan et al. (1996), with a sample of 229 children aged 8 –13 years. In this study, boys reported significantly higher scores of avoidant coping, whereas girls reported more preoccupied coping. The authors noted this association with sex and attributed it to gender stereotyping. Similar results were obtained in a Canadian study (Karavasilis et al. 2003), which investigated the relation between parenting and attachment to mother in a sample of 202 children aged 9 – 11 years. Boys reported more avoidant coping, while girls reported more preoccupied coping at the CSQ; both associations were statistically significant and of remarkable size. In another US study, Corby (2006) administered an expanded version of the CSQ to 199 children aged 8 –14 years (mean age: 11). Again, she found significantly higher avoidance scores in boys and higher preoccupation scores in girls. The only contrasting result so far comes from a recent study in the United States by Kerns et al. (2007), in which the doll-play procedure used by Granot and Mayseless (2001) was administered to a sample of 52 children aged 9 – 11 years. In this study, (K.A. Kerns, personal communication, December 12, 2007), females were more likely to be classified as avoidant than ambivalent (35% A vs. 4% C); the same was true for males, to a lesser degree (19% A vs. 4% C). Boys were more often classified as disorganized (42% of boys vs. 4% of girls). This is the only study which departed from the overall pattern, at least for females; note, however, that sample size was substantially smaller compared to the other studies.

There are three reasons for the relatively small number of relevant studies in this age group. First, the lack of ageappropriate measures and tasks has led attachment researchers to neglect middle childhood until recently, so the sheer number of studies in this age range is much smaller than in infants or adults (Kerns et al. 2000; 2005). Second, attachment studies in middle childhood often focus solely on the security – insecurity dimension, without assessing avoidant/ambivalent insecure styles. Third, many researchers still omit reporting and analyzing their data by sex, probably based on the tacit assumption that sex differences in children’s attachment patterns do not exist. Hopefully, the accumulating evidence for strong sex effects in this age group will prompt more researchers to include this variable in their studies. 4.3. Adolescence and adulthood

When examining sex differences in adult attachment, the issue of measurement methods (interviews vs. questionnaires) becomes crucial. The first surveys of adult attachment styles were based on the AAI, and consistently failed to reveal any sex difference (e.g., van IJzendoorn & Bakermans-Kranenburg 1996). The same seemed to happen, at first, with questionnaire-based measures: indeed, most early studies failed to find significant sex differences in styles of romantic attachment (e.g., Collins & Read 1990; Feeney & Noller 1990; Hazan & Shaver 1987). However, early self-report attachment measures had a categorical response format and very low reliability (Baldwin & Fehr 1995). Newer studies, employing continuous ratings, soon began to find sex effects on attachment selfreports: notably, men (on average) have higher avoidance scores and lower anxiety scores than women, or (depending on the instrument) rate themselves as more dismissing (e.g., Bartholomew & Horowitz 1991; Brassard et al. 2007; Brennan et al. 1998; Kirkpatrick 1998; Picardi et al. 2002; Scharfe & Bartholomew 1994). Not all questionnaire studies found sex differences, however (e.g., Gentzler & Kerns 2004; Jang et al. 2002). Questionnaire studies with adults often find smaller sex differences than those reported in middle childhood, especially compared to those found with doll-play procedures. Part of this effect may depend on the lower accuracy of self-reports compared with experimenter-coded measures (distinct from psychometric reliability, which is usually high). There is, however, a more interesting explanation: When age is taken into account, it becomes apparent that sex differences are stronger in young adulthood and decline markedly approaching middle age. In a large Italian validation sample for the ECR (Experiences in Close Relationships; Brennan et al. 1998), for example, standardized sex differences in the anxiety dimension were d ¼ .57 at 18 –20 years, d ¼ .48 at 21– 35 years, and d ¼ 2.02 at 36 –65 years (Picardi et al. 2002). The same age-related decline of sex differences was apparent in the cross-cultural study by Schmitt et al. (2003a; see following). Thus, depending on participants’ age, the size of sex differences can vary considerably. I discuss the relevance of this finding in section 7. Schmitt et al. (2003a) performed a cross-cultural study of adult attachment in 62 cultural regions employing the Relationships Questionnaire (RQ; Bartholomew & Horowitz 1991). While male and female dismissiveness scores BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

7

Del Giudice: Sex, attachment, and the development of reproductive strategies were highly correlated across samples (r ¼ .87), males described themselves as more dismissing than women in most countries, with smaller differences in Oceania and East Asia, and no significant difference in Africa. The overall effect size was d ¼ .18. There was substantial cultural variation in the degree of sex differences (d ranging from 2.26 to .43), and it was mostly driven by female dismissiveness scores. Smaller sex differences were related to higher dismissiveness (especially in females), and both were predicted by high levels of mortality, high fertility, and high AIDS rates, but not by indexes of gender inequality and cultural stereotypes. Overall, cultural stereotypes do not seem to explain much variation in sex differences, which appear to be more closely related to levels of environmental stress: where mortality and fertility are high, women (and, to a lower extent, men) are more dismissing, and sex differences tend to become smaller. Mean age in the 62 samples ranged from 19 to 38 years (median, 22 years; reported in Schmitt et al. 2004). I correlated the mean age of each sample with the corresponding effect size d; Pearson’s correlation was negative and significant (r ¼ 2.30, p ¼ .016, N ¼ 62), showing that sex differences in dismissiveness get smaller with increasing age. (Of course, longitudinal data would be necessary to rule out cohort effects.) 5. Attachment and life history theory 5.1. Life history strategies

Life history theory (see Hill 1993; Kaplan & Gangestad 2005; McNamara & Houston 1996; Roff 2002) is a branch of theoretical evolutionary biology, dealing with the trade-offs in the allocation of time and resources over an organism’s life span. The starting point of life history theory is that time and resources are inherently limited, so organisms have to make decisions about how to invest them to optimize their fitness. The way resources are allocated constitutes the organism’s life history strategy. Different ecological constraints will result in different optimal strategies, both at the between-species and at the within-species level. The basic trade-off in life history theory is that of somatic effort versus reproductive effort. The former is defined as resources devoted to growth during development and maintenance during adulthood; it also includes the accumulation of resources that augment the reproductive potential. The latter is typical of mature stages and is distributed between mating effort (resources invested to attract mates, increasing opportunities for reproduction) and parenting effort (resources invested in raising already-conceived offspring). The balance between mating and parenting is another key trade-off in life history strategies. Another way to conceptualize life histories is by considering two reproduction-related trade-offs: between current versus future reproduction and between quality versus quantity of offspring (for an introduction, see Chisholm 1993; Hill 1993; Pennington & Harpending 1988). The optimal solution of these trade-offs is related to the ecological pattern of extrinsic mortality, that is, mortality that cannot be prevented or diminished by altering the organism’s behavior. Predation, pathogens, and warfare are usually considered extrinsic sources of mortality; but, more generally, all factors that negatively affect reproductive success independent on the organism’s decisions can be considered sources 8

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

of extrinsic risk (Quinlan 2007). When adult mortality is high, it is adaptive to favor current reproduction by starting mating early, even at a cost for one’s future reproductive potential (costs may arise, for example, because waiting longer gives higher benefits to offspring, or because reproducing earlier increases parent’s mortality). When juvenile mortality is high, it pays for parents to avoid the risk of lineage extinction by producing more offspring and investing fewer resources in each (Promislow & Harvey 1990; 1991). In addition, high extrinsic risk means that investing in parental care has quickly diminishing returns: Since (by definition) parental effort cannot decrease extrinsic risk, offspring’s fitness will not respond to parental care beyond a certain amount (the “saturation point”; see Fig. 1). Thus, elevated environmental risk favors quantity versus quality of offspring and current versus future reproduction, and selects for life histories that invest in mating at the expense of parenting (Chisholm 1993; Pennington & Harpending 1988; Quinlan 2007). 5.1.1. The human life history. When compared with other

species, humans show many traits characteristic of an extremely “slow” life history strategy. We reproduce late, and pass through a prolonged stage of reproductive immaturity which has no equivalent in other primates (Ellison 2001; Flinn & Ward 2005). Moreover, we invest considerable time and effort in parental care, which involves a lot of teaching and social training in addition to mere energetic investment (Bjorklund & Rosenberg 2005; Geary & Flinn 2001; Hewlett et al. 2000). As an exception to this pattern, humans show high fertility compared with their close primate relatives, with inter-birth intervals of about 2.5– 3.5 years. The peculiar pattern of human life history traits can been explained by the coevolution of a bigger brain, extended skill learning and slow growth rate in childhood, longevity, and skill-intensive foraging practices such as hunting and complex food processing. This suite of characters leads to a unique combination of high fertility and slow development, obtained through massive intergenerational transfer of resources (see Gurven & Walker 2006; Kaplan et al. 2000; Kaplan & Robson 2002). In addition to the need for extended learning of foraging

Figure 1. Effect of extrinsic risk on function of parental effort (adapted Increased extrinsic risk (dashed line) point of parental effort (i.e., the point at does not increase offspring fitness).

offspring fitness as a from Quinlan 2007). lowers the saturation which additional effort

Del Giudice: Sex, attachment, and the development of reproductive strategies abilities, the social complexity of human coalitions (and of foraging practices themselves) is thought to have further increased selective pressures for bigger brain and slower development (for an overview, see Dunbar & Schultz 2007). 5.1.2. Adaptive plasticity. As discussed earlier, humans as a species have a recognizable life history strategy, and show a distinctive pattern of life history traits. However, as in most species, there is also room for substantial variation between individuals. While some of this variation (e.g., in timing of maturation and reproduction) is heritable, organisms are also expected to embody mechanisms that evaluate the current (and expected) state of the environment and adjust their life history traits accordingly. In other words, life histories show adaptive plasticity. Mathematical models clearly show that the concept of a single “best” strategy is an illusion: what is expected (and found) is a variety of strategies, contingent on local conditions. The best strategy in a safe, predictable environment does not work well in a threatening and unpredictable one; the aim of maximizing long-term fitness can be targeted effectively only by organisms capable of context-sensitive (or state-dependent) adjustment of life history decisions (Houston & McNamara 1999; McNamara & Houston 1996). As Chisholm (1999) puts it, in the realm of life histories, “contingency rules.” The study of context sensitivity in life history decisions has always been one of the key research topics in evolutionary anthropology (e.g., Blurton Jones 1989; Borgerhoff Mulder 1989; Hill & Kaplan 1988; Low 2000; Mace 2000a). The key assumption of life history models of attachment is that, in humans, attachment relationships in infancy and early childhood (the first 5–7 years) provide the child with crucial information about the safety and predictability of his/her local environment. In turn, childhood attachment patterns are thought to translate into different reproductive strategies,2 involving different trade-offs between current and future reproductive investment, and between mating and parenting effort. Of course, there are many other factors involved in the development of relational and sexual styles, including heritable dispositions, attractiveness, cultural practices, and the local sex ratio. The link between environmental stress, attachment, and adult reproductive strategy is thus expected to be only probabilistic (Gangestad & Simpson 2000; Schmitt 2005a; see also sect. 6.4 for a more detailed discussion). 5.2. The Belsky, Steinberg, and Draper model

The first systematic attempt to reframe attachment theory from a life history perspective was made by Belsky, Steinberg, and Draper (Belsky et al. 1991), drawing on previous work by Draper and Harpending (1982) on the effects of father absence on children’s behavioral development. Belsky et al. noted that, in stressful conditions, parenting style becomes harsher and less sensitive and marital discord increases, causing the child to experience chronic psychosocial stress and leading to insecure attachment patterns. Insecure children thus receive crucial (albeit indirect) information about their environment: that resources are scarce and unpredictable, that people cannot be trusted, and that mating relationships tend to be short and

uncommitted. This should switch development towards a reproductive style based on opportunistic interpersonal orientation, early reproduction, and low parental investment (offspring quantity vs. quality). Secure attachment/low stress, on the other hand, should lead to delayed mating, high parental investment, and a trusting and reciprocally oriented attitude. The reproductive strategies following secure versus insecure attachment are thought to be implemented by a suite of covarying traits, both behavioral/psychological (e.g., interpersonal orientation, sexual style) and somatic (e.g., accelerated sexual maturation). The result would be an adaptive polymorphism, based on condition-sensitive, developmentally contingent variation in life-history-related traits. The theory predicted that early relational stress (and, by extension, insecure attachment) would relate to earlier maturation, earlier age of intercourse, and a tendency to entertain short-term relationships with mates. Although there is as yet no longitudinal study using attachment security to predict later maturation and sexual style, the evidence on the effects of relational stress and parenting consistently supports the main predictions made by Belsky et al. (1991). Note, however, that the weight of heritable genetic effects in linking mating and parenting across generations has yet to be fully evaluated; for example, age at menarche is known to be substantially heritable, at least in industrialized societies (current estimates are in the .40 to .50 range; see Campbell & Udry 1995; Chasiotis et al. 1998; Comings et al. 2002; Kirk et al. 2001; Moffitt et al. 1992; Rowe 2000a; Treloar & Martin 1990). For reviews of the evidence on early stress and accelerated sexual maturation, see Chisholm et al. (2005a) and Ellis (2004; 2005). See also Chisholm et al. (2005b) for recent data on early first birth related to insecure attachment. Recently, two longitudinal studies further confirmed the effects of parent –child relationships on sexual maturation. Ellis and Essex (2007) found that low-quality investment and marital conflict predicted earlier pubertal development in girls, and earlier onset of adrenarche in both sexes (see sect. 7.2). Belsky et al. (2007b) found that negative parenting predicted earlier pubertal development, but only in girls. They also found a moderating effect of early temperament, with infants low in negative emotionality showing the opposite pattern (i.e., negative parenting predicted later development; the meaning of this finding is still unclear). The first version of the Belsky, Steinberg, and Draper theory had two main limitations. First, it treated insecure attachment as a whole, without distinction between avoidant and ambivalent strategies. Second, it assumed that the same reproductive strategy would be optimal for both males and females – an assumption that was immediately criticized by Maccoby (1991). The issue of sex differences has never been fully addressed by the theory (as noted also by Simpson 1999), perhaps because of the lack of sex-related differences in published attachment research. On the other hand, Belsky (1999) provided an updated version of the model, specifically addressing the issue of possible differences between ambivalent and avoidant strategies. Belsky (1997a; 1999) argued that his original analysis (predicting low-investment, short-term mating) was in fact more relevant to avoidant attachment, which is associated with parental rejection and high-risk, unpredictable environments. Similarly, BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

9

Del Giudice: Sex, attachment, and the development of reproductive strategies Kirkpatrick (1998) conceptualized adult dismissiveness as a male-biased, short-term reproductive strategy. The hypothesized link between avoidance and short-term mating style is well supported by research in adult attachment. Dismissing adults are consistently found to be more promiscuous and sexually unrestrained; they are less committed in romantic relationships, are more likely to be sexually coercive, and tend to avoid intimacy with partners (reviewed in Allen & Baucom 2004; Belsky 1999; Brassard et al. 2007; Feeney 1999; Gentzler & Kerns 2004). In apparent contrast with these data, Cooper, Shaver and Collins (1998) found that avoidant adolescents of both sexes were less likely than secure ones to ever have had intercourse. However, the categorical measure they employed is likely to have misclassified as avoidant many fearful adolescents, high in both anxiety and avoidance (see Bartholomew 1990). This possibility is consistent with the study by Gentzler and Kerns (2004), who found high avoidance and moderately high anxiety both in students reporting no intercourse and in those reporting early intercourse (before 15 years). Finally, Bogaert and Sadava (2002) found that avoidance and anxiety both correlated with earlier age of first intercourse in women, but not in men. In contrast with avoidant attachment (associated with a behavioral profile of self-reliance, pseudo-maturity, and aggression with peers in childhood), the ambivalent pattern is characterized by dependency, exaggeration of need signalling, behavioral immaturity, and passivity in peer relations. Belsky (1997a; 1999) speculated that ambivalent attachment was likely to develop in a form of delayed, indirect reproductive strategy. He hypothesized that ambivalent children would tend not to reach autonomy from the family, rather becoming their parents’ “helpers-at-the-nest,” and enhancing their own inclusive fitness by helping to raise younger siblings. Belsky then suggested that, for this reason, ambivalent children should often be first-borns. Interestingly, a study by Vondra et al. (1999) lent some empirical support to this specific prediction: Infants who became ambivalent at 18 months were more likely to be firstborn (and male), whereas avoidant children were more often later-born. To date, however, there is very limited support for the “helper” hypothesis; the sexual style associated with adult anxious attachment is also more complex than the avoidant one. Preoccupied individuals eagerly look for intimate relationships, and (if males) describe themselves as less accepting of casual sex (see Allen & Baucom 2004; Belsky 1999; Brassard et al. 2007; Feeney 1999; Gentzler & Kerns 2004); however, they also experience intense and impulsive sexual attractions (Hazan & Shaver 1987) and report higher frequency of infidelity (Bogaert & Sadava 2002). They are also more likely than secures to engage in unwanted sex following relational pressures and in order to keep their partner close (Impett & Peplau 2002; Schachner & Shaver 2002). This behavioral pattern, though different from that of dismissing adults, can nevertheless result in early onset of reproduction and in a relational style characterized by multiple matings, especially in women. Indeed, attachment-anxious women (but not men) are likely to initiate sexual activity earlier in adolescence (Bogaert & Sadava 2002; Cooper et al. 1998; Gentzler & Kerns 2004); and in a study of sexual fantasies, it was found that, in women, attachment 10

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

anxiety predicts both more “romance” and more “unrestricted-emotionless sex” content (Birnbaum 2007). In a recent study by Jackson and Kirkpatrick (2007), anxiety showed a weak negative correlation with short-term mating orientation, and virtually no correlation with long-term orientation; in contrast, avoidance was weakly and positively related to short-term orientation, but strongly and negatively related to long-term orientation. 5.3. The Chisholm model

Chisholm (1993; 1996; 1999), drawing on the model of Belsky et al., proposed a complex theory of conditiondependent reproductive development based on attachment security. In Chisholm’s model, attachment experiences provide children with information about the availability, sensitivity, and responsivity of their future social relations (a socioassay of their local environment). In a narrower sense, Chisholm (1993) proposed that caregiving conditions are used as a proxy for a critical life history parameter: the local mortality rate. This should lead to (unconscious) estimates of one’s own expected lifespan and of the probability that one’s offspring will survive, orienting towards earlier reproduction and higher mating efforts in the case of a risky environment. In addition, reproducing at a younger age means having fewer social resources (e.g., status, support networks) and, as a consequence, being less able to “make a difference” in offspring’s quality; thus, early reproduction would further push reproductive strategies towards offspring quantity (vs. quality) and high reproductive rate. Intriguingly, there is evidence that insecurely attached adults tend to make shorter estimates of their own life expectancy (see Chisholm et al. 2005b), and that local mortality rates do correlate with familial environments and fertility patterns (Bereczkei & Csanaky 2001). In Chisholm’s words, parents act as “vectors” through which the risk and uncertainty of the environment is transferred to children. Recent cross-cultural analyses by Quinlan (2007; Quinlan & Quinlan 2007b) strongly support the link between extrinsic risk factors (famine, warfare, and pathogens) and lower parental (especially maternal) investment. Furthermore, they suggest that unresponsive parenting and its developmental consequences (matingoriented, risk-taking strategies) can influence cultural patterns related to casual sex, aggression, theft, and social hostility, thus contributing to a self-sustaining “culture of risk”. While it is predicted that both avoidant and ambivalent children will adopt life history strategies maximizing current reproductive effort, the two attachment strategies are thought to respond to different safety threats. In the case of parents who are willing, but unable to consistently invest in offspring, for example, because of scarce resources or competing demands on parents’ time, the ambivalent strategy maximizes the available investment by increasing signals of need and behaving immaturely. When parents are unwilling to invest, however, the avoidant strategy is favored, pushing towards self-reliance and protecting the child from being abandoned or abused (Chisholm 1996). When insecure children grow up, they are expected to engage in low-commitment mating and low-investment parenting. Based on sexual selection theory (see further on), the model then describes two sex-specific developmental

Del Giudice: Sex, attachment, and the development of reproductive strategies pathways leading to maximization of current reproduction (Chisholm 1999). Males growing in high-risk environments should adopt a strategy based on increased sex drive, aggression, impulsivity, and risk-taking, given the evocative label of Young Male Syndrome (from Wilson & Daly 1985). Insecure females should mature quickly as well, and their strategy should be characterized by impulsive mate choice (based on mate’s genetic quality and immediate benefits), early and frequent childbearing, and single motherhood: the Young Female Syndrome. Of course, the two strategies represent the ends of a graded continuum, rather than being all-or-none choices. What, then, about adult attachment styles? In Chisholm’s model, ambivalent and avoidant insecure patterns are differently tuned to safety threats in the caregiving environment, but have no special role after reaching reproductive maturity. It would then make sense to think of attachment styles as ontogenetic adaptations – disposable phenotypes which have no reproductive value outside the caregiving environment. This contrasts with the empirical observation that insecure adults differ considerably in attachment styles, with measurable consequences for sexual and caretaking behavior. This is the main point of divergence from Belsky’s revised theory; however, neither model succeeds in fully taking into account the different adaptive consequences of a given attachment pattern for males and females. 5.4. Some theoretical refinements 5.4.1. Environmental stability and “child development theory.” A somewhat problematic assumption in life

history models of attachment is that of substantial environmental stability in the time span from early childhood to puberty. In fact, only if ecological conditions are relatively stable is it adaptive for the child to set his or her future reproductive behavior according to current indices of mortality and risk (for a critique of this assumption, see Rowe 2000a). It is not clear to what degree ecological conditions have been (relatively) stable or fluctuating over our evolutionary history; nevertheless, the possibility of environmental fluctuations certainly reduces the reliability of parental behavior as a cue for expected risk. Belsky (2005) has suggested that cross-generational instability in environmental conditions could select for genotypic diversification in sensitivity to rearing influences, with some infants being genetically predisposed to be less affected by parental behavior than others (see also sect. 6.4). In a discussion of parental effects on pubertal timing, Ellis (2004) proposed a new explanation of why lowquality parenting should accelerate the onset of puberty, which he labelled the “child development theory.” The key idea is that children are not choosing their future reproductive strategy; rather, they are using information about parental investment in order to regulate the length of childhood. In this framework, the child is not responding to indirect macro-ecological cues (such as mortality), but to direct micro-ecological cues about his or her own rearing environment. If parental care is of high quality, the child can benefit by prolonging childhood and maximizing parental investment (e.g., food, wealth, skills teaching, status). If, on the other hand, parental investment is hard to come by, it might pay to shorten childhood and reach independence from parents at an earlier age.

Child development theory is not incompatible with ecological risk models; in fact, it could help explain why some decisions concerning reproductive strategies are made so early in ontogeny, and disentangle the macroand micro-ecological levels that make up a child’s environment (see also sect. 7.1.1). Moreover, it is consistent with recent data on the anticipation of adrenarche (sect. 7.2). In section 6, I argue that an additional reason for early strategy switching is the importance of sexually selected traits in the context of children’s peer relationships. 5.4.2. Attachment to mother and to father: Do they provide different cues? Although the models reviewed

here focus on attachment security as a cue of ecological risk (and, in child development theory, of the future quality of parental investment), a whole literature inspired by Draper and Harpending (1982) has singled out paternal investment (and, in particular, father absence) as a crucial factor influencing pubertal timing in daughters (see Ellis 2004, for a review). The theoretical basis for focusing on paternal investment is that paternal care, much more than maternal care, is contingent on the mating system (monogamy vs. polygyny) and on the degree of local male – male competition for status, in addition to environmental risk (see also sect. 6.2). Thus, inconsistent or detached paternal care would act as a cue that (1) mating is polygynous (Kanazawa 2001), and/or that (2) paternal investment is unreliable, is probably not crucial for successful reproduction, and should not be expected from future partners. This would prompt daughters to adopt a reproductive strategy based on early sexual maturation (which is advantageous for females in polygynous systems; see Kanazawa 2001) and low commitment in long-term relationships (since paternal investment is not forthcoming). On the male side, sons from fatherabsent families tend to show increased aggressiveness and hypermasculine behavior (Draper & Harpending 1982), which can be seen as preparation for increased male – male competition for status (the Young Male Syndrome). Consistent with the idea of the father as a vector of mating-related cues, there is also evidence that harsh or insensitive fathering has a distinctive role in predicting the onset of “conduct disorder” in boys (reviewed in DeKlyen et al. 1999). This suggests that security of attachment to mother and father may have different (and partly independent) effects on the development of boys and girls. Unfortunately, research on the developmental correlates of maternal versus paternal attachment is still carried out with virtually no reference to evolutionary hypotheses, so that the dependent variables employed in most studies lack direct biological relevance to male–male competition, status-seeking, and sexual style. Nevertheless, there is some interesting (if inconclusive) evidence of parent-specific effects: maternal attachment better predicts scholastic skills and “emotional maturity” in adolescence (Aviezer et al. 2002), a range of measures related to play quality and interpersonal conflict resolution (Suess et al. 1992), and “positiveness of self” in preschoolers (Verschueren & Marcoen 1999). On the other hand, paternal attachment, sensitivity, and availability seem to be more related to anxious/withdrawn behavior in preschoolers (Verschueren & Marcoen 1999), aggression with peers and peer rejection in middle childhood (Booth-Laforce et al. 2006; Verschueren & Marcoen BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

11

Del Giudice: Sex, attachment, and the development of reproductive strategies 2002), and friend support, depression, and conflict with friends in early adolescence (Lieberman et al. 1999; Ducharme et al. 2002; Liu 2007). Most studies found no significant interaction with children’s sex, but this may often depend on small sample size. 6. Sexual selection, parental investment, and sex differences in optimal life histories

aggressive and status-oriented competition. Moreover, men develop more slowly than women, and tend to marry at a later age in most cultures (Eibl-Eibesfeldt 1989; Ellison 2001; Geary 1998; 2002; Mace 2000b). In addition, human reproduction features two important characteristics that, without being unique to our species, contribute to define the evolutionary landscape of sexspecific life histories: facultative paternal care and cooperative breeding.

6.1. Sex-specific life history strategies

6.2.1. Facultative paternal care. Human fathers partici-

Sexual selection (see Andersson 1994; Geary 1998; 2002) is, simply stated, natural selection following from the behavioral correlates of sexual reproduction. A character can be sexually selected in two main cases: when it is involved in same-sex competition for access to mates (intrasexual competition), and/or when it is related to mate choice by the opposite sex (intersexual choice). At a deeper level, the dynamics of sexual selection are driven by sex asymmetries in reproduction, especially those concerning parental investment (Clutton-Brock 1991; Trivers 1972). Members of one sex, usually females, provide higher investment in offspring production, both energetically and in parenting behavior; members of the other sex (usually males) invest less, and sometimes considerably less. In some species, this can consist of just the provision of sperm to females. As a result, the investing sex becomes the “choosing” side, while the other sex engages in strenuous competition for mates. Sexual selection and asymmetries in parental investment are intimately related to the shaping of life histories (Ho¨glund & Sheldon 1998; Kokko 1997; Svensson & Sheldon 1998). Life history strategies are essentially centered on achieving maximal reproductive success, and the way to attain the latter usually differs between males and females. As a consequence, males and females within a species experience different constraints, costs, and benefits, resulting in sex-specific life history strategies. The primary sex difference is in the balance of mating versus parenting effort (see sect. 5.1), with females investing more in parenting and males spending more time and energy in finding mates. This results from the fact that males enjoy higher maximum reproductive rates (i.e., potentially, a male can reproduce much more often than a female), so their benefit in pursuing additional matings can be much higher than it is for females. In addition, because males engage in same-sex competition, they usually need more developmental time and effort than females to reach an optimal degree of competitive ability. This is obviously true for physical size and strength, but it applies just as well to other sexually selected traits: Fighting ability, social competence, and courting displays all take time and energy to develop, as do the resources and social status needed for successful competition in humans. As a result, males tend to be slow developers and typically reach maturity later than females (Ellison 2001; Geary 2002).

pate in parental care of their children, sometimes rivaling mothers in the amount of investment they provide. This attenuates the asymmetry between the choosing and the chosen sex, and leads to female –female sexual competition and increased mate choice by males. However, paternal investment in humans is not an obligate trait: some men invest more and help mothers considerably, whereas others look for short-term mating opportunities and invest less in parenting (if at all). Such variation arises both between and within cultures, so paternal care in humans is best characterized as a facultative adaptation (see Geary 2005a; Hrdy 2005a; Miller 1994). The degree of paternal investment found in a given population is determined by many factors; for example, paternal investment is expected to be enhanced if the fitness advantage conferred on offspring by the father’s presence is substantial, if the opportunity for new matings is low, and if paternity certainty is high (see Geary 2005a). Availability of new mates stands out as an especially powerful factor: A survey of four hunter-gatherer societies by Blurton Jones et al. (2000) showed that mating opportunity (expressed as a function of fertile women-to-men ratio) was the strongest negative predictor of marital stability, which in turn is a major determinant of continued paternal investment across cultures (e.g., Betzig 1989; Draper 1989; Furstenberg & Nord 1985; Furstenberg et al. 1983). More generally, “father-absent” societies are characterized by aloof couple relationships, polygyny, warfare, and high levels of male competitive displays. This pattern is often seen in resource-rich ecologies, and in societies practicing low-level agriculture (Draper & Harpending 1988). A cross-cultural study by Quinlan (2007) showed that paternal and maternal care respond differently to environmental risk: Whereas maternal care decreased steadily with increasing famine and warfare, and with high levels of pathogen stress (while increasing at moderate levels), paternal involvement was negatively related to pathogen stress, and only weakly related to famine. Moreover, the local degree of polygyny seemed to partly mediate the association between pathogen stress and paternal involvement. Polygyny is strongly predicted by high pathogen stress (Marlowe 2003), so in general it may covary with extrinsic risk and mediate its effects on paternal care. In polygynous societies, males contribute less to subsistence, and direct paternal care is reduced (see also Marlowe 2000; 2003); counterintuitively, couple stability is increased overall, probably due to the shortage of women created by polygyny (Quinlan & Quinlan 2007a) and to the fact that polygynous men need not divorce and remarry if they can afford to acquire new wives (Marlowe 2000). Couple stability, however, can well coexist with high mating effort, as shown by the lower investment in paternal care and the

6.2. Human reproductive ecology

What about humans? Most of the above applies to our species as well: Women invest more in offspring than men, and men show much of the standard repertoire of 12

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

Del Giudice: Sex, attachment, and the development of reproductive strategies increased frequency of extramarital affairs in polygynous/ low male provision societies (Marlowe 2000).

considering how sex-specific selective pressures and trade-offs might relate to insecure attachment styles.

6.2.2. Cooperative breeding and alloparental care. The

second characteristic is that humans, like a small number of other primates, probably have an evolutionary history of cooperative breeding (Hrdy 1999; 2005a; 2005b; Kramer 2005; Mace & Sear 2005). Mothers living in ancestral groups were not alone in caring for their children, but could elicit alloparental care from others, especially older children and female relatives (i.e., grandmothers and aunts). The presence of maternal grandmothers and older sisters has been shown to increase child survival in many populations, with smaller but similar effects for maternal grandfathers and older brothers (reviewed in Mace & Sear 2005). Cooperative breeding has been documented in many traditional societies, and shows remarkable cross-cultural invariance. In a survey of traditional cultures by Kramer (2005), infants received about 50% of their daily care time from mothers, with little variation in percentage between cultures; the remaining 50% was distributed among siblings (10 –33%), grandmothers (1 – 12%), other alloparents (3 – 21%), and fathers. Fathers’ contribution to direct infant care was rather variable and, one would say, not very impressive, ranging from less than 1% to about 6% of time. Similar figures are found in nontraditional societies as well (e.g., Geary 2000; Lamb et al. 1982: Lampert & Friedman 1992). Note that I am not equating paternal investment with direct caring time; paternal investment in humans comes in many forms, including food provision, protection, teaching, and social status (Geary 2000; Geary & Flinn 2001). However, these data strongly underline the importance of alloparenting in human reproductive ecology. Hrdy (1999; 2005a; 2005b) argued that, as happens in other cooperatively breeding primates, human mothers without available alloparents (partly synonymous with “social support”) are more likely to abandon, neglect, or even kill their children (see also Hill & Hurtado 1996; Wasser & Barash 1983). If alloparental care is really so vital for human mothers, its absence (or low quality) should be considered a specific, major source of environmental stress; for females, the “socioassay” taken in the first 5 – 7 years of life should definitely include the availability of alloparents, including one’s own parents and siblings. The evolutionary importance of cooperative breeding would also explain the tendency of young children to form multiple attachments, while at the same time maintaining a “special” relationship with a primary attachment figure (usually the mother) – a phenomenon labeled monotropy (see Ahnert 2005; Cassidy 1999). 6.3. Implications for attachment theory

Because of asymmetries in parental investment and sexual selection, males and females face different trade-offs in their life history decisions. This must be taken into account by life history models of attachment, and related to expected costs and benefits of different strategies. Chisholm (1999) described sex-specific reproductive strategies under the labels of Young Male and Young Female Syndrome (sect. 5.3); in this section, I extend his analysis and those of Belsky (1999) and Kirkpatrick (1998) by

6.3.1. Sex differences in the mating versus parenting balance. Current life history models link insecure attach-

ment to a developmental trajectory of early reproduction, low commitment in long-term couple relationships, and low parental investment. However, from the above discussion, it is apparent that low parental investment involves different cost/benefit ratios for males and females. Males, much more than females, are facultative investors: for a male, low parental investment can be a very effective strategy, especially if the costs of raising a child are borne by the mother and her alloparental network. Moreover, uncertainty of paternity lowers the benefits of investing in one’s (probable) offspring. The decrease in fitness following lower investment in already-born children can be compensated for by additional matings; thus, males are able (in certain conditions) to employ a zero-parenting strategy wherein virtually all resources are devoted to mating. For women, such an extreme low-investment strategy is not feasible, since they are usually the primary caregivers and infants need at least some basic parental investment to survive. Even more crucially, women’s fitness does not benefit as much from mating with additional partners, given the strong limitations on female maximum reproductive rate. In other words, women cannot shift the balance between parenting and mating effort as easily as men (Archer & Mehdikhani 2000). In conditions of heightened environmental risk, both males and females are expected to adopt reproductive strategies focused on current reproduction and increased mating effort at the expense of parenting. However, the strategic balance in resource allocation will differ between the sexes, with males engaging in lower levels of parental effort than females. With increasing risk, the optimal investment for males will fall off rapidly because, when approaching the “saturation point” of offspring fitness, parental effort has diminishing returns (sect. 5.1), and the resources needed to increase offspring quality by a small amount can bring higher benefits if successfully diverted to mating. This is not the case for females, who cannot increase their reproductive output beyond a certain amount, nor avoid the basic (but by no means trivial) investment of pregnancy and lactation. Females will then favor a higher level of parental investment in each offspring, and are thus expected to (1) invest more than males in parental effort, and (2) try to elicit additional investment from mates and/or alloparents. Only at high levels of risk (i.e., when the saturation point becomes very low), can females meet the optimal investment even with little or no contribution from mates and/or alloparents.3 Then, at high levels of risk, females are expected to engage in low-investment mating without demanding additional investment. On the contrary, when the environment is safe (high saturation point), the optimal level of parental investment can become very high, and both males and females can gain by investing considerably in parental effort, thus maximizing offspring quality and their own long-term fitness. However, even at low levels of risk, paternal linvestment is more contingent on the degree of polygyny and on the availability of new partners, BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

13

Del Giudice: Sex, attachment, and the development of reproductive strategies which can considerably increase the benefit of diverting some resources from parenting to mating. Although increasing risk generally favors quantityoriented strategies, scarcity of social resources in absence of elevated risk could instead favor reproductive suppression (i.e., delaying reproduction until resources become available) in females (Wasser & Barash 1983). The same happens when energetic resources are scarce, which delays maturation and, in mature females, induces temporary suppression of fertility (see Ellison 2001; Ellis 2004). I suggest that lack of available alloparents (including potential mates) might lead to temporary reproductive suppression in women who are not adopting a low-investment strategy; this may involve behavioral correlates such as reduced sexual motivation and/or reduced interest in couple relationships (either long- or short-term). While this suggestion is openly speculative, it is crucial to keep resource scarcity as distinct as possible from environmental risk, since the two are expected to exert different effects on reproductive strategies (see also Clark & Daly 2005). Translating the above considerations into the framework of attachment theory, insecure males should readily adopt avoidant strategies, which are most likely to maximize their fitness in a threatening environment. Insecure females, on the other hand, should preferentially adopt anxious, care-eliciting strategies (i.e., preoccupied/ ambivalent attachment). This would keep them in close contact with kin, and help spread the costs of reproduction onto relatives. In cooperative breeding systems, the “helper” strategy can coexist with the helper’s own reproduction (Clutton-Brock 2002); thus, Belsky’s hypothesis about ambivalent attachment as helping-at-the-nest (sect. 5.2) could be partly correct, although without involving delayed reproduction. Maximizing help from kin is not the only benefit of anxious attachment: In the context of couple relationships, anxious behaviors (e.g., dependence, preoccupation with intimacy and partner availability) can help to maintain closeness with one’s partner, especially if the latter is avoidantly attached, and to maximize the available paternal investment (thus enacting a sort of “counter-strategy” to male avoidance). Intriguingly, an early study by Kirkpatrick and Davis (1994) found that couples made up of dismissing men and preoccupied women can be long-lasting, as much as the secure-secure ones (even while enjoying less couple satisfaction). There is also evidence that preoccupied women find it especially difficult to end couple relationships, even following abuse or deceit by their partner (Henderson et al. 1997; Jang et al. 2002). This is consistent with a strategy aimed at maximizing closeness and continued investment, even by reluctant or uncommitted mates. Thus, in the context of cooperative breeding, the anxious phenotype can be a useful device for adult females to extract investment and care from both relatives and mates. In this framework, females are expected to adopt male-like avoidant strategies as a second choice, tailored to high levels of environmental risk. This view is supported by the cross-cultural survey by Schmitt et al. (2003a), who found dismissiveness in women to covary with environmental risk and mortality rates, and to do so much more strongly than in males (on average, r ¼ .40 versus r ¼ .23). Finally, scarcity of social resources (in absence of high risk) can be expected to induce temporary 14

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

reproductive suppression in females. This possibility has not been yet considered in evolutionary models of attachment, but (if supported by evidence) it would have important implications. For example, some avoidant women may be actually adopting a temporary suppression strategy (e.g., by avoiding sexual relationships altogether), and shift to anxious or secure patterns when social conditions improve; even if suppression was not related to specific attachment patterns, it could alter the expected relationship between attachment and sexual behavior. The idea that women’s reproductive strategies are highly condition-sensitive is not new and has been championed by Lancaster (1989) and Hrdy (1999; 2000). Basically, they contend that, when monogamy is not a viable choice (e.g., because of low paternal investment and/or low quality of potential mates), women can adopt “facultative polyandry” as their optimal strategy. Mating with many partners allows the exchange of sexual access for immediate benefits and, even more importantly, creates a network of “possible fathers,” who can then provide protection and help, and be tolerant (non-aggressive) towards the mother’s children. The theory I propose is compatible with the facultative polyandry hypothesis, while adding another layer of complexity to female strategies. Avoidant strategies can easily lead to polyandry; for reasons discussed earlier (sect. 5.2), anxious strategies can lead to multiple matings as well, while retaining an orientation to long-term commitment. Finally, some women may be actually adopting a third kind of strategy – one aimed at temporarily suppressing reproduction because of scarcity of available alloparenting and social support. 6.3.2. Intrasexual competition. There is one more reason

to consider attachment in the light of sexual selection: Because attachment patterns are related to the development of personality traits (and, in general, exert a powerful organizational effect on social development), their costs and benefits should differ between males and females if the traits in question are sexually selected. Avoidant attachment, which is related to inflated self-esteem (Cassidy 1988) and aggression with peers (e.g., Card & Hodges 2003; Erickson et al. 1985; Finnegan et al. 1996; Renken et al. 1989), is likely to be more adaptive for young males who need to defend their social status in anticipation of early reproduction (male– male competition); moreover, the above-mentioned traits can be attractive to females, thus contributing to mate choice (see Sadalla et al. 1987; Weisfeld et al. 1983). The opposite applies to ambivalent attachment, which predicts fearfulness, withdrawal, and passive behavior with peers (e.g., Card & Hodges 2003; Cassidy 1988; Cassidy & Berlin 1994; Erickson et al. 1985). These traits are not nearly as damaging for females as they are for males. For example, Morison and Masten (1991) found that the “sensitive-isolated” behavioral profile in middle childhood was associated with lower self-worth in males, but higher self-worth in females. Also, studies correlating attachment with externalizing (e.g., aggression, disruptiveness) and internalizing (e.g., anxiety, withdrawal) behavioral problems often find effects that are sexspecific to some degree. In the study by Renken et al. (1989), avoidance predicted externalizing symptoms only in males; and Finnegan et al. (1996) found that avoidant coping was correlated with more externalizing problems,

Del Giudice: Sex, attachment, and the development of reproductive strategies but fewer internalizing problems in males; preoccupied coping, on the other hand, was related to higher internalizing problems in males but not in females. Ambivalent children also tend to be victimized in peer groups, especially by avoidant children, who instead take the role of bullies (Troy & Sroufe 1987; Finnegan et al. 1996). Although bullying has traditionally been associated with peer rejection (e.g., Pellegrini 1995; Weisfeld 1999), being a bully is usually better than being a victim (e.g., Juvonen et al. 2003), and may be a reasonable way to secure one’s place in male dominance hierarchies and gain access to social resources. Indeed, researchers are becoming increasingly aware that not all bullies are equal; especially among older children and young adolescents, there is a subgroup of aggressive children who are accorded high peer status, are rated as “cool” and attractive by girls, and date more often (e.g., Juvonen et al. 2003; Pellegrini & Bartini 2001; Rodkin et al. 2006). Whereas, in early childhood, aggression is associated with peer rejection (note, however, that rejection and dominance status are only weakly correlated), at later ages it becomes increasingly predictive of peer acceptance (Bukowski et al. 2000). In synthesis, the overtly aggressive and self-aggrandizing style of avoidant children can provide a competitive way to gain status and dominance in male groups. The ambivalent pattern has been less studied, because of its relatively low frequency in early childhood, so it is more difficult to relate it to possible adaptive benefits for girls (at the same time, we know less about the mechanisms of dominance and status in female peer groups). What has been observed empirically is that insecure patterns are more extremely skewed in boys than in girls, and it is then possible that intrasexual competition has stronger implications for male attachment as well. 6.3.3. Why middle childhood? Middle childhood (approx.

age 7 –11 years) is the human homologue of the primate juvenile phase. Children of this age, like other juvenile primates, no longer depend exclusively on parental care for survival, and can forage effectively if they need to do so. At the same time, they are still sexually immature and have limited competitive abilities (see Bogin 1999; Geary & Bjorklund 2000; Kramer 2005). In this developmental phase, the peer group becomes the child’s primary interpersonal world; fight play, parenting play, and same-sex grouping all peak between 6 and 11 years of age, with little cultural variation (e.g. Geary 1998; 2002; Serbin et al. 1993). Conventional wisdom about children’s peer interactions is that they allow for “safe experimentation” with adult social skills. However, childhood peer relationships also have lasting effects on people’s lives, much more so than is usually realized. For example, dominance ranks in childhood tend to carry over into adolescence: In a longitudinal study by Weisfeld et al. (1987), “toughness” ranks of boys at 7 years of age correlated about .70 with dominance, popularity, and leadership at age 16. Other studies found stability from childhood to adolescence in related personality traits, such as dominance and passivity (reviewed in Weisfeld 1999). Although the influence of heritable traits on these results has not been quantified (social dominance is also related to strength and physical attractiveness, and shows moderate heritability; e.g., Gottesmann 1966), it seems reasonable to consider that early outcomes,

especially in the field of dominance and status, can have a lasting influence on later development. “Social inertia” is a well-known phenomenon in animal dominance hierarchies: After the first encounters settle the initial ranking, individuals tend to keep their position in the hierarchy as long as they remain in the same group, even if their hormonal levels are experimentally manipulated to match those typical of higher-ranking animals. The same manipulations, however, have dramatic effects if an individual is placed in a new, unfamiliar group (Adkins-Regan 2005). Human juveniles are not just “preparing” for their adult roles; they are actually establishing a starting place in the social group, which in turn will influence their future reproductive opportunities to some degree. In this phase, conflict between early behavioral strategies (influenced by parental care) and new environmental demands (driven by the peer group) could start becoming apparent, to culminate later during adolescence. 6.4. Multiple factors affecting reproductive strategies 6.4.1. Other environmental factors. A cautionary note is

warranted at this point. As mentioned above, individual reproductive strategies are not completely determined by developmental factors such as attachment security and early stress. Many researchers have argued for a multifactorial view of reproductive decisions, and have described other factors affecting mating strategies: local sex ratio, pathogen load, social and economic structure, self-perceived mate value, attractiveness, and age (see Barber 2000; Campbell 2000; Cashdan 1996; Gangestad & Simpson 2000; Landolt et al. 1995; Schmitt 2005a; Schmitt et al. 2003b; Voland 1998). Of course, some of these factors (e.g., sex ratio, pathogen load) are known to affect parental investment, and their effects could turn out to be partly or fully mediated by early stress and attachment; other factors (such as attractiveness and age), however, are likely to have independent effects on reproductive strategies. For example, a cross-cultural study by Schmitt (2005a) found that, while low interpersonal trust and insecure attachment correlated with short-term mating, there were also many associations between short-term mating and “positive” traits, such as low psychological symptoms and high self-esteem – especially in males. In addition, short-term mating in men tended to increase with age, and men were on average more oriented to short-term mating, regardless of attachment style. Similar findings were reported by Egan and Angus (2004). They found that rate of sexual infidelity correlated positively with psychopathic traits (i.e., manipulative and egocentric behavior) and negatively with agreeableness and social desirability in both sexes; however, men who had been unfaithful at least once were higher in socially desirable personality traits such as agreeableness, extraversion, and conscientiousness. Clearly, these results support a multifactorial model of reproductive choices. Generally speaking, it must be stressed again that males have a bias towards short-term mating, probably because of the high benefit/cost ratio of this behavioral option (e.g., Buss & Schmitt 1993; Schmitt et al. 2003b); highly attractive men may actually look for an increased number of short-term partners (Gangestad & Simpson 2000; Jackson & Kirkpatrick 2007; Landolt et al. 1995). Schmitt (2005a) also found a tendency for short-term mating orientation to increase in people either BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

15

Del Giudice: Sex, attachment, and the development of reproductive strategies living with a partner or married, making it clear that commitment in a long-term relationship and high parental investment can coexist with the occasional pursuit of short-term mating opportunities (see also Geary 1998). 6.4.2. Heritable individual differences. Individual differences in reproductive strategies are also affected by heritable genetic variation. In section 5, I mentioned findings of substantial genotypic variance in life history traits, such as pubertal maturation and age at menarche. Other researchers have tried to investigate the broader spectrum of life history strategies, and they also found support for the idea that heritable factors play an important role in shaping human reproductive styles. Figueredo et al. (2006) summarized their research on the “K-factor,”4 a global measure (derived by factor analysis) accounting for a large portion (about 70–90%) of reliable variance in a broad class of life-history-related traits (such as attachment to parents, romantic attachment security, mating effort, manipulativeness, risk-taking, and altruistic feelings). In a twin study, the heritability of the K-factor was estimated at .65 (Figueredo et al. 2004), suggesting considerable genotypic influence. As the authors note, the same result also implies that a substantial portion of variance in reproductive strategy is influenced by environmental factors (and/or genotypeenvironment interactions). A common evolutionary explanation of heritable individual differences in sexual strategies invokes the concept of frequency-dependent selection (e.g., Campbell 2000; Gangestad & Simpson 1990; Rowe et al. 1997). Genetic variation can be maintained if the resulting phenotypes are adaptive in the context of the other phenotypes present in the population. For example, if all males provided low parental investment and engaged in high competition, it could be adaptive for some of them to become less competitive and to invest more in parenting, so that genes favoring parental investment would spread; the two strategies might eventually reach an equilibrium state, each at its optimal frequency, and coexist in the population (note that more complex and “dynamical” outcomes are also possible; e.g., involving cycles in genotype frequencies). Nettle (2006; 2007) provides a good introduction to the possible sources of heritable variation in behavioral traits, including frequency-dependent selection. The empirical data support the view that human reproductive strategies are shaped by both heritable and environmental factors, with the possibility of substantial gene– environment interaction (different genotypes showing different conditional responses to the environment). It is also possible that some genotypes are more responsive than others to environmental cues; this hypothesis of “differential susceptibility” to rearing environment was proposed by Belsky (1997b; 1999; 2005) in the context of early stress and attachment styles, and is gathering increasing empirical support in various domains (Belsky et al. 2007a).

7. A synthesis: Attachment and the development of reproductive strategies I now draw upon theory and evidence presented up to this point, and outline an updated synthesis of the life-history model of attachment. First, I follow the main phases in the development of the attachment system, from infancy 16

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

to adulthood (Fig. 2), highlighting the empirical predictions derived from the model at each stage. Then, I propose a hormonal basis for the middle childhood transition, discuss the evidence for the influence of sex hormones on attachment, and speculate on its implications for theory and research. In dealing with insecure attachment patterns, I will focus on the avoidance/ambivalence dimensions, leaving aside the issue of disorganization. There are three pragmatic reasons for this choice: first, the literature on attachment and life histories has been mainly concerned with explaining avoidant and ambivalent patterns. Second, disorganization is in many ways orthogonal to the three-way (ABC) classification, and (especially if low to moderate) can coexist with a primary, organized pattern; thus, its effects are likely to interact with those of the avoidant/ ambivalent components. Last and most important, adult attachment questionnaires do not assess disorganization or its sequelae, so that at present it is hard to track the disorganized developmental trajectory from childhood to adulthood (but see Simpson & Rholes 2002, for the interesting proposal that fearful-avoidant romantic patterns may reflect disorganization). That said, I have no intention of downplaying the importance of disorganized attachment, and investigating this issue will be a crucial task for future research and theoretical work. Disorganized attachment is often related to severe levels of psychosocial stress and is characterized by sex differences from early childhood – two features that suggest high relevance to the life-history model proposed here. 7.1. Developmental phases and transitions 7.1.1. Infancy and early childhood. In infancy and early

childhood, the attachment system has the function of eliciting care and protection during the period of maximal dependence on parents. For this reason, attachment styles can be expected to track caregivers’ behavior quite closely, adapting to changes in environmental and parenting conditions as shown by longitudinal studies. No strong sex differences in attachment patterns are predicted (nor have been reported) in this developmental phase; as an exception, males are more often and severely disorganized than females (sect. 4.1). At the same time, young children use attachment security as a “socioassay” of their current (and expected) local ecology. Insecure attachment acts as a cue that (1) the environment is risky, possibly involving high mortality rates, and/or (2) the child should expect to receive low parental investment in the future; this includes physical resources, teaching, protection, and alloparenting. As discussed in section 6.3, the presence and availability of alloparents (including siblings, other relatives, and potential mates) could be an important cue driving female reproductive strategies, and possibly influencing attachment security and/or style in girls. The macro- and micro-ecological cues gathered in early childhood also affect the timing of the child’s transition to juvenility (marked by adrenarche; see sects. 5.2 and 7.2). Child development theory (sect. 5.4.1) suggests that lowquality parental investment should lead to shortened childhood, as recently confirmed by Ellis and Essex (2007), who found that early stress anticipates adrenarche in both sexes. At the same time, insecure attachment acts

Del Giudice: Sex, attachment, and the development of reproductive strategies

Figure 2. Schematic diagram of the development of reproductive strategies. Black lines represent typical developmental pathways, with thickness roughly proportional to phenotype frequency in relatively low-risk populations. The core causal relationships discussed in the text are indicated by solid arrows; dashed arrows represent other possible effects, relevant to the model but not discussed in detail here. For ease of presentation, the diagram shows high stability in attachment security across infancy and childhood. This is a simplification, as changes in security are possible throughout the lifespan. In addition, attachment strategies are depicted as discrete categories, although they are probably better described as continuous dimensions.

as a cue of environmental risk, thus leading to current reproduction-oriented life history strategies and earlier maturation. Note that low parental investment can act as a indication of risk for two reasons: it can inform the child that extrinsic risk is generally high, but it can also mean that parents are not willing to invest in that specific child for other reasons (e.g., low phenotypic quality, presence of step-parents, cultural bias against males or females, etc.), resulting in higher mortality risk for the child itself. Of course, the likelihood of such discrimination by parents will increase when social resources are scarce and/or ecological risk is high (see also Chisholm 1999). Thus, parental investment can act as a cue of risk at both the macro- and micro-ecological levels, which helps to explain why one finds variable degrees of attachment security within the same social group and the same family (another reason being genotypic variation in environmental susceptibility). 7.1.2. Middle childhood: A transitional phase. At the

beginning of middle childhood, or human juvenile stage, the attachment system undergoes a phase of remarkable reorganization. The available data show that insecure children’s attachment patterns become highly sex-biased, so as to switch their reproductive strategies towards sex-optimal developmental pathways. Attachment security/insecurity is a reliable index of socioecological risk, and, as such, it is retained as a relatively stable, prototype-like behavioral trait5 (Fraley 2002). The specific insecure strategies adopted in early childhood, on the contrary, can be viewed as disposable phenotypes, to be modified during development if they do not suit the adaptive interest of the growing children (see sect. 3).

The most immediate selection pressure on attachment styles in middle childhood probably comes from intrasexual competition in the peer group (sect. 6.3.2). In middle childhood, children begin to fight their way through social reality, and the first outcomes can have long-lasting effects on future development – perhaps more so for insecure children, who cannot count on a protective family network to buffer them against difficulties and failures. Thus, the behavioral correlates of attachment patterns are likely to be sexually selected already at this stage. In particular, the avoidant pattern is associated with aggression, self-reliance, and inflated self-esteem – all traits that can be useful to males as a high-risk status-seeking strategy. Indeed, empirical studies in middle childhood show that nearly all insecure boys can be classified as avoidant. Girls, on the other hand, tend to shift to ambivalent styles (while in a less extreme fashion); it is less clear whether this particular pattern gives them some advantage in the peer group, or if it just anticipates adult strategies. Notably, attachment in middle childhood does not relate only to aggression: Sroufe and colleagues (1993) found that insecure children aged 10–11 years were more likely than secure ones to “violate gender boundaries,” which included flirting, physical contact, and sexual gestures. This is fully consistent with life-history models, and indicates that insecure strategies may relate to earlier initiation of sexual activity already in middle childhood. A question may arise at this point: If avoidant attachment can be such a rewarding strategy for males, why don’t all males (including secure ones) shift to this pattern? In fact, there is evidence that, among secure boys, a sizeable proportion shows secondary elements of avoidance (sect. 4.2; BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

17

Del Giudice: Sex, attachment, and the development of reproductive strategies Del Giudice 2008), so this could be partly correct. However, it is not the case that successful social strategies always involve competitive status-seeking. If environmental conditions are safe and predictable, and if the local mating system favors monogamy and high paternal investment, then low-risk, cooperative strategies can be more successful in maximizing males’ long-term fitness (sect. 6.3.1); lower male avoidance is predicted in this case, consistent with weaker male-male competition and reduced conflict of interest between mothers and fathers. As mentioned in section 5.4.2, it is possible that maternal and paternal attachment play different roles in determining the behavioral strategy adopted in middle childhood. Maternal and paternal investment seem to respond differently to extrinsic risk (Quinlan 2007; see sect. 6.3.1), and could, therefore, be expected to shape children’s life history strategies in somewhat different ways. In particular, paternal investment can carry useful information about the local degree of male–male competition, polygyny, and paternal involvement, which could affect specific aspects of children’s attachment and behavior (e.g., competitiveness and risktaking in boys, avoidance vs. ambivalence in girls). Further research is needed to assess the merit of this hypothesis, and to make sharper predictions about the possible effects of maternal versus paternal attachment on sons and daughters. Finally, although attachment-based models focus on environmental cueing of secure versus insecure and avoidant versus ambivalent strategies, heritable effects are also likely to influence development in this phase. As I explain in section 7.2, the physiological changes taking place in middle childhood are likely to reveal previously unexpressed genotypic variability, which could significantly affect the regulation of life-history-related traits (including attachment). In fact, heritable effects on adult romantic attachment styles have been recently found using the RSQ (Brussoni et al. 2000); no comparable study has been yet performed in middle childhood. Genotypic variation can affect reproductive strategies in many ways; one possibility is that the genotype influences the degree of environmental sensitivity, leading to relatively flexible versus fixed strategies, as proposed by Belsky (2000; 2005; Belsky et al. 2007a). 7.1.3. Adolescence and adulthood. With the coming of

adolescence, the attachment system finally takes on its mature function, that of regulating couple relationships and mating strategies. In general, insecure strategies can be seen as maximizing current reproductive success and privileging mating effort at the expense of parenting. However, the mating versus parenting balance entails different tradeoffs for males and females (sect. 6.3.1). Insecure males are predicted to favor avoidant strategies (low parental investment, short-term and uncommitted mating), whereas insecure females are expected to show a broader range of phenotypes: they should adopt anxious, investment-eliciting strategies when environmental risk is moderate, and avoidant strategies when faced with challenging conditions. Both strategies are usually related to impulsive mating and to short-term sexual orientation in females; however, whereas avoidant women show little desire for commitment and intimacy, anxious women also show heightened desire for long-term relationships, intimacy, and romance (sect. 5.2). Female anxiety could partly act as a “counter-strategy” to male avoidance, by coupling impulsive mating (which can 18

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

initially attract avoidant males) with high requests for commitment and investment. To render the picture more complex, some women may adopt a temporary strategy of reproductive suppression, possibly marked by low interest in both short- and long-term relationships, due to perceived lack of social resources. Finally, secure adolescents of both sexes are expected to follow parenting-oriented reproductive strategies maximizing future reproduction (later onset of reproduction, high commitment in couple relationships, high parental investment, and later puberty in girls). In the present model, avoidant/ambivalent patterns after the middle childhood transition are expected to predict avoidant/anxious attachment strategies in adulthood. This would provide a developmental basis for adult measures of romantic attachment, which to date lack a clear theoretical and empirical link to childhood antecedents (Belsky 2002). A related prediction is that the size of sex differences in middle childhood should mirror that in the adult population. One should find larger differences in cultural regions characterized by moderate environmental risk, somewhat smaller differences in low-risk regions (because of higher security in both sexes), and the smallest differences (mostly driven by higher female avoidance) in high-risk regions. It is noteworthy that, until now, reports of marked sex differences in middle childhood all came from regions showing aboveaverage adult sex differences in the cross-cultural study by Schmitt et al. (2003a): Italy (d ¼ .21), Israel (d ¼ .21), western United States (d ¼ .26), and Canada (d ¼ .35). To end this section, I briefly turn to the issue of plasticity in adulthood. Current life-history models tend to confine plasticity to the first years of life, without considering how individuals may adjust their strategies later in development. Nevertheless, it is quite reasonable to assume that, given the long reproductive lifespan of humans, there is room for strategic change in adolescence and adulthood as well. Improvements in social support and in the quality of couple relationship are known to affect parenting and attachment, as noted by Chisholm (1993; 1999); ultimately, changing socioecological contexts could lead adults to “revise” their reproductive strategies, with behavioral as well as physiological consequences (see also Cashdan 1996). Age itself is a key variable affecting the main life history trade-offs, with likely consequences for mating and parenting strategies (e.g., Delton et al. 2006). For example, males throughout the world tend to shift from high mating effort in young adulthood to a phase of increased parental investment (Winking et al. 2007); this transition may be especially dramatic in insecurely attached men. In section 4.3, I reported initial evidence that sex differences in attachment styles peak in late adolescence and decline towards middle age; this supports the idea that sexual selection on reproductive strategies is stronger in early adulthood, and that the relational styles of men and women become substantially more similar at later ages. Finally, researchers are beginning to study how attachment representations change following marriage and parenthood (e.g., Crowell et al. 2002; Treboux et al. 2004); results from this research field will be extremely useful in increasing the realism of life-history models. 7.2. Hormonal basis of the middle childhood transition

So far, I have argued that middle childhood is an important transitional period, and have provided the evolutionary

Del Giudice: Sex, attachment, and the development of reproductive strategies rationale to support this view. However, the idea that behavioral change at this age is related to sexually selected strategies may sound strange, since this kind of transition is usually associated with the later onset of puberty. On the biological side, middle childhood appears to be characterized by stasis rather than change – perhaps a legacy of the Freudian concept of “latency.” This view is incorrect. On the contrary, middle childhood is a phase of intense (though physically concealed) endocrine development, anticipating puberty in many respects. As I show in this section, there are reasons to consider middle childhood as the actual beginning of “adult” sexual differentiation at the neurobehavioral level. The possibility that sex differences in attachment styles (precursors of reproductive strategies) are primed by such hormonal changes should definitely be considered and investigated. 7.2.1. Adrenarche. At about 6 years of age, with little difference in timing between males and females, the adrenal cortex of both sexes begins to secrete a growing amount of androgens into the bloodstream. These do not include the familiar androgen testosterone, which will begin to rise later in puberty. The main products of adrenal glands are dehydroepiandrosterone (DHEA), dehydroepiandrosterone sulfate (DHEAS), and androstenedione (A4), three chemical precursors of testosterone and estrogen. Secretion of adrenal androgens increases steadily for about 10 years, reaches a peak in early adulthood, and then slowly declines. The onset of adrenal androgen production is called adrenarche, and marks the beginning of the developmental phase known as adrenal puberty (Auchus & Rainey 2004; Iba´n˜ez et al. 2000; Palmert et al. 2001; Spear 2000). Adrenal puberty is a peculiar feature of human development, absent in most other mammalian species (including primates); to date, it has only been documented in chimpanzees and gorillas, which also undergo a prolonged juvenile phase before reproduction (Iba´n˜ez et al. 2000). DHEA and DHEAS were once thought to be “weak” androgens, because they show low affinity with androgen receptors, and as such were largely ignored by researchers. However, it has been recently discovered that brain cells (and other peripheral tissues) express the enzymes needed to convert precursors such as DHEA into “active” testosterone and/ or estrogens (see Adkins-Regan 2005; Labrie et al. 2005). According to current estimates, such “intracrine” production of sex hormones in peripheral tissues accounts for about 75% of total estrogen in women and 50% of total androgens in men (Labrie et al. 2005). Thus, adrenal androgens contribute significantly to sex-hormone production in adults; in children, they can drive development along sexspecific developmental pathways before full reproductive maturity. Through local conversion to testosterone and estrogen, adrenal androgens can be behaviorally active even if they have only minimal effect on bodily development (i.e., initial growth of axillary and pubic hair, increased oil in the skin, and a slight acceleration of skeletal growth). They may also exert direct behavioral effects, via neuromodulation of GABA receptors and upregulation of the androgen receptor (see Simon & Lu 2006). Indeed, adrenal androgens have been shown to influence brain function in laboratory animals, and are included in the family of neuroactive steroids (Spear 2000). There is preliminary evidence

that DHEAS levels may be linked to aggression in middle childhood, and high levels of DHEAS have been found in samples of children (mostly boys) diagnosed with Conduct Disorder (CD) and Oppositional Defiant Disorder (ODD; van Goozen et al. 1998; 2000). In female rodents and primates, DHEA reduces aggression, although it is still unknown whether this also applies to humans (Simon & Lu 2006). Hyperactivity symptoms in a sample of children with Attention Deficit Hyperactivity Disorder (ADHD) correlated with lower DHEA and DHEAS levels in a study by Strous et al. (2001). Finally, it has been proposed (Herdt & McClintock 2000; McClintock & Herdt 1996) that adrenarche could be responsible for the onset of the first sexual/romantic attractions, usually happening at about 7–10 years of age. Thus, adrenal androgens appear to be involved in sexual differentiation and in the initiation of early reproduction-related behavior in middle childhood. When adrenal androgens start to be secreted, and locally converted to active molecules, previously unexpressed genetic variation in the sex-hormones pathways will suddenly be uncovered and rendered effective. Such variation may include allelic variants in the many enzymes involved in hormone production, conversion, transport, reception, and degradation, all of which can potentially affect behavior. For example, sequence variants in the androgen receptor (AR) gene have been linked to life-history variables such as aggression, impulsivity, number of sexual partners, age of menarche, and likelihood of having divorced parents (Comings et al. 2002; however, the results were not replicated in a large study by Jorm et al. 2004). In addition, the activation of sex-hormone pathways is bound to interact with the organizational effects of prenatal and perinatal hormone levels. The rising levels of sex hormones in the brain, coupled with the release of sex-hormone related genetic variation, would determine a “modular” phenotypic transition between childhood and juvenility, where both sex-specific and heritable factors would come into play. This is consistent with the evidence of a relatively rapid, sex-specific reorganization of attachment patterns at about 7 years of age. But what is the relationship between sex hormones and attachment? 7.2.2. Sex hormones, stress, and attachment behavior: A complex interplay. Life-history models are usually cen-

tered on the effects of stress and attachment on sexual development: psychosocial stress and insecure attachment are expected to (1) accelerate sexual maturation (adrenarche in both sexes, and puberty in girls) and (2) affect a suite of reproduction-related behaviors (e.g., aggression, impulsivity), many of which are under the influence of sex hormones. The hypothesis I propose focuses precisely on the reverse effect, that of sexual development on attachment and, by definition, on stress regulation. Experimental evidence from nonhuman animals strongly suggests that sex hormones can directly affect attachment-related behaviors: testosterone administration dramatically reduces separation-induced distress vocalizations in chicks, quails, and guinea pigs (Bernroider et al. 1996; Panksepp 1998), whereas prenatal administration of estrogens seems to exert the opposite effect. In rhesus monkeys, too, prenatal testosterone has been found to influence the sex-specific development of separation vocalizations (Tomaszycki et al. Wallen 2001; Wallen 2005). BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

19

Del Giudice: Sex, attachment, and the development of reproductive strategies In a broader perspective, there is ample evidence that sex hormones deeply interact with stress-regulation mechanisms. Taylor et al. (2000) summarized a wide array of studies and argued that, in many animals (including humans), the stress response system of adults shows adaptive sexual dimorphism. The classic fight-or-flight response seems to be more characteristic of males, whereas the primary response of females is better described as “tendand-befriend.” In other words, stress tends to induce aggression and/or avoidance in males, but it solicits caregiving (protection of offspring) and seeking of social support and affiliation in females. The prevalence of the tend-andbefriend response is thought to depend on the kind and level of experienced stressors, and on the female’s reproductive status (e.g., on whether she is sexually mature, has dependent offspring, and is in a fertile phase). At the neurobiological level, this sexual dimorphism could be mediated by oxytocin and endogenous opioids, and would therefore be closely linked to the neural substrate of the attachment system (Keverne et al. 1999; Taylor 2006; Taylor et al. 2000). Sex hormones have a critical role in this pathway: androgens, for example, inhibit the stress-induced release of oxytocin, while estrogen enhances the anxiolytic effects of oxytocin (see Jezova et al. 1996; McCarthy 1995; McCarthy et al. 1996). Sex hormones can also directly affect the stress system through regulation of neuroendocrine activity in the amygdala (Viau 2002). Other mechanisms of interplay among sex hormones, stress, and aggression have been described in an evolutionary perspective by Korte et al. (2005). It is noteworthy that the fight-or-flight versus tend-andbefriend model closely mirrors the sex differences observed in avoidant versus anxious attachment styles; in particular, anxious strategies involve heightened seeking of support and closeness, which is typical of the female response to stress. From the above discussion, it is apparent that the stress/attachment system and the sexual system can interact bidirectionally across the life cycle; for example, psychosocial stress leads to accelerated adrenarche, which, in turn, could affect the stress-regulatory pathways (already primed by early experience) with both sex-specific and genotype-dependent effects. Such positive feedback mechanism could then function as effective developmental “switches,” leading to diverging life history trajectories. Many interesting questions arise from this hypothesis. Could attachment behaviors be related to prenatal sex hormone levels (which organize early brain development)? Might the relationship become manifest only starting from middle childhood, because of the activational effects of adrenal androgens? Are sex hormones the vehicle of genetic effects on the K-factor, with its overlap with attachment and mating styles? And are atypical degrees of masculinity/femininity related to “sex-atypical” attachment patterns, for example, in highly anxious men? A preliminary study (cited in Greenberg 1999) found an association between ambivalent attachment and gender identity disorders in childhood, but the evidence regarding this issue is still very limited. In summary, powerful hormonal changes take place at the start of middle childhood, and we are only beginning to understand their full implications for human development. The “hidden” onset of pubertal maturation brought about by adrenarche might be the neuroendocrine switch that begins to reorganize attachment towards its mating-related functions, even before the coming of full reproductive 20

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

maturity. While this hypothesis is still speculative, it provides a starting point to investigate the relationships among attachment, reproduction, and the hormonal mechanisms that regulate life history strategies (Adkins-Regan 2005; Ellison 2001). 8. Conclusion The study of reproductive strategies has become a fruitful, fast-expanding area of research in evolutionary psychology and anthropology. More broadly, life-history theory is emerging as a truly integrative paradigm in the study of individual differences, encompassing traditionally separated fields of inquiry such as attachment theory, the psychology of aggression and sexuality, personality theory, behavior genetics, and the anthropology of mating systems. The next frontier will be achieving integration with psychobiology and neuroscience, and this enterprise is already underway; see, for example, the evolutionary model of stress reactivity by Boyce and Ellis (2005; Ellis et al. 2006), and the work by Korte et al. (2005) on alternative phenotypes in stress regulation and aggression. I anticipate that, in this integration process, the study of sex differences and sexual selection will play an increasingly central role. Males and females face different strategic choices, and different evolutionary pressures, at each stage of their life cycle. Understanding how male and female strategies unfold (and interact) in the course of development is an essential step for appreciating the fascinating complexity and the deep evolved logic of human life histories. ACKNOWLEDGMENTS

I am deeply grateful to Jay Belsky and Alasdair Houston for their precious comments, and for the creative and rewarding discussions we had on attachment theory and evolutionary biology. I also owe much to Jim Chisholm, Sarah Hrdy, Susan Johnson, Bruce Ellis, and an anonymous reviewer for their help in sharpening the theory, catching mistakes, and (not least) writing a much clearer paper. Kathryn Kerns and Ildiko Toth kindly shared their original data for reanalysis. Finally, Marinus van IJzendoorn, Marion Bakermans-Kranenburg, Mattie Tops, Fabie¨nne Naber, Cristina Becchio, Angela Ciaramidaro, Livia Colle, Elisa Grandi, Valeria Manera, and Davide Mate all provided useful feedback on previous versions of the manuscript. NOTES 1. The distinction I draw here between past-focused interviews and present-focused questionnaires is a simplification: there are also interviews about current relationships and questionnaires about past attachment experiences. However, they are employed much less often than the “paradigmatic” instruments I discuss here, so my distinction holds as a general approximation. 2. When adaptive variations in life history parameters are condition-dependent, rather than genetically determined, some authors suggest they should be labeled life history tactics, since they can be seen as conditional branches of a single statedependent strategy (e.g. Gross 1996). However, since the term “strategy” has gained common use and other authors use “tactic” with different meanings, I refer to reproductive strategies throughout this article. 3. This kind of mechanism might explain the finding by Quinlan (2007) that, with increasing pathogen stress, paternal

Commentary/Del Giudice: Sex, attachment, and the development of reproductive strategies involvement decreases linearly while maternal care follows a quadratic function (first increasing, then decreasing). The increase in maternal care at moderate levels of risk could reflect mothers “compensating” for lowered paternal investment. 4. The letter “K” in the K-factor derives from the classic approach to life history theory, based on population growth rates. In this theoretical framework, slow and high-investment strategies are labeled “K-strategies,” while fast and low-investment strategies are labeled “r-strategies.” While this approach was popular some decades ago, most life-history theorists presently favor models based on age-specific mortality patterns and extrinsic risk (see sect. 5). 5. As discussed in section 2.3, stability in attachment security is not always expected to be high; in particular, negative life events often lead to shifts towards insecurity, which have the effect of decreasing measured stability. The “prototype” model discussed by Fraley (2002), however, implies that early security exerts continuing effects during development, both in low-risk and high-risk environments.

Open Peer Commentary Selfishness and sex or cooperation and family values? doi:10.1017/S0140525X09000028 Joshua M. Ackermana and Douglas T. Kenrickb a

Department of Psychology, Yale University, New Haven, CT 06520-8205; Department of Psychology, Arizona State University, Tempe, AZ 85287-1104. [email protected] [email protected] b

Abstract: Evolutionary models of behavior often encounter resistance due to an apparent focus on themes of sex, selfishness, and gender differences. The target article might seem ripe for such criticism. However, life history theory suggests that these themes, and their counterparts, including cooperation, generosity, and gender similarities, represent two sides of the same coin – all are consequences of reproductive trade-offs made throughout development.

Evolutionary approaches to human behavior, such as Del Giudice’s target article, have been accused of overemphasizing: (1) a limited (and often unsavory) set of topics, especially sex and mating; (2) selfishness; and (3) gender differences (see also Kenrick et al. 2003). Each of these accusations is superficially true but is misguided at a deeper level. The approach presented in the target article helps emphasize this latter point. 1. Sex. An evolutionary approach is fundamentally concerned with reproduction, but reproduction involves more than just sex. Life history theory presumes that all organisms are designed to allocate resources over the lifespan in a way that maximizes reproductive success. Mating effort is certainly a central phase of any organism’s life history. For salmon and century plants, which reproduce in a single burst and then die, the effort ends there. But for many organisms, including all mammals, the effort expended in attracting a mate and copulating is relatively minor compared to the effort devoted to parenting. In the case of humans, decades of parenting effort may precede decades of grandparental effort. From this perspective, reproductive strategy incorporates not only parent – child and parent – parent bonding, but also many other non-obvious aspects of life, including religious

attendance and political affiliation (e.g., Weeden et al. 2008). As suggested by Del Giudice, interpersonal attachment patterns may also reflect general reproductive strategies, and trade-offs between mating and parenting effort. 2. Selfishness. Selfish genes do not necessarily produce selfish organisms. Although genes tend to promote their own relative replication rates (Dawkins 2006; Williams 1966), those genes often construct organisms that make trade-offs favoring cooperation and kindness over narrow selfish behaviors. At the simplest level, people are often kind and generous to kin and to those with whom they expect to have future interaction (e.g., Ackerman et al. 2007; Burnstein et al. 1994). Economists and game theorists have been surprised at people’s willingness to cooperate with others even when payoffs favor competition, but given the diffusion of genetic returns generated by common human social arrangements, people’s generosity seems less irrational (Kenrick et al. 2008). For instance, a view of organisms as selfish might suggest that in a domain such as romantic relationship formation, which is so closely linked to genetic payoffs, competitiveness will rule the day. Yet, people cooperate in order to improve a range of reproductive goals (e.g., Ackerman & Kenrick, under review). Perhaps even more economically puzzling is the fact that people regularly turn down offers of resources from others, even when they are actually in need of help. Viewed within the framework of life history theory, such self-denial is often self-serving at a more ultimate level (Ackerman & Kenrick 2008). Indeed, “irrational” behaviors, such as cooperative courtship and refusing offered aid, may be consequences of the very same trade-offs that produce secure and insecure attachment patterns. 3. Gender differences. Gender differences reflect critically important trade-offs, but so do gender similarities. Life history theorists have noted that reproductive strategies often diverge within a species, many times along gender lines. The underlying cause of such variation – differences in minimal obligatory parental investment – can result in a wide range of sex differences, from mate preferences (for willingness to engage in short-term relationships; Li & Kenrick 2006) to benevolence (between friends; Ackerman et al. 2007) to attachment patterns (in insecure attachment; Del Giudice, target article). For instance, in research on cooperative courtship, we find evidence that parental investment differences may motivate women to prefer assistance in building thresholds for potential mates, but motivate men to prefer assistance in overcoming these thresholds (Ackerman & Kenrick, under review). At the same time, gender similarities reveal equally important tradeoffs. In biparental species (e.g., humans), many of the problems of parenting and mating are the same for males and females, and solutions to these problems play out through sex similarities in some of the same arenas that also exhibit sex differences – mate preferences (for partner qualities within short-term relationships; Li & Kenrick 2006), benevolence (between kin; Ackerman et al. 2007), and attachment patterns (in secure attachment; Del Giudice, target article). Indeed, in our studies of cooperative courtship, we find substantial similarities between men’s and women’s desires to cooperate, despite sex differences in how this cooperation materializes (Ackerman & Kenrick, under review). 4. Conclusion. Life history theory is garnering increased attention among behavioral researchers. At first blush, the evolutionary foundations of this theory might inspire familiar objections. On careful consideration, however, a theoretical and empirical response to these objections reveals even broader application for an evolutionary life history perspective. This approach suggests that variations in a wide range of phenomena (e.g., attachment patterns, cooperation and competition, selfishness and generosity) emerge from a common source: trade-offs concerning reproduction (but not just sex!). As Del Giudice’s target article implies, the next stage of inquiry involves a full conceptual integration of these important ideas with the range of phenomena in the behavioral sciences. BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

21

Commentary/Del Giudice: Sex, attachment, and the development of reproductive strategies

No reliable gender differences in attachment across the lifespan doi:10.1017/S0140525X0900003X Marian J. Bakermans-Kranenburg and Marinus H. van IJzendoorn Centre for Child and Family Studies, Leiden University, NL-2300 RB Leiden, The Netherlands. [email protected] [email protected] http://www.socialsciences.leidenuniv.nl/educationandchildstudies/ childandfamilystudies/organisation/bakermans-kranenburg.jsp http://www.socialsciences.leidenuniv.nl/educationandchildstudies/ childandfamilystudies/organisation/van-ijzendoorn.jsp

Abstract: In middle childhood, boys show more avoidant attachments and girls more ambivalent attachments as a prelude to gender differentiation in reproductive strategies. However, we have failed to find systematic and method-independent gender differences in middle or late childhood attachments, nor in adult attachment representations. We conclude that Del Giudice’s model rests on a brittle empirical basis.

Del Giudice’s model of different reproductive strategies for females versus males hinges critically on the assumption that boys show more avoidant attachments and girls more ambivalence in the developmental period after early childhood. Attachment research in infancy and early childhood did not detect gender differences, whereas, according to Del Giudice in the target article, the picture changes “dramatically” in middle childhood. The question we address here is: Does the picture indeed change, and if so, in what respect? Del Giudice lists seven studies to document this radical change. The three studies using the Coping Strategies Questionnaire (CSQ) should, however, be discounted as sources of evidence because patterns of attachment behavior and mental representations of attachment cannot be validly assessed by means of self-reports. Children as well as parents lack insight into their own attachment interactions and relationships, in

particular when they are insecurely attached. Their insecurity distorts their self-perception (for meta-analytical evidence, see Van IJzendoorn et al. 2004). Four pertinent studies remain (Del Giudice, in press; Granot & Mayseless 2001; Kerns et al. 2007; Toth et al., personal communication, October 19, 2007). Only the Kerns et al. (2007) study produced results contrasting with Del Giudice’s model, showing that female 9–11-year-olds were more often classified as avoidantly attached relative to male participants. The other studies pointed to the expected direction of insecure boys being more often avoidant and insecure girls more often ambivalent. In our Leiden Attachment Research Program, we assessed quality of attachment in two studies on 7-year-old children (Gilissen et al. 2008; Pannebakker 2007), and in one study on 14-yearolds (Beijersbergen et al., in press). Furthermore, after a brief literature search, we found pertinent studies by Ammaniti et al. (2000), Bureau et al. (2006), and Gloger-Tippelt and Koenig (2007) on children aged 6– 10 years. Following on Del Giudice’s focus on the secure, avoidant, and ambivalent categories, we excluded the category of disorganized attachment, or used forced classifications when available. The combined distribution of secure, avoidant, and ambivalent attachments across all samples (including those discussed by Del Giudice) is 49% secure, 37% avoidant, and 14% ambivalent attachments for boys. For girls, the distribution is 64% secure, 22% avoidant, and 14% ambivalent attachments (see Table 1). Haberman’s adjusted standardized residuals show significant differences between boys and girls for the secure (fewer boys) and the avoidant (more boys) classifications, but not for the ambivalent classification. Dividing the studies according to their assessment procedures (doll play narratives based on Bretherton et al. 1990; Cassidy 1988), observations of separation/reunion (Main & Cassidy 1988), and modified Adult Attachment Interview (AAI; Hesse 1999; Main et al. 1985), we only found gender differences in the set of studies using narratives (see Table 1). Apparently, the gender effect is measurement-specific, and systematic errors of measurement

Table 1 (Bakermans-Kranenburg & Van IJzendoorn). Distributions of attachment in middle childhood and adulthood Secure (n; %) Total, 6– 14 years Boys Girls Doll play narratives1 Boys Girls Observation (separation-reunion)2 Boys Girls AAI (modified for younger ages)3 Boys Girls AAI 4 Males Females

Avoidant (n; %)

Ambivalent (n; %)

Total (N)

Secure

Avoidant

n%

n%

n%

N

179; 49% 342; 64%

134; 37% 117; 22%

53; 14% 72; 14%

366 531

123; 51% 210; 67%

88; 36% 61; 20%

31; 13% 41; 20%

242 312

2 3.9 3.9

4.4 2 4.4

2 0.1 0.1

27; 69% 84; 69%

7; 18% 24; 20%

5; 13% 13; 11%

39 121

0.2 2 0.2

2 0.2 0.2

0.4 2 0.4

29; 34% 48; 49%

39; 46% 32; 33%

17; 20% 18; 18%

85 98

2 2.0 2.0

1.8 2 1.8

0.3 2 0.3

331; 48% 1507; 50%

230; 33% 901; 30%

126; 18% 589; 20%

687 2997

2 1.0 1.0

1.8 2 1.8

2 0.8 0.8

 Significant 1

Haberman’s adjusted standardized residuals 2 4.6 4.8 4.6 2 4.8

Ambivalent residuals

0.4 2 0.4

adjusted standard residuals in bold Del Giudice (2008); Gilissen et al. (2008); Gloger-Tippelt et al. (2007); Granot & Mayseless (2001); Kerns et al. (2007); Toth et al. (2006) 2 Bureau et al. (2006); Pannebakker (2007) 3 Beijersbergen et al. (in press); Ammaniti et al. (2000) 4 Van IJzendoorn & Bakermans-Kranenburg (in preparation)

22

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

Commentary/Del Giudice: Sex, attachment, and the development of reproductive strategies might be the source of these differences. Narrative procedures might focus more on content than on coherence, contaminating formal avoidance with substantive “macho” accounts. In addition, differences in verbal abilities between boys and girls in this age group may play a role (Verschueren & Marcoen 1999). It should be noted that in Granot and Mayseless’ (2001) study, secure children (mostly girls) had significantly higher scores on cognitive achievement, and avoidant children (predominantly boys) had the lowest scores on cognition. Furthermore, the gender difference in studies with doll play narratives is restricted to boys showing more avoidant attachments, and not less or more ambivalence than girls. In the narrative approach, boys do follow the predicted pathway of avoidance, but girls do not prefer the ambivalent attachment strategy (see Table 1). In the observational studies, no gender differences in middle childhood have been documented. In the two studies using the AAI, boys were only slightly under-represented in the secure category and were not significantly over-represented in one of the insecure categories (Table 1). It should be noted that numerous studies on attachment representations in adulthood using the Adult Attachment Interview – the gold standard for assessing attachment representations, which is independent of cognitive abilities (Hesse 1999) – have not come up with any replicable gender differences in dismissing versus preoccupied attachments. In a meta-analysis of studies using the AAI (Van IJzendoorn & Bakermans-Kranenburg 1996), and in a recent update (Van IJzendoorn & Bakermans-Kranenburg 2008; in preparation), we traced 17 studies on 687 males and 59 studies on 2997 females (mostly parents of reproductive age). The distribution of males was 33% dismissing, 48% secure, and 18% preoccupied attachments, while the distribution of females was 30% dismissing, 50% secure, and 20% preoccupied attachments. No significant gender difference was found (see Table 1). We conclude that Del Giudice’s model has a brittle empirical basis. We did not find systematic and method-independent gender differences in middle- or late-childhood attachments, nor were any gender differences in adulthood present in studies using the gold standard to assess attachment representations. The speculative model badly needs some repairs to accommodate with an obstinate empirical reality.

Attachment, reproduction, and life history trade-offs: A broader view of human mating doi:10.1017/S0140525X09000041 Lane Beckes and Jeffry A. Simpson Department of Psychology, University of Minnesota, Minneapolis, MN 55455. [email protected] [email protected]

Abstract: In this commentary, we attempt to broaden thinking and dialogue about how our ancestral past might have affected attachment and reproductive strategies. We highlight the theoretical benefits of formulating specific predictions of how different sources of stress might impact attachment and reproductive strategies differently, and we integrate some of these ideas with another recent evolutionary model of human mating.

According to Del Giudice’s model, sex differences in insecure attachment orientations should emerge most strongly when individuals are exposed to moderate levels of stress during development. In such environments, males should be more likely to develop avoidant attachment orientations, and females ought to develop anxious/preoccupied orientations, especially from middle childhood extending into early adulthood. There are two potential problems with this claim. First, valid attachment measures have proven difficult to develop for middle childhood

(Kerns 2008), which may partially explain the dearth of studies focusing on this age group. Second, gender differences are rarely found or are quite small in adult samples, regardless of whether self-report or interview measures of attachment are used (Crowell et al. 1999/2008). It is not clear how Del Giudice’s model reconciles these issues. When contemplating how stress impacted social development in the environment of evolutionary adaptedness (EEA), one must consider the different sources of stress our ancestors faced and which specific sources were most strongly associated with certain environmental risks, given humans’ comparatively long life histories. According to life history theory (Kaplan & Gangestad 2005), organisms make trade-offs between current versus future reproduction, quality versus quantity of offspring, parental investment versus offspring genetic quality, and reproduction versus survival over the lifetime. Different sources of stress might have had unique implications for certain life history trade-offs and, ultimately, the adoption of certain adult reproductive strategies. Four prevalent environmental threats during evolutionary history were predation pressures, the availability of food, the prevalence of diseases, and intergroup or intragroup conflict (Simpson & Belsky 2008). Each type of threat could have produced a unique source of stress that affected how life history trade-offs were made. For example, if diseases were prevalent in the local environment, most men might have placed more weight on either offspring quantity or their genetic quality, contingent on other mating factors. Most women, however, should have emphasized offspring genetic quality on account of the more limited lifetime reproductive capacity of women and the need to bear the most disease-resistant offspring. These tradeoffs may have oriented both sexes toward greater avoidance, which might have facilitated greater male promiscuity and more male – male intrasexual competition (Schmitt 2005b). Women should have been more strongly attracted to mates who displayed better health or more viability, and women should have competed more intensely to reproduce with these men, putting less emphasis on paternal investment. In environments characterized by intergroup conflict, many men may have shifted to a shorter-term, higher-quantity mating strategy, given the greater risk of death in ancestral men (Cronin 1991). Most women, however, may have worked to keep their mates invested to secure more protection for themselves and their offspring. Adopting an avoidant orientation might have been the best way for most men to increase their fitness in these environments, whereas an anxious/preoccupied orientation might have been the best strategy for most women. The main point is that different sources of stress might have shifted men and women in the same or in different directions with respect to adult attachment orientations and reproductive strategies, even if the absolute levels of stress experienced during development were similar. Although the purpose of the target article is to present an integrated life-history-based evolutionary model of reproductive strategies, relatively little is said about how life history tradeoffs might intersect with proximal (current) trade-offs. In section 6.3.1 of the target article, the author notes that “women cannot shift the balance between parenting and mating effort as easily as men.” This is precisely why ecologically contingent evolutionary models of human mating such as the Strategic Pluralism Model (SPM; Gangestad & Simpson 2000) have been developed. Rather than viewing human mating in terms of how an individual’s history results in specific adult reproductive strategies, SPM proposes that women evolved to make mating decisions on the basis of the nature of their current local environments. Specifically, women should place relatively greater weight on men’s viability (i.e., their health, vigor, and ability to withstand diseases), especially in pathogen-prevalent environments. However, they should place greater emphasis on men’s ability and willingness to invest in themselves and their offspring when local BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

23

Commentary/Del Giudice: Sex, attachment, and the development of reproductive strategies environments require more biparental care. Indeed, it is precisely because women cannot easily “shift the balance” between parenting and mating effort that they should have evolved to select mates by making judicious trade-offs between these mate attributes. SPM, therefore, explains how and why women evolved to make the best of one “constraint” imposed on them by nature – the unique way in which they reproduce. A complete evolutionary account of human mating must consider how and why both ontogenetic (life history) and proximal (current environmental) factors led both sexes to make adaptive decisions with respect to the allocation of mating effort versus parenting effort. How might life-history and ecological-contingency models intersect? An individual’s developmental history could set thresholds for judging the acceptability of a mate’s viability or investment potential in adulthood. For example, females exposed to early life stress stemming from prevalent diseases should have placed greater value on – and may have had higher thresholds of acceptability for – a mate’s degree of viability. Conversely, women should have placed more weight on the willingness and ability of mates to invest if the primary source of early stress was poor or unpredictable food supplies. The important point is that the specific source of psychosocial stress in a person’s past could influence her or his tendency to value, attend to, and hold higher or lower standards of acceptability for a potential mate’s viability in relation to investment potential, or vice versa. These standards may also have influenced decisions about whether and when to terminate relationships. Though taken for granted in modern environments, our ancestors faced major obstacles in raising even a few offspring to adult reproductive age. Selection pressures should have led people to generally make adaptive trade-offs on the basis of reproductively relevant events that occurred earlier in their lives. However, selection pressures should also have led people to make adaptive trade-offs in response to reproductively relevant conditions in their current environments. Long-term fitness returns might actually have been more strongly linked to the mating and parenting trade-offs that individuals made in response to their current environments in adulthood than to more distant lifehistory factors, especially if factors that affected mating or parenting changed within a person’s lifetime (e.g., the prevalence of disease, sex-ratios, changes in the food supply).

“Fatal attraction” syndrome: Not a good way to keep your man doi:10.1017/S0140525X09000053 Anne Campbell Psychology Department, Durham University, Durham DH1 3LE, United Kingdom. [email protected]

Abstract: Female behavior that is driven by ambivalent attachment is far from passive or withdrawn. As dramatised in the movie “Fatal Attraction,” such women’s emotional hyper-reactivity is often expressed in violence, which is antithetical to securing investment from mates or peers. Single motherhood, rather than reflecting an avoidant strategy in which close relationships are devalued, is often the result of ecological conditions in which paternal investment is desired but unavailable.

Does the “fearfulness, withdrawal, and passive behavior” that Del Giudice notes (target article, sect. 6.3.2, para. 2) is associated with an ambivalent attachment style, really assist women in retaining their partner and in maximising paternal investment? Preoccupied attachment is characterised by intense desire for closeness, discomfort when not intimately involved with another, and nagging worry about rejection. It is a short step from here to the clinical condition of borderline personality

24

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

disorder (BPD), with its pervasive relationship instability and frantic efforts to avoid separation or abandonment. Initial idealisation of the target is coupled with demands for their exclusive attention, but at the first sign of real or imagined rejection, the emotion switches from infatuation to bitter devaluation. Dependent and borderline personality traits characterise up to 50% of male perpetrators of partner violence, with these men’s extreme dependency resulting in a violent response to the wife’s perceived rejection or insubordination (Holtzworth-Munroe & Stuart 1994). Following the belated recognition of symmetry in partner violence, women’s partner violence has also been related to attachment style. Women receiving mandated treatment for domestic violence show elevated rates of both anxious and avoidant attachment styles and evidence of borderline personality traits (Goldenson et al. 2007). Perhaps because 75% of BPD sufferers are women, there has been a tendency to consider it as a predominantly internalising disorder by emphasising the diagnostic criteria of chronic feelings of emptiness, suicidal behaviour, and self-mutilation.But there are other externalizing diagnostic criteria, including affective instability, impulsivity, and “inappropriate, intense anger or difficulty controlling anger (e.g., frequent displays of temper, constant anger, recurrent physical fights)” (American Psychiatric Association 2000, p. 710). Self-harm is more strongly associated with an avoidant attachment style, while it is anxious attachment that is associated with reactive, other-directed intimate aggression (Critchfield et al. 2008). College women with attachment styles characterised by high anxiety and low avoidance are more likely to report having used violence than women who are high in both styles (Orcutt et al. 2005). These violent reactive outbursts, characteristic of BPD, have been linked to hyper-responsiveness of the amygdala and modulatory failure of the prefrontal cortex (Siever 2008). In short, although Del Giudice suggests that “anxious behaviors (e.g., dependence, preoccupation with intimacy and partner availability) can help to maintain closeness with one’s partner” (sect. 6.3.1, para. 4), there is evidence that insecurity of this kind can have exactly the opposite effect. With regard to same-sex relationships, I agree with Del Giudice that the adaptive benefits of an anxious attachment style for women are unclear. That anxious traits are “not nearly as damaging for females as they are for males” (sect. 6.3.2, para. 2) is hardly a ringing endorsement of their functionality. Female friendships are characterised in the psychological literature as more exclusive, self-disclosing, supportive, and lacking the competitive edge of male groups (see Rose & Rudolph 2006). However, this rosy view belies a more complex picture. A preoccupation with friendships and emotional intimacy fuels girls’ need for social approval, increases possessive feelings and jealousy about best friends, and results in depression when friendships terminate (which they do with greater frequency among girls than among male friends). If these are the downsides of “normal” girls’ friendships, they are surely multiplied for anxiously attached girls. Low self-worth is associated with heightened “best friend” jealousy, loneliness, social rejection, and aggression (Parker et al. 2005). Conflict between teenage girls is often driven by rivalry about attractiveness to the opposite sex and “ownership” of desirable boys (Campbell 1995). If anxious girls accord high strategic priority to securing male investment, this should increase their willingness to compete for it, drawing them into indirect or direct aggression against their peers. Before considering a female’s avoidant strategy, I need to pose a more wide-ranging question about life history strategies. What is their psychological status? Attachment security might shape later reproductive strategy in several ways. Early experience might affect personality traits, molding an individual’s general level of anxiety or avoidance in response to others. These traits would alter interpersonal competence (the ability to form stable relationships). If early experiences build a “behavioralmotivational” system, they might act through preference, by causing the individual to value some relationship styles more

Commentary/Del Giudice: Sex, attachment, and the development of reproductive strategies than others, as exemplified by Del Giudice’s statement “preoccupied individuals eagerly look for intimate relationships” (sect. 5.2, para. 4). More cognitively, early experiences may lead to different “decisions concerning reproductive strategy.” The implication here is that individuals weigh the costs and benefits of different strategies (with those weightings informed by early experience) to arrive at some personal utility function. In short, is the link between early experience and later strategy mediated by personality (avoidant women are unable to cope with long-term relationships), preference (avoidant women prefer short-term sexual encounters), or cognition (avoidant women decide that the net utility of pair-bonding is lower than that of single motherhood)? Psychologically damaged individuals are certainly poor at maintaining stable relationships, but this is not equivalent to preferring short-term relationships or devaluing stable relationships. Among girl gang members, most of whom came from very disturbed families, I was surprised by the extent to which they idealised marriage and traditional family values (Campbell 1992). Their chaotic and marginal lifestyles, coupled with volatile emotions, often conspired to defeat long-term relationships; but there was little doubt that the girls very much aspired to them (LeBlanc 2004). As for an active preference for shortterm encounters, recent data suggest that women find onenight stands very much less emotionally satisfying than do men (Campbell 2008b). Young single mothers rarely choose their situation: Local female-biased sex ratios (resulting from male death, imprisonment, and addiction) create a paucity of men who are able or willing to make paternal investment (Campbell 1995). “Avoidance” may be less of a female strategy than a default option forced by ecological circumstance.

Avoidant strategy in insecure females doi:10.1017/S0140525X09000065 Bin-Bin Chen and Dan Li Department of Psychology, Shanghai Normal University, Shanghai 200234, China. [email protected] [email protected]

Abstract: This commentary cites evidence to argue that girls growing up in a competitive and aggressive environment are more likely to shift to avoidant attachment than to ambivalent attachment in middle childhood. These avoidant women are also more likely to favor a shortterm mating strategy. The role of oxytocin (OT) and early experience in shaping an avoidant attachment in females is also discussed.

Del Giudice’s evolutionary model integrating attachment with human reproductive strategies contributes to our understanding of the adaptive significance of gender differences in attachment, and of the role of sex-related endocrine mechanisms on development of attachment styles. He argues that gender differences in attachment patterns emerge during middle childhood, and specifically, that insecure girls tend to develop ambivalent attachment, whereas insecure boys develop avoidant attachment, when faced with new social demands driven by the peer group. If this argument is correct, we should expect that insecure girls, similar to boys, would be more likely to shift to avoidant attachment in middle childhood. Thus, this commentary cites evidence supporting why this shift to avoidant attachment is essential. We follow Del Giudice’s logic in analyzing the emergence of aggression as a correlate of avoidant attachment in girls at this stage of childhood. According to Del Giudice’s analysis of attachment styles in middle childhood, insecure children without a secure family environment must “begin to fight their way through social

reality” (target article, sect. 7.1.2). Therefore, insecure girls are more likely to shift to avoidant styles than ambivalent styles,1 which emerge from competition in the peer group. The behavioral correlates of avoidant attachment (i.e., aggression) are already selected at this stage because they are more adaptive than traits associated with ambivalent attachment (e.g., dependency, behavioral immaturity, and passivity in peer relations) in terms of independent resource acquisition because insecure girls cannot rely on their parents under extreme high-risk environmental conditions. Guttmann-Steinmetz and Crowell (2006) provide a conceptual model demonstrating that externalizing behavior by children develops from parental failure to provide a secure base to protect children from environmental risks. With the recent development of measures of attachment quality during middle childhood (Finnegan et al. 1996; Kerns et al. 2000), a line of research has emerged suggesting that girls’ aggression is associated with avoidant rather than ambivalent attachment style. No gender differences were found in the association between avoidant attachment and aggression (Davies & Forman 2002; Granot & Mayseless 2001). Finnegan et al.’s (1996) research showed that for both sexes, aggression was predicted by an avoidant strategy but not by a preoccupied strategy, according to their specific linkage hypothesis. More recent research (BoothLaForce et al. 2006) showed that avoidant coping in relation to the mother, and lower security with the father, was related to aggression. These associations did not vary by sex of child. In addition, in order to reduce the risk of retaliation, girls are more likely to use subtle forms of aggression such as relational aggression or coercive and prosocial strategies of resource control (i.e., bi-strategy; Hawley 2007) to compete for resources, such as better-resourced men (Campbell 1999). Active reaction (the correlates of avoidant attachment) to secure a scarce resource is likely to be more effective than a passive one (the correlates of ambivalent attachment). We would expect insecure females to be more likely to adopt avoidant rather than anxious strategies to maximize current reproductive success, no matter how challenging the environment is,2 because there is very limited support for the proposal that an anxious strategy enhances partner investment or reproductive success. However, “avoidant” women who show little desire for commitment and adopt a short-term mating strategy (Feeney 1999; Gentzler & Kerns 2004) by mating with multiple males could acquire considerable resources and simultaneously reduce the possibility of future male attacks against her and her offspring (Hrdy 1981). In accordance with Del Giudice’s hormonal basis of the middle childhood transition, we now consider the evidence for the role of oxytocin (OT) in females’ avoidant attachment. OT is closely implicated in female behavior because its effects are strongly modulated by estrogen. Early social experience can alter social behavior by affecting the development of neuroendocrine systems including OT (see Cushing & Kramer 2005, for a review). In rats, female offspring raised by high-licking and high-grooming mothers show a significant increase in oxytocin receptors, OTRs (Francis Young et al. 2002). Furthermore, changes in the oxytocinergic system in response to social interaction could alter brain development and, thereby, the subsequent expression of social behavior. However, OT has little or no effect on regions of the brain in adults due to a lack of receptors, compared with before adolescence (Cushing & Kramer 2005). Thus, middle childhood appears to be the “last chance” for OT to exert long-lasting effects on behavior. In a high-risk environment lacking sufficient parental investment, we would expect to see a decrease in oxytocin affecting the childhood organization and development of the brain and resulting in an increase in aggression among girls with insecure attachments. In an environment in which well-resourced males are in short supply, it would seem adaptive that female competition and female assault (see Campbell 1999) would be primed BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

25

Commentary/Del Giudice: Sex, attachment, and the development of reproductive strategies by epigenetic mechanisms of early social experience (Cushing & Kramer 2005). Overall, we welcome and accept most of Del Giudice’s arguments as a significant contribution to our understanding of the development of attachment from an evolutionary perspective. However, Del Giudice’s analysis of insecure attachment in females should include avoidant pattern, which should be further examined empirically. ACKNOWLEDGMENTS We thank Anne Campbell and Lei Chang for comments on this manuscript. NOTES 1. Del Giudice admits that it is not clear whether ambivalent pattern is adaptive to girls (see target article, sect. 7.1.2). 2. According to Del Giudice’s argument, there is no opportunity for insecure girls to shift to a secure attachment even in a secure environment because insecure attachment patterns after the middle childhood transition are expected to be stable in adulthood.

Attachment and life history strategy doi:10.1017/S0140525X09000077 Aurelio Jose´ Figueredo, Jon A. Sefcek, and Sally G. Olderbak Department of Psychology, University of Arizona, Tucson, AZ 85721-0068. [email protected] [email protected] [email protected] www.u.arizona.edu/~ajf

Abstract: Del Giudice addresses a complex and pertinent theoretical issue: the evolutionary adaptiveness of sex differences in attachment styles in relation to life history strategy. Although we applaud Del Giudice for calling attention to the problem, we regret that he does not sufficiently specify how attachment styles serve as an integral part of a coordinate life history strategy for either sex.

Del Giudice’s target article addresses a complex and pertinent theoretical issue: the evolutionary adaptiveness of sex differences in attachment styles as they relate to life-history strategy. This is an extremely important theoretical problem in evolutionary psychology, and we applaud Del Giudice for bringing it to the attention of the field; but we are not convinced that the present article makes a very substantial contribution to its resolution. Although the theory seems plausible, Del Giudice does not make the necessary theoretical connections to sufficiently integrate these two constructs. One major limitation is that Del Giudice mentions certain important issues that are critically relevant to the topic, but rather than explaining them or taking a position on them, he claims that they are beyond the scope of the article. For example, he does not address whether or not attachment patterns are categorical or continuous, and instead references the work of Fraley and Spiker. Del Giudice must take a stand on this issue if he is going to take an evolutionary stance on attachment patterns based on individual differences – in our understanding, specifically in the direction of the patterns being continuous. He also does not fully address the degree of persistence versus plasticity of “internal working models” across the life cycle, stating that “a detailed treatment of the issue is beyond the scope of this article” (sect. 2.3, para. 1). Without explaining these issues in more detail, the conclusions seem based more on belief than on a principled deliberation on the evidence. Aside from the lack of explanation, Del Giudice also makes contradictory statements without offering a resolution. For example, Del Giudice discusses the affect of increasing age on life history trade-offs, with aging men switching from mating effort to parental investment. He argues that “this transition may be especially dramatic in insecurely attached men” (sect.

26

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

7.1.3, para. 3). However, prior to this he discusses a couple of studies finding that “short-term mating in men tended to increase with age, and men were on average more oriented with short-term mating regardless of attachment style” (sect. 6.4.1, para. 2). This is an apparent contradiction that is left unresolved. Del Giudice also discusses “disorganized” attachment style early in the article, stating later that it will not be integrated into his synthesis on account of three concerns about the concept. However, in the very next paragraph he mentions it again, arguing that it is the only infant attachment pattern showing a sex difference. If one of the main points of this article is to discuss theoretical reasons for sex differences in attachment, this issue may be of critical importance. Granted, Del Giudice suggests that this is an important area of research for the future, but because he has written this BBS target article specifically on the topic of evolutionary perspectives on sex differences, he needs to offer his perspective on why that one notable infant sex difference may exist. One might argue, for example, that an infant’s attachment style has a different function than an attachment style displayed during adolescence or young adulthood. At each point in time, the goals and relationship of the child with their parent differ, and so we would expect attachment styles to vary as well. The attachment style towards romantic partners, friends, and other members of society would also require a different relationship, and possibly a different style of attachment. It seems that at each developmental stage, one would want or need different things from the relationship, so the same attachment style might not always be appropriate. On the other hand, there are also reasons to expect some degree of stability in attachment styles over time. For example, Del Giudice addresses genetic influences on attachment styles during the course of development, but fails to properly discuss genetic predispositions (e.g., temperament) of the child which might be present at birth, thus setting a baseline on which to build during development. Del Giudice partially addresses the stability of attachment style over time, but does not suggest which factors might differentially influence those individuals who do and who do not have stable attachment styles over time, given the probable existence of such predispositions. As Del Giudice is aware, our laboratory has published multivariate models of latent life-history constructs that support the cross-situational consistency of attachment styles across both childhood attachment to parents and adult attachment to romantic partners (Figueredo et al. 2004; 2006; 2007). This latent variable has been shown to have a high heritability coefficient (h 2 ¼ .65) and is probably a good candidate both for early childhood predispositions that shape later developments and an individual difference variable that predicts temporal stability versus instability in attachment styles over time. Although Del Giudice cites this work, he does not pursue its implications for the resolution of this theoretical problem. There are also various miscellaneous concerns to which we would have liked Del Giudice to respond. For example, we disagree with the statement (in sect. 3.1, para. 2) where Del Giudice writes, “In order to be selected for, traits need to solve two problems: being adaptive at the present time and being adaptive in the future of the organism.” Selection has nothing to do with the future of the individual organism. It has to do with what worked in the ancestral past (teleonomy). Does the author understand this principle, or is this just a very poorly worded sentence? We are also confused by Del Giudice’s analysis of polygyny and attachment. We can understand how the arguments apply to serial polygyny, where the father ends one relationship and moves on to another one, potentially decreasing the resources invested in the previous family. However, we do not understand how this relates to societies with simultaneous polygyny, where the father will continue to invest in previous wives/families, even if it is only through material resources rather than bioenergetic ones.

Commentary/Del Giudice: Sex, attachment, and the development of reproductive strategies Despite our criticism of the target article, we do feel that it is important to make the issue of evolutionary bases for sex differences known to a wider audience. However, in its present execution, the article seems to be lacking in ability to resolve some of the key issues concerning this topic.

Evolution of neuroendocrine mechanisms linking attachment and life history: The social neuroendocrinology of middle childhood doi:10.1017/S0140525X09000089 Mark V. Flinn,a Michael P. Muehlenbein,b and Davide Ponzic a

Department of Anthropology, University of Missouri, Columbia, MO 65211; Department of Anthropology, Indiana University, Bloomington, IN 47405; c Neurobiology Program, Department of Biological Sciences, University of Missouri, Columbia, MO 65211. [email protected] [email protected] [email protected]. b

Abstract: An extended period of childhood and juvenility is a distinctive aspect of human life history. This stage appears to be important for learning cultural, social, and ecological skills that help prepare the child for the adult socio-competitive environment. The unusual pattern of adrenarche in humans (and chimpanzees) may facilitate adaptive modification of the neurobiological mechanisms that underpin reproductive strategies. Longitudinal monitoring of DHEA/S in naturalistic context could provide important new insights into these aspects of child development.

Living organisms are flexible; they can respond to changing conditions with a variety of morphological, physiological, and behavioral mechanisms. The processes that organisms use to change and respond to environmental challenges are posited to be evolved adaptations (West-Eberhard 2003). The development of the psychology of the human child may be viewed as a complex set of flexible adaptations (Bjorklund & Pellegrini 2002). Del Giudice proposes a comprehensive model for how humans might adjust future reproductive strategies – such as timing of puberty, mate choice, and parental behavior – in response to environmental conditions during middle childhood. His ideas are exciting because they suggest new research directions, including investigation of the stimuli that affect release of adrenal androgens (DHEA/S), and the associated effects on neurobiological development. The human child must master the dynamics of social networks and culture, supported by the extraordinary information-processing capacities of the human brain (Adolphs 2003; Roth & Dicke 2005). We are interested in the unusual sensitivity of the fetus and child to the social environment – interpersonal relationships – and the consequent changes that occur in neuroendocrine systems. Our curiosity is piqued by both the paradoxical nature of this phenomenon – for some hormonal responses have attendant somatic costs (e.g., Flinn 2006; Muehlenbein & Bribiescas 2005) – and its importance for human health. For example, maternal depression and high levels of social anxiety during pregnancy are associated with low birth weight, elevated stress reactivity, and subsequent disease risk for offspring (Barker 1998; Gluckman & Hanson 2006; Weinstock 2005). The processes that underlay this biological embedding of information from the social environment in humans remain obscure. On the basis of life history theory, the delayed reproductive maturity represented by an extended period of childhood and juvenility in humans is predicted to be important for learning cultural, social, and ecological skills that help prepare the child for the adult socio-competitive environment (Flinn et al. 2005; Geary & Bjorklund 2000). During this developmental period,

boys and girls show behavioral sex differences in play and social interactions: boys tend to invest more time in organizing groups of peers, among which they form hierarchies, and compete with other groups. Conversely, girls usually invest more time in dyadic interactions with similar-age girls, caring for siblings, and doing domestic chores (Geary & Flinn 2002; Quinlan et al. 2003). How the onset of male coalitional and female dyadic psychobiology and life history trajectories might be influenced by family environment is yet an open question. In the target article, Del Giudice draws our attention to the life history stage of middle childhood, which has not received much attention in comparison with infancy and early childhood on the one hand, and adolescence and the transition to early adulthood on the other. Yet it is childhood that distinguishes humans from other primates (Bogin 1999), and it is the period during which some of the most uniquely human socio-cognitive abilities are developed and refined (Geary 2005b). Adrenarche appears to be a key neuroendocrine transition in middle childhood that may facilitate the ontogeny of the neural mechanisms that underpin human sociality (Campbell 2006). Humans (and chimpanzees), moreover, have an unusual pattern to adrenarche. A few months after birth, the fetal zone of the adrenal cortex disappears and the levels of DHEA(S) diminish to near zero. In other primates, such as macaques and baboons (Muehlenbein et al. 2003; Nguyen & Conley 2008), newborns have high levels of DHEA(S) (compared to chimpanzees and humans) that slowly and steadily decrease with age. In humans, adrenarche typically begins around age 7 years for females and 9 years for males (Dhom 1973), and DHEAS levels peak earlier in females than in males (Orentreich et al. 1984; Sulcova et al. 1997). The zona reticularis of the adrenal gland gradually begins developing at around age 3 years, at which time the production of DHEA(S) begins (Palmert et al. 2001; Remer et al. 2005). DHEA(S) is produced by the adrenal gland and converted to DHEA within target cells (Labrie et al. 1998). DHEA can then be converted into other androgens, such as androstenedione, as well as estrogens. DHEA acts antagonistically to cortisol (Hennebert et al. 2007; Kimonides et al. 1999), and both are derived from the pregnenolone precursor (Rainey et al. 2002). Timing of adrenarche also appears to be sensitive to environmental conditions; low-quality parental investment predicts earlier onset of adrenarche (Ellis & Essex 2007). Infants with a lower birth weight also exhibit earlier age at adrenarche (Ong et al. 2004). The links between adrenarche and the socio-cognitive demands of middle childhood are uncertain. Androgens associated with adrenarche are likely important for brain maturation (Campbell 2006). DHEA(S) is a neurosteroid expressed in the developing brain in a region-specific fashion and involved in regulating the organization of the neocortex (Compagnone & Mellon 1998). While brain growth in humans is almost complete by age 7 years (the typical time of onset of adrenarche), cortical maturation with synaptic pruning continues (Gogtay et al. 2004). Glucose metabolism rates in the brain also remain high until around age 10 years (Chugani et al. 1998). DHEA(S) promotes neurological functions in rodents (Karishma & Herbert 2002), and DHEA(S) binds to various receptors in the human brain, including GABAa (Majewska et al. 1990). DHEA(S) may also affect mood in humans (Arlt et al. 2000; Micheal et al. 2000). Functional connections between the middle childhood attachment transitions and adrenarche, however, remain speculative; the timing could be coincidental. DHEA(S) production has several physiological roles in childhood, including muscle and bone growth (Zemel & Katz 1986) and immuno-stimulation (Chen & Parker 2004). Rises in DHEA(S) during adrenarche may also be important in the hypothalamic desensitization associated with the onset of puberty (Tanner 1978). “Attachment” in humans appears to involve additional functions beyond security and protection; the flow of information from parents and other relatives, and recruitment into kin-based BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

27

Commentary/Del Giudice: Sex, attachment, and the development of reproductive strategies coalitions, emerge as critical challenges for the child. Predicting what one’s social environment will be as an adult and modifying phenotypic trajectories of the hormonal, neurological, and psychological mechanisms that comprise “internal working models” seems extraordinarily complex, and unlikely to favor early canalization of reproductive strategy. A more flexible system that allows inclusion of input throughout childhood and adolescence would have advantages over one primarily contingent on conditions during infancy (cf. Belsky 2002; Draper & Harpending 1982; Quinlan & Flinn 2003). To evaluate the exciting new ideas about the adjustment of reproductive strategies in middle childhood proposed by Del Giudice, it would be useful to have research designed to specifically evaluate causal relations among the key proximate mechanisms. Comparisons with other primates will help identify what aspects of human adrenarche are shared or derived. Analyses of patterns of attachment and adrenarche in apes would be most interesting. Longitudinal studies of human child development that monitor DHEA(S) levels in naturalistic context could provide detailed data on links among hormone levels, family environment, affiliative bonds, and long-term outcome measures.

Attachment strategies across sex, ontogeny, and relationship type doi:10.1017/S0140525X09000090 Cari D. Goetz, Carin Perilloux, and David M. Buss Psychology Department, University of Texas at Austin, Austin, TX 78712. [email protected] [email protected] [email protected] http://homepage.psy.utexas.edu/homepage/students/Perilloux/ http://www.davidbuss.com/

Abstract: We propose that middle childhood female ambivalent attachment, given the adaptive problem of uncertainty of future investment, is designed to evoke immediate investment from current caregivers, rather than new investment sources. We suggest greater specificity of strategic attachment solutions to adaptive problems that differ by sex, time, and relationship type.

The target article represents a major theoretical contribution on several fronts. First, it highlights evolutionary functional accounts as indispensable for any comprehensive theory of attachment (as initially envisioned by Bowlby [1969/1982], but neglected by many subsequent attachment theorists). Second, it focuses on sex differences in adaptive problems and the resultant attachment-related strategic solutions as males and females enter the arena of mate competition. And third, it provides an elegant theoretical integration of the evolutionary psychological work on sex differences in mating strategies with important dimensions of individual differences – something urgently needed, but relatively neglected by much previous work in evolutionary psychology (Buss & Greiling 1999). In this commentary, we build upon these important advances, and propose some directions for exploring additional attachment differences across time, sex, and adaptive problem domain. Evolution by selection tends to produce domain-specific and context-specific adaptations. It is theoretically problematic to assume that the attachment strategy as an adaptive response to environmental cues at one point during development will be adaptive if implemented in interactions in relationships later in life. To the degree that mother –infant attachment bonds serve functions that differ from those of friendship bonds and mateship bonds, and to the degree that they differ by sex, we expect selection to favor specificity of strategic solutions by relationship type, sex, and life phase.

28

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

It is reasonable to hypothesize that the sex differences in relative frequencies of insecure attachment styles that emerge at middle childhood are a result of sex differences in adaptive problems faced recurrently over deep evolutionary time during this stage of development. Del Giudice provides a compelling functional account for the shift in males towards an avoidant attachment style, but notes that the function of the female skew towards an ambivalent attachment style at this age is “less clear” (sect. 7.1.2, para. 2). If we focus on the finding that girls of parents who provide only inconsistent and unpredictable investment are the ones who tend to develop ambivalent attachment styles, we can conceptualize their ambivalent attachment psychology during middle childhood as a solution to the adaptive problems predicted by their parent-specific and kin-specific input. The ambivalent attachment style is characterized by extreme dependence, emotional instability, and a desire to exert influence over the caregiver (sect. 2.2). These strategies, perhaps effective in eliciting as much investment as possible from parents, are unlikely to succeed in establishing strong alloparental bonds if directed toward same-sex peers during middle childhood. There is evidence that female–female friendships during childhood are shorter and more fragile than male–male friendships (Benenson & Christakos 2003). If female friendships at this age lack the stability to endure until reproductive age, it strains credulity that they function as precursors to lasting alloparental bonds. Exhibiting high dependence and emotional instability toward female friends could easily backfire, as mechanisms that caused individuals to invest in non-kin who deplete resources, but who are unable or unlikely to reciprocate, would be selected against (Trivers 1971). From this perspective, we would not predict that attitudes and characteristics associated with ambivalent attachment at middle childhood would result in behavior that successfully initiates or maintains same-sex peer relationships. If we instead hypothesize that this strategy is designed to be directed towards parents or kin, a functional hypothesis presents itself. In an environment in which male parental investment is inconsistent or nonexistent, such as that associated with ambivalent attachment, females who waited longer to reproduce, or who attempted to obtain all the necessary investment from their future mate, would have been out-competed by females who began reproducing early and who attempted to extract resources and investment from kin (Belsky 1997a; 1999; Hoier 2003; Quinlan 2003). Moreover, the earlier a female reproduces, the more likely she is to receive grandparental assistance and resources. Given cues to low or unstable male parental investment, one adaptive solution would be to reproduce early and maintain high levels of dependence upon close kin. Because kin are usually already invested in the survival and reproductive success of their genetic relatives, we hypothesize that the care-eliciting strategies associated with ambivalent attachment directed specifically toward kin have the best chance of successfully extracting investment for a female’s offspring. In contrast, similar strategies directed toward peers may result in alienation. In an environment in which male parental investment is unreliable or unlikely to be forthcoming, reproducing early capitalizes on as much grandparental investment as possible while these extended kin are still alive and can invest. Whereas securely attached individuals expect consistent support from family members, insecurely attached females may do best to focus on obtaining as much support as possible during middle childhood, because they cannot rely on it for the future. Building on the theoretical advances proposed by Del Giudice, we have focused on functional explanations for the female switch to ambivalent attachment patterns during middle childhood. We suggest that this domain-specific approach will yield a psychological harvest for each sex, life stage, and relationship type. To the degree that sex-specific adaptive problems are associated with different types of dyads such as mateships, friendships, and kinships, we expect that selection will favor

Commentary/Del Giudice: Sex, attachment, and the development of reproductive strategies sex-specific, relationship-specific, and life-stage-specific strategic solutions.

Co-regulation of stress in uterus and during early infancy mediates early programming of gender differences in attachment styles: Evolutionary, genetic, and endocrinal perspectives doi:10.1017/S0140525X09000107 Sari Goldstein Ferber Division of Developmental Neuroscience, New York State Psychiatric Institute and Department of Psychiatry, Columbia University, New York, NY 10032; Department of Neonatology, Wolfson Medical Center, Sackler School of Medicine, and Department of Neurobiochemistry, Tel Aviv University, Tel Aviv 58100, Israel. [email protected]

Abstract: According to evolutionary, genetic, and endocrinal perspectives, gender differences are modulated by the interaction between intra-uterine stress, genetic equipments, and the availability of the facilitating environment during the newborn period. The social message of fitness over obstacles during socialization and the discussion of secure/non-secure attachment styles should take into consideration the brain functions, which are altered differently in response to intra- and extra-uterine stress in each gender.

Sexual selection is governed by social selection, and social regulation should also be taken into account. The “helper” hypothesis raised by Del Giudice is exactly this kind of mechanism. Del Giudice’s characterization of the stress-stricken male as being more aggressive, more dominant, and more competitive may be considered as the early selection of boys to raise the cutoff point of survival, leaving only the highly fit individuals for adult stages when same-sex competition determines the odds for reproduction. This is buttressed by a series of studies in the obstetric literature from this decade, indicating the greater vulnerability of male fetuses as compared to female fetuses (Anderson & Doyle 2008; Deulofeut et al. 2007; Ingemarsson 2003; Jones et al. 2005; Pressler & Hepworth 2002). Evolutionary logic may also explain the observed phenomenon. Ingemarson (2003) offers a view of the fetus as basically female. The masculinization process is regarded as an excess process that brings an additional set of risks to the fetuses’ development. This could be interpreted similarly to the evolutionary Handicap Principal, according to which, excessive survival risks characterizing animals convey a social message of fitness over obstacles (Cellerino & Jannini 2005). It may be argued that males have a genetic disadvantage because they are equipped with two distinct sex determination chromosomes (XY) instead of the parity backup that exists in females, who have two of the same kind of sex determination chromosomes (XX). Ingemarsson (2003) claims that in comparison with the X chromosome, the Y chromosome has degenerated through evolution and includes only a small number of genes, all of which are heterozygote. Having only one copy of each gene means that every “bad” allele is expressed. A mediating factor for the gender genetic differences may be stress (Als 1986; Als et al. 1994; 2004; Heckman et al. 2005). Phillips (2007) claims that antenatal stress has life-long effects that vary among men and women, and suggests the possibility of gender dimorphic environmental fetal programming. Davis and Emory (1995) show gender dimorphic stress reaction in healthy, full-term infants prior to extensive socialization. Therefore, I suggest that these differences found in the response to stress are constructed and programmed in utero and continue to develop differently during the socialization

process, so that the factor of stress, as suggested by Del Giudice, only secondarily affects gender differences, interacting after birth with this early programming. Meaney and Szyf (2005) examined this issue with an animal model and found that increased levels of pup licking/grooming by rat mothers in the first week of life alter the DNA structure at a glucocorticoid receptor gene promoter in the hippocampus of the offspring. These effects are far more robust in females, suggesting that girls may be better equipped genetically but may also have a greater vulnerability to non-supportive environmental signals, and that this is the manner in which the non-secure girl develops an anxious attachment style, as claimed by Del Giudice. Findings from Francis et al.’s (2002) animal studies are in accordance with the above. Francis et al. found that in the central nucleus of the amygdala and bed nucleus of the stria terminalis (which encompasses discrete, anatomically separate populations of oxytocin and vasopressin receptors; Huber et al. 2005), oxytocine receptors binding – which decreases anxiety and stress, and facilitates social encounters, maternal care, and the extinction of conditioned avoidance behavior – was increased in adult females, but not adult males, that had received high levels of maternal licking and grooming as pups. Conversely, amygdala vasopressin receptor binding – which is involved in modulating fear and aggression – was increased in males, but not females, that had received high levels of maternal licking and grooming. This further explains the manner in which the non-secure boy, in order to appear robust in the environment, develops – as suggested by Del Giudice – an avoidant attachment style. The attachment styles may be understood not as gender differences so much as differences in the gender construct of society and its interaction with the genetic fetal equipment for dealing with intra- and extra-uterine stress, leading female offspring to be better fitted to deal with social pressure by seeking bonding and male offspring to be better fitted to deal with social pressure by avoiding collision. Therefore, finally, the gender effect could be explained by socialization: It could be that mothers care for boys and girls differently, according to certain social gender roles, as well as certain cues coming from the infant which are already formatted in utero. The social context of my assumptions suggests that the interplay of sex hormones and stresses is an interfering factor in the aromatization process of masculinization, whereas Del Giudice refers to sex hormones as mitigating stress. In the face of contradicting results and theories, the concept of co-regulation may be considered (Als 1986; Goldstein Ferber 2008; Hofer 1994). That is, in situations of good co-regulation between neurobehavioral subsystems in utero and in the newborn period, a boy may benefit from the interplay between these subsystems, including the development of the sex hormones; whereas, in cases where such a co-regulation state is lacking, especially in the deregulation of the development of the hypothalamic-pituitary-adrenal (HPA) responses, the interplay may turn into an interfering process and result in difficulties as early as during gestation. Additional perspectives suggest that within a regulatory process in early parental interactions, with the offspring’s cues already shaped in utero, the dyadic reciprocal regulation between the child and his or her parents in these early interactions (Archer 1996; Cho et al. 2007) may determine (1) whether the boy or the girl develops a secure attachment style, and (2) whether the boy or the girl develops his or her gender’s typical disadvantage in attachment style (i.e., either avoidant or anxious). Therefore, I suggest that the level of fitness between the gender-type cues shaped during gestation, the ability of the parents to reciprocate with those cues, and the social interest the parent represents, work in feedback circuits. Having said that, it seems that socialization and the development of attachment styles are processes provided with windows of BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

29

Commentary/Del Giudice: Sex, attachment, and the development of reproductive strategies opportunities during critical ages in early childhood, and that anthromorphologic structural theories are able to explain contradicting research data by functional ante- and postnatal processes.

Attachment theory underestimates the child doi:10.1017/S0140525X09000119 Judith Rich Harris 54 Crawford Road, Middletown, NJ 07748. [email protected] http://xchar.home.att.net/tna/

Abstract: The problem with elaborations of attachment theory is attachment theory itself. How would a mind that works the way the theory posits have increased its owner’s fitness in hunter-gatherer times? The child’s mind is more capacious and discerning than attachment theorists give it credit for. Early-appearing, long-lasting personality characteristics, often mistaken for the lingering effects of early experiences, are more likely due to genetic influences on personality.

The human mind was shaped by natural selection. The mind we inherited from our ancestors was tested against alternative versions and won out against the competition because it was more successful in dealing with the challenges of hunter-gatherer life. Would the mind depicted by attachment theorists have passed this test? Would it have helped its owner survive and reproduce in the Paleolithic? According to attachment theory, an infant develops expectations about the world and its people on the basis of the way his mother takes care of him. How would this be beneficial to him? Why should a child with an attentive mother expect other people – siblings, peers, strangers – to be equally attentive? And how would it profit the child of a negligent mother to go into the world with the expectation that no one will love him? Isn’t the child’s mind capacious enough to form separate “working models” for all the important people in his life, so that he can behave appropriately with each of them? According to the “life history” elaboration of attachment theory, a girl who experiences a stressful environment in the first few years of life, or whose father was absent or insufficiently attentive during that period, is more likely to reach sexual maturity early and to hasten into sexual relationships with temporary partners. But didn’t every child grow up in a stressful environment in the Paleolithic? Many parents died before their children were grown. Polygyny was probably universal, limited only by men’s ability to support additional wives. As noted in the target article, paternal childcare is rare in polygynous societies. According to life history theorists, the child’s experiences with parents in infancy and early childhood influence her reproductive strategy 10 or 12 years later. But why should a girl’s reproductive strategy be based on such inadequate data? Many things might change in the interim, so why not wait as long as possible to settle on a strategy? And why look only at her own parents? Birds are capable of making a decision about next year’s nesting site on the basis of “public information” they collect by observing other nests in the area (Doligez et al. 2002). If a bird can do it, why not a child? Her family may be atypical, so the more data she collects the better. Attachment theory underrates the child. There is good evidence that children construct separate working models for each of the important people in their lives, and that expectations developed in one relationship are not carried over to other relationships. Babies can be securely attached to one caregiver and insecurely attached to another. Infants of depressed mothers are subdued with their mothers but

30

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

behave normally with other familiar caregivers. Children who are pushed around by older siblings at home are fully capable of taking a dominant role with peers (see Harris 1998; 2006). Although behavior is highly sensitive to context, it is not ruled solely by context. Researchers observed children interacting with their parents and the same children interacting with their peers, and recorded instances of uncooperative, angry, or bossy behavior. The correlation between behavior with parents and behavior with peers was .19 (Dishion et al. 1994). I attribute this correlation not to generalization or transfer but to heritable personality traits (e.g., disagreeableness) that the child takes along to every social context. Confirmation comes from behavioral genetic studies of other behaviors, such as shyness. Correlations between shy (or bold) behavior in two different contexts are almost entirely due to genetic influences on these behaviors. In contrast, environmental influences on shyness are context-specific (Saudino 1997). The same holds true for consistency over time. The long-term stability of personality traits is primarily due to the heritable component of such traits (Caspi & Roberts 2001). This finding has an important implication: Personality characteristics that appear early in development and persist into adulthood are largely genetic and not, as most developmental psychologists have assumed, due to the lingering effects of early experiences. Though the role of genes in producing resemblances between siblings is now generally acknowledged, the fact that children also share half their genes with their biological parents – both parents – is widely ignored. For genetic reasons alone, people who are impulsive, quickly bored, easily angered, or prone to depression have a greater-than-average risk of producing offspring with similar disadvantages. A pretest for temperament in infancy is a hopelessly inadequate way to control for genetic influences on behaviors measured in infant, child, and adult attachment tests (Rowe 2000b). Attachment theory is based almost exclusively on studies that examine one child per household; almost always, the child is the biological offspring of the mother. This method makes it impossible to disentangle environmental and genetic influences. Studies of twins, siblings, and adoptees – two per household – have shown that about half the variance in most psychological characteristics can be attributed to genetic influences, and most of the remainder to unknown “unique” influences (see Harris 2006). The environment shared by children who grow up in the same household accounts for little or none of the variance. Physical maturation – age at menarche – follows the same pattern; although heritability is higher, most studies find little effect of shared environment (see Mendle et al. 2006). Whether or not a father is present, and whether or not the home is stressful, are aspects of the shared environment. As Nettle (2006) has argued, most individual differences in mating behavior can be attributed to heritable personality characteristics. Differences in rate of maturation are also heritable, and the environmental factors that play a role may not be the ones discussed in the target article. For instance, the role of socioeconomic status (SES) is often overlooked. Girls in troubled households may mature more rapidly because, in developed nations, obesity is negatively correlated with SES: children in lower-SES homes and neighborhoods tend to be fatter (Shrewsbury & Wardle 2008). Because age at menarche depends on body weight, overweight girls mature faster (Frisch 1988). The link between body weight and rate of maturation can also explain why girls today are maturing faster than they used to (Cesario & Hughes 2007). The increased prevalence of obesity is probably to blame. In conclusion, correlations between infant attachment behaviors and later sexual relationships are most likely due to persisting, heritable personality traits. Evolutionary psychologists and biologists can explain sex differences in sexual relationships without recourse to attachment theory.

Commentary/Del Giudice: Sex, attachment, and the development of reproductive strategies

Pre-adjustment of adult attachment style to extrinsic risk levels via early attachment style is neither specific, nor reliable, nor effective, and is thus not an adaptation doi:10.1017/S0140525X09000120 Johannes Ho¨nekopp School of Psychology and Sport Sciences, Northumbria University, Newcastle upon Tyne NE1 8ST, United Kingdom. [email protected]

Abstract: The mechanism proposed by Del Giudice by which adult attachment style is adapted to the extrinsic risk in the local environment via attachment style during the early years does not fulfill important criteria of an adaptation. The proposed mechanism is neither specific, nor developmentally reliable, nor effective. Therefore, it should not be considered an adaptation.

In the wake of similar models (Belsky et al. 1991; Chisholm 1993; 1996; 1999), Del Giudice proposes an adaptation consisting of multiple steps that adapt adults’ attachment style and reproductive strategy to the level of extrinsic risk in their local environment via their attachment style during early years. Unfortunately, Del Giudice never discusses which criteria we should use to establish if a certain mechanism or structure is an adaptation. Andrews et al.’s (2002) summary of the debate on this issue shows that specificity, developmental reliability, and efficiency figure prominently among these criteria. The mechanism envisaged by Del Giudice fails on all three of them. Although parental behavior has a moderate effect on children’s attachment style (De Wolff & van IJzendoorn 1997), the latter appears to be insensitive to the extrinsic risk of the environment. In a cross-cultural review of attachment patterns, van IJzendoorn and Sagi (1999) find one of the highest proportions of secure attachment (88% with a tripartite coding system) in the Dogon, for which child mortality is extraordinarily high (25% of children die within their first five years). Thus, deviation from secure attachment does not appear to be very sensitive to extrinsic risk. In addition, unusually high deviations from secure attachment were found in kibbutz children (31% to 52% not securely attached). But the reason for this high level of insecure attachment is not an extrinsically risky environment but, rather, that infants sleep away from their parents, with Israeli day-care infants showing secure attachment rates between 75% and 80%. Thus, the effect of extrinsic risk on infant attachment style lacks specificity because risk hardly (if at all) affects attachment style, whereas an irrelevant variable (sleeping away from parents) has a strong effect on attachment style. However, even if the proportion of non-securely attached children clearly covaried with the extrinsic risk of the environment, the overall low variability in the relative frequency of attachment patterns should be noted. Van IJzendoorn and Sagi’s (1999) review strongly suggests that secure attachment is the norm across very different environments. If deviation from secure attachment were adaptive in the face of high extrinsic risk, we would thus find that the majority of children in cultures that suffer from high childhood mortality or other high risks are maladapted because they are securely attached. However, if deviation from secure attachment were adaptive under these circumstances, we should observe developmental reliability; that is, the shift away from secure attachment should be ubiquitous and not only seen in a fraction of children. Women’s esthetic judgment of male bodies may serve as an example for the ubiquity of an alleged psychological adaptation in humans. Ho¨nekopp et al. (2007) hypothesized that women evolved an adaptive preference for the bodies of athletic men. In line with the notion that an adaptation should be developmentally reliable, a preference for the bodies of athletic men was found for all women in their sample. Even if the mechanism envisaged by Del Giudice were specific and developmentally reliable, it would still be unlikely to be

effective. The reason is that the proposed chain of information transmission is simply too long: In the ideal case, a match between A’s reproductive strategy and the extrinsic risk of the environment would come about because (1) the risk during A’s adulthood correlates with the risk during A’s early years, (2) the risk during A’s early years correlates with A’s parents’ caregiving behavior, (3) A’s parents’ care-giving behavior correlates with A’s early attachment style, and (4) A’s early attachment style correlates with A’s adult attachment style and reproductive strategy. Even if we assume an unusually high correlation of r ¼ .7 within each of the four links, half of the relevant information would be lost during each transmission stage (because two variables that correlate with r ¼ .7 have 0.72 ¼ 49% variance in common). This information loss multiplies over the whole chain, so that 94% of the relevant information is lost at the end of the chain (1 2 12  12  12  12). Of course, even a mechanism that loses 94% of the relevant information may be adaptive if there is no better alternative. But obviously, A has ample opportunity to directly observe the extrinsic risk in the local environment. As humans’ sexual strategies appear flexible enough to take such information into account (Gangestad & Simpson 2000), A’s mating strategy should be based on the observation of the current environment. And the availability of this more direct strategy renders the mechanism proposed by Del Giudice utterly ineffective. In sum, the proposed mechanism of adjusting adult romantic attachment style and sexual strategy to the extrinsic risk level of the local environment via attachment style during the early years lacks specificity, developmental reliability, and efficiency, and should therefore not be considered an adaptation.

Synthesizing life history theory with sexual selection: Toward a comprehensive model of alternative reproductive strategies doi:10.1017/S0140525X09000132 Jene´e James Jackson and Bruce J. Ellis Division of Family Studies and Human Development, University of Arizona, Tucson, AZ 85721-0078. [email protected] [email protected]

Abstract: Del Giudice’s model of sex-specific attachment patterns demonstrates the usefulness of infusing life history theory with principles of sexual selection. We believe a full synthesis between the two theories provides a foundation for a comprehensive model of alternative reproductive strategies. We extend Del Giudice’s ideas based on our own program of research, focusing specifically on the importance of intrasexual competition and the individual phenotype during development.

Del Giudice’s explication of sex-specific patterns of insecure attachment substantively advances our understanding of attachment organization across development and its role in shaping adult reproductive strategies. An important strength of the model is the incorporation of parental investment and sexual selection theory into current life history models, enabling a better account of sex-differentiated life histories. The focus on sex-specific reorganization of attachment patterns in middle childhood as a critical phase in the formation of reproductive strategies is a novel and exciting idea that should stimulate future research. Despite these strengths, the synthesis of life history theory and sexual selection theory needs further development. A comprehensive model must incorporate not only the concept of asymmetries in parental investment between the sexes, but also the alternative reproductive strategies that arise within each sex as a result of intrasexual competition. Current sexual selection models, such as Gangestad and Simpson’s Strategic Pluralism BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

31

Commentary/Del Giudice: Sex, attachment, and the development of reproductive strategies Theory (2000), emphasize social and sexual competition as important factors shaping adaptive variation in reproductive strategies. The incorporation of ideas from sexual selection theory into life history models should be especially valuable because the weakness of one theory is often the strength of the other. For instance, life history theory has provided the basis for a useful set of models for explaining development of variation in women’s reproductive strategies (e.g., explaining developmental trade-offs between current vs. future reproduction; see Belsky et al. 1991; Ellis 2004). However, because life history models have not adequately addressed social-sexual competition for mates, they have been limited in their ability to explain strategic variation among men. Sexual selection models, by contrast, have been successful in explaining how males apportion reproductive effort to mating versus parenting in relation to current levels of social-sexual competitiveness (e.g., Gangestad & Simpson 2000; Gross 1996), but do not provide an adequate developmental model for how and when males should make these adaptive shifts. To address these limitations, we have constructed a more comprehensive theoretical framework based on the strengths of each model (Jackson & Ellis, submitted; Jackson et al., submitted). Our approach toward this synthesis has been to integrate the components of social and sexual competition into the developmental trajectories laid out by life history theorists. Our theory and research in this area focus on how early familial environments and status obtained in adolescence and early adulthood contribute, additively and in interaction, to the development of reproductive strategies in men and women. Because males and females must solve qualitatively different adaptive problems when negotiating life history trade-offs, there should be important sex differences in the tracking of environmental information, or at least the weight given to specific environmental cues (see Gangestad & Simpson 2000; Thiessen 1994). Females are ultimately constrained by the resources that they can extract from the environment, their relatives, and their mates in order to successfully produce and rear offspring; accordingly, they should be especially attuned to the nature of the local ecology and support in and around their home environments, per life history models. Males, on the other hand, are ultimately constrained by their ability to access, attract, and retain females; accordingly, their own reproductive strategies should be especially attuned to the demands and desires of females and their ability to successfully engage in intrasexual competition, per sexual selection models. Recent empirical studies lend support to these claims. In a longitudinal study on adolescent development, Ellis and Garber (2000) found that early psychosocial stress (i.e., discordant family relationships, father absence/stepfather presence) predicted early pubertal maturation among girls. In the same sample, Jackson et al. (submitted) examined the effects of early psychosocial stress within the home and competency within the peer group on attachment orientation and sexual behavior in late adolescence. As in the previous analyses of pubertal timing, in females, but not in males, early psychosocial stress was a reliable predictor of heightened sexual activity, risky forms of sexual behavior, and avoidant romantic attachment patterns. By contrast, in males, but not females, self-perceived social competence and athletic ability during adolescence were reliable predictors of heightened sexual activity and secure romantic attachment in late adolescence. Thus, consistent with our model, development of female reproductive strategies was more closely linked to variation in familial and ecological conditions, whereas development of male reproductive strategies was more closely linked to social-competitive ability and status. All children do not respond equally, however, to family environments and social-sexual competition. An important factor noted in

32

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

section 6.4, but not explicitly modeled by Del Giudice, is that the effects of attachment on social and reproductive strategies may depend on the phenotype of the developing child. Phenotypic characteristics likely moderate the developmental pathways presented by Del Giudice in the following ways. Among females, phenotypic quality could alter the extent to which they are able to extract resources from relatives and mates, moderating in part the relationship between extrinsic risk, attachment orientation, and sexual behavior in adulthood. Among males, phenotypic quality could alter the extent to which they are able to enact specific competitive and reproductive strategies. For instance, the male avoidant strategy detailed by Del Giudice includes a personality profile of inflated self-esteem, self-reliance, and aggression that facilitates status seeking. Although we agree that these traits may aid in intrasexual competition in certain social niches, it is likely that only those males who possess the physical prowess needed to successfully compete in such a manner would benefit from the strategy. Along these lines, Figueredo and Jacobs (2000) have argued for a kind of reactive heritability model whereby the inheritance of strategically relevant traits biases individuals toward different strategies. They contend that, “psychosexual development involves a self-assessment of sociosexual capabilities and opportunities, calibrating optimal utilization of physical assets such as size, strength, health, and attractiveness, as well as psychosocial assets such as intelligence, self-efficacy, social skills, personality, and socioeconomic status and/or prospects” (p. 603). In sum, extant phenotypic characteristics should interact with social status and developmental experience to determine life history strategies. In conclusion, Del Giudice’s ideas concerning the sex-specific organization of attachment beginning in middle childhood are likely to assume an important role in research on development of life history strategies. We see Del Giudice’s model, together with our own, as affording meaningful steps toward the integration of life history and sexual selection models of human reproductive strategies.

Attachment patterns of homeless youth: Choices of stress and confusion doi:10.1017/S0140525X09000144 Min Ju Kanga and Michael Glassmanb a Department of Child and Family Studies, Yonsei University, Seoul 120-749, Korea; bDepartment of Human Development and Family Science, Ohio State University, Columbus, OH 43210. [email protected] [email protected]

Abstract: This commentary explores the reproductive strategies and attachment patterns among homeless youths. Del Giudice’s integrated evolutionary model is applied to a homeless youth population that must function in ecological settings of constant high risk and stress. Different reproductive needs result in different patterns of high-risk behaviors. Intervention considering the sex differences, life history, and early caregiver– child relationships is suggested.

It is difficult to imagine situations with higher levels of stress and greater relative extrinsic dangers than those faced by homeless youth on a given day. There are between 500,000 and two million homeless youth in the United States (Cooper 2006), with few social service or non-governmental agencies to help them (Wright 1990). Homeless youth have a number of hygiene and health vulnerabilities that make them outcasts in society (Staller 2004); but they also tend to avoid and be mistrustful of what society might have to offer. Del Giudice’s target article informs the issue of homeless youths in two ways: (1) it offers insight into what we believe to be the extreme avoidant

Commentary/Del Giudice: Sex, attachment, and the development of reproductive strategies behavior of many homeless youths, especially males, and (2) it provides a possible context for many of their behavioral choices, especially those pertaining to sexuality and reproduction. Del Giudice’s thesis examining the interconnection between attachment patterns, life histories, and reproductive strategies allows us to view the extreme avoidant behavior, and seemingly illogical reproductive choices (e.g., young girls having children with transient partners while they are homeless), through the lens of an adaptive model rather than a deficit model; except, it is adaptation turned on its head by brutal and uncaring social ecologies. Many homeless youth begin to engage in sexual activity at an early age – sometimes as the result of relationships and sometimes as a means to an end (survival sex). Del Giudice’s reference to Sroufe et al. (1993) concerning the intersection between life stress, insecure attachment patterns, and early violation of gender boundaries in middle childhood has particular resonance. A number of youths seem to become highly sexualized in their behavior relatively early in life, with flirtation becoming a dominant form of communication and connectivity. For males, flirtation and early sexual behavior are part of the initiation into the same-sex social hierarchy that will eventually determine their place in the street economy. The need to develop an avoidant stance towards relationships is critical. The females may be more confused about what their flirtation means. A number of homeless females have been sexually abused, been raped at some point, or used sex for survival purposes. At the same time, flirtation and sex through middle childhood and into adolescence is one of their surest relational strategies. When homeless youth do engage in sex, many times it is unprotected, leading to both pregnancy and the spread of sexually transmitted diseases (STDs). Although this may seem like foolish or self-destructive activity to a casual observer, Del Giudice’s thesis suggests that these youths are following natural reproductive strategies that meet needs determined by their circumstances. Many of the males will meet violent ends, become alienated homeless adults, or wind up in jail. There is a drive to impregnate females while remaining distant. The females are often left alone with children while they are still homeless or living on the margins of society, leading them to exhibit depressive symptoms (Meadows-Oliver 2007). Research has suggested that depressed mothers who have low emotional availability (e.g., are less sensitive, less structuring, and more intrusive and hostile) during the early years of child rearing, tend to have children with low emotional availability (e.g., less responsive, less involving) (Easterbrooks et al. 2000). And this quality of emotional interaction can lead the insecure attachment patterns (Ziv et al. 2000), perpetuating a cycle of avoidance and alienation. Stress becomes a regular part of the developing child’s life, and the impact of life histories on choices becomes more evident and more dominant, with somatic resources used primarily for survival. Because the street economy often plays a major role in the lives of homeless youth, many males move in the direction of highly avoidant insecure attachment patterns. To establish long-lasting relationships is in many ways to become more vulnerable. They understand that their transient relationships – the members of their “posse” – are critical for their survival. On the other hand, many of the females who have been raised by single mothers understand that they represent the sole opportunity for their children to survive. They necessarily have to choose more ambivalent insecure attachment patterns to protect their child. However, once stress and/or extrinsic danger reaches such a high level that their attempts at caretaking strategies end in failure or prove to be impossible (for ambivalent homeless females’ behavior choices don’t so much involve “helping at the nest,” as attempts to create a nest), their relationship strategies can change. We have seen a number of occasions where failure in attempts to develop a middle-class-type lifestyle (i.e., with

job, stable living conditions) has caused once hopeful females to be even more avoidant and alienated than the males. It is our hope that some of the ideas from Del Giudice’s article might give us a better handle on the development of interventions for these homeless youth. To date, interventions have had only limited success, with one of the major impediments being lack of trust and extreme avoidant behavior. By focusing on reproductive strategies, we can hypothesize that one of the best times to engage female homeless youths is when they are pregnant. It is possible that this is the point at which they are most ready for and in need of establishing relationships, even if these relationships are based on ambivalent strategies. Unfortunately, the same phenomena that cause these females to want to establish linkages and relationships, make them highly vulnerable to failure or avoidance by society at large. If we can develop programs that are nonjudgmental and focused on success, especially involving “nesting” trajectories (e.g., getting stable housing that is not contingent and the resources necessary to raise a healthy child), there is the possibility of reinforcing lifetime linkages. And perhaps more importantly, it would allow these mothers to develop stronger, more emotionally available relationships with their children, breaking the cycle of depressive and self-destructive behavioral trajectories. Thus, an intervention that first helps the homeless females to successfully find the stability in housing, jobs, and social support, will reduce the chances of females being so passive in relation to males’ reproductive strategies, leading to fewer opportunities for multi-mating (a major cause of paternal absence in parenting).

Developmental transformations in attachment in middle childhood doi:10.1017/S0140525X09000156 Kathryn A. Kerns Department of Psychology, Kent State University, Kent, OH 44242. [email protected]

Abstract: The target article proposes a model to explain the emergence of sex differences in attachment in middle childhood and their implications for reproductive strategies. While biological factors are prominent in the model, little is said about the social context of middle childhood and its contributions. There is also a need to clarify the fundamental nature of attachment in middle childhood.

The Del Giudice target article raises important questions about how gender may play a role in the development of attachment in middle childhood. The author rightly notes that the influence of gender has been largely ignored in the attachment literature. The proposed model generates several novel hypotheses and is likely to lead to new research. Especially intriguing is the idea that insecure attachment may be reorganized differently for boys and girls in middle childhood in support of reproductive strategies that have evolutionary advantage. Given that attachment theory is, in large part, a theory of social influence, it is surprising that the model does not specify the impact and role of social partners during middle childhood. The model depicts an “early experience” role for parenting, in which a parent’s main role is to influence the initial development of attachment. Parents are, however, still the primary attachment figures for children in middle childhood (Kerns et al. 2006), which raises the question of what role they play in the proposed developmental transformation of attachment. For example, if a girl switches from an avoidant to a heightening strategy with a parent (as predicted by the model), presumably this would invoke some changes from the parent (e.g., increased rejection). The dyadic nature of parent –child relationships is not captured by the model. Relatedly, although peers play an important role BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

33

Commentary/Del Giudice: Sex, attachment, and the development of reproductive strategies in socializing gendered behaviors, very little is said regarding how peers may influence and amplify any emerging sex differentiation in attachment. Thus, while biological influences on attachment deserve greater consideration, the proposed model would be enhanced by also incorporating social influences in middle childhood. A complexity in interpreting the model arises from the multiple meanings of the word “attachment.” Does it refer to a child’s relationship with a specific target, or to a child’s general orientation (style) across different attachment relationships? Both meanings are used in the target article. There has been a lack of research on how experiences in multiple attachment relationships come to be integrated into a general orientation (e.g., “state of mind”) to attachment. It is possible that this integration begins in middle childhood (Kerns et al. 2005), which may have some implications for the model. Specifically, the proposed sex-specific reorganization in insecure attachments could occur within the context of specific attachments (which would produce unstable relationships) or could be reflected in a child’s (emerging) attachment style. It is possible that specific attachments may be influenced primarily by patterns of interaction with a partner, whereas attachment “style” is influenced by some combination of experiences in specific relationships, gender socialization, genetic propensities in personality traits, and evolutionary pressures for adaptive mating and reproduction strategies. If the author is correct in suggesting that the proposed sex differences in attachment are in the service of reproductive strategies, then it is also possible that sex differences in insecure attachment will be most pronounced in attachments to peers that develop in late adolescence (i.e., in relationships where mating and reproduction are more relevant). A key prediction from the model is that, in middle childhood, most insecure boys will be avoidant, and most insecure girls will be ambivalent in their attachments. The literature review on gender differences in attachment, which was based on studies using doll play interviews or questionnaires, did provide evidence for sex differences in the distributions of insecure attachment in 6to 12-year-old children. The review could have been more extensive. There are additional studies of this age range that have employed observational measures (e.g., Graham & Easterbrooks 2000; Moss et al. 2004) or autobiographical interviews (e.g., Ammaniti et al. 2000; Target et al. 2003) to assess attachment, but these studies were not included in the review. Evidence that sex differences are found using these other methods (as well as evidence for sex differences in Adult Attachment Interview [AAI] insecure classifications in adolescence) would provide stronger evidence that the pattern is not confounded with choice of methods. (Del Giudice’s point about inattention to gender in the attachment literature is well taken, as most of the studies cited in this paragraph did not provide information regarding the sex breakdown within the insecure attachment groups.) Interestingly, the review of studies also showed that many insecurely attached boys are disorganized. The author provides a cogent discussion of why insecure boys might be predisposed to adopt avoidant strategies; but why are so many boys in middle childhood disorganized in their attachments, and can this be accounted for by the model? Perhaps the model could be elaborated to identify factors in the social ecology (e.g., level of family stress) that might distinguish between boys who develop disorganized rather than avoidant attachments, but it is less clear how disorganized attachment might confer an adaptive advantage. Despite these limitations, the article does generate many testable hypotheses, some of which are not intuitive and are not consistent with current theory. If there is a reorganization of insecure attachment in middle childhood that is not a result of changes in parenting, and an overall waning of the influence of parents, then presumably both cross-generational continuity in attachment (parent – child correspondence) and stability of child – parent attachment would be lower in middle childhood than in early childhood (although how much lower is not quite clear from

34

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

the article). These corollary hypotheses are currently difficult to evaluate given the lack of relevant data (although see Target et al. [2003] for an exception). Longitudinal studies in early middle childhood are especially needed to test the hypothesis that girls and boys shift toward different insecure attachment patterns around age 7 years. Given the modest links between attachment and parenting in middle childhood (Kerns, in press), more elaborated models of the influences on attachment are needed. The model proposed by Del Giudice provides additional processes (specifically, biological mechanisms) that could be tested, as well as specifying how levels of stress may predispose girls and boys to different developmental pathways.

Life history as an integrative theoretical framework advancing the understanding of the attachment system doi:10.1017/S0140525X09000168 Daniel J. Kruger Health Behaviour & Health Education, University of Michigan, Ann Arbor, MI 48109-2029. [email protected] http://www-personal.umich.edu/~kruger

Abstract: Evolutionary Life History Theory (LHT) is a powerful framework that can be used for understanding behavioral strategies as contingent adaptations to environmental conditions. Del Giudice uses LHT as a foundation for describing the attachment process as an evolved psychological system which evaluates life conditions and chooses reproductive strategies appropriate in the developmental environment, integrating findings across several literatures.

Evolutionary Life History Theory (LHT) emerged in evolutionary biology (e.g., Pianka 1970) to describe the relationship between environmental conditions and reproductive patterns across species. The vast majority of research addressing LHT has been conducted in nonhuman species, and usually makes comparisons between species. In recent years, there has been a growing interest in human life history and life history variation within species (e.g., Heath & Hadley 1998). LHT holds great promise for promoting the understanding of our own species and integrating findings from diverse research methodologies spanning multiple levels of analyses. Following the behavioral ecology principle that behavioral strategies related to reproductive success are conditional based on the characteristics of the physical, economic, and social environment (Crawford & Anderson 1989), Belsky et al. (1991) proposed that the attachment process is an evolved psychological system that evaluates life conditions and chooses reproductive strategies appropriate to the developmental environment. Del Giudice provides an overview of research related to this theory across a variety of perspectives and enhances its precision of prediction by adding insights from sex differences in human reproductive strategies. The target article, “Sex, attachment, and the development of reproductive strategies,” demonstrates the value of LHT as a framework for integrating converging evidence across fields and levels of analyses to result in a more complete and comprehensive understanding of the complex factors underlying human behavioral patterns. Evolution by natural and sexual selection is the most powerful theory in the life sciences, and in recent decades there has been considerable progress in using evolutionary theory to explain behavior, especially human behavior. E. O. Wilson’s (1975) book, Sociobiology: The New Synthesis, generated considerable political controversy in the heyday of social relativism, where theories describing the biological bases for behavior were interpreted as a threat to the goal of social equality. More recently,

Commentary/Del Giudice: Sex, attachment, and the development of reproductive strategies “evolutionary psychology” has emerged as the dominant moniker for the study of evolution and human behavior. The added value that evolutionary psychology brings to the study of evolution and human behavior is the identification of proximal mental processes or mechanisms, which guide behavioral strategies (see Cosmides & Tooby 1994). Darwinian anthropologists and others rightly argue that “evolutionary psychology” may be too narrow as a proper descriptive term. As can be seen in Del Giudice’s article, the evolutionary framework crosses traditional disciplinary boundaries and promotes a universally intelligible account that may enhance communication between researchers using quite divergent methodologies. Tinbergen’s (1963) four questions may be used to assess the maturation of explanations for behavior. These questions entail the evolutionary (ultimate) explanations of function (adaptation) and phylogeny (evolution), as well as the proximate explanations of causal mechanisms and ontological development. Del Giudice’s overview addresses each of these areas. Del Giudice rightly identifies Darwin’s theory of evolution by selection as the only viable account for biological design. He describes how the psychological attachment system is an adaptation to promote reproductive success through the selection of sexspecific reproductive strategies based on the availability of caregivers during development. Biological (including psychological) processes are designed by selection to promote inclusive fitness, and Del Giudice’s overview explains why seemingly dysfunctional behavioral patterns are actually useful in promoting an individual’s contribution to future generations. Del Giudice details cross-cultural studies which strongly suggest that the attachment system is a human universal, rather than a product of cultures with particular familiar patterns. In fact, variations in attachment patterns follow the expected direction based on local mating systems and family structures. Del Giudice emphasizes the unique combination of facultative paternal care and cooperative breeding in humans as the key distinction of our species. He notes the existence of adrenal puberty in gorillas and chimpanzees and its absence in other primates and most other mammalian species, as well as the experimental evidence from the manipulation of hormonal levels across a variety of species. However, the phylogenetic context of the attachment system is perhaps the weakest aspect of his overview. This may be easily remedied in part by inclusion of research such as Harry Harlow’s maternal-deprivation and social isolation experiments on rhesus monkeys (e.g., Harlow 1964), which were inspired by Bowlby’s work on human child institutionalization. A fuller phylogenetic understanding may require further research on species that vary in phylogenetic proximity to humans and on the extent of parental care. Del Giudice’s overview is particularly adept at summarizing proximate explanations. He describes the neuroendocrine bases for the attachment system. Our current understanding of the interplay between psychology and underlying neuroendocrine mechanisms is incomplete, though research is flourishing and likely far from the point of diminishing returns. As models of neuroendocrine systems increase in precision, they will enhance the ability of attachment theorists to specify the causal cascade contributing to patterns of reproductive behaviors. The theory’s depiction of causal mechanisms is properly allied with an account of the developmental processes guiding behavioral strategies. The quality of parental care serves as a proxy for environmental conditions and suggests locally effective strategies. Early caregiving experiences influence attachment styles, which in turn influence reproductive strategies later in life. The LHT emphasis on the importance of the interaction between genes and environment will hopefully further alleviate misconceptions that evolutionary explanations of behavior entail genetic determinism, providing an additional reason to move beyond the unfruitful nature versus nurture debates. The recognition and understanding of adaptive developmental plasticity in Del Giudice’s model also provides encouragement and, more importantly, guidance for intervention.

In sum, Del Giudice’s account enhances our understanding of the attachment system by integrating findings across several literatures in an evolutionary life history framework. He describes a constellation of phenomena that would be difficult to explain with a competing paradigm or to discount as a “just so” story. His addition of sexual dimorphism in attachment processes proposes further refinement in the description, prediction, and understanding of human psychology and behavior.

Disorganized attachment and reproductive strategies doi:10.1017/S0140525X0900017X Andrew J. Lewis and Gregory Tooley School of Psychology, Faculty of Health, Medicine, Nursing, and Behavioural Sciences, Deakin University, Burwood 3125, Victoria, Australia. [email protected] http://www.deakin.edu.au/hmnbs/psychology/ [email protected] http://www.deakin.edu.au/hmnbs/psychology/

Abstract: Del Giudice provides an extension of the life history theory of attachment that incorporates emerging data suggestive of sex differences in avoidant male and preoccupied female attachment patterns emerging in middle childhood. This commentary considers the place of disorganized attachment within this theory and why male children may be more prone to disorganized attachment by drawing on Trivers’s parental investment theory.

Del Giudice is to be congratulated for such a bold expansion of attachment theory. The target article resumes the dialogue between evolutionary and developmental theories that proved to be such fertile ground for Bowlby himself. Sex differences in child attachment patterns and their possible preparation for adult mating strategies have been largely neglected within attachment theory, possibly in an attempt to distinguish it from its psychoanalytic origins. Accordingly, a new set of evolutionary hypotheses has emerged that will provide a welcome direction for theoretical integration and future research. In this commentary, we focus on the integration of disorganized attachment into the life history model, arguing that it may represent a limit to the adaptive function of attachment, and we also aim to briefly draw out some of the clinical consequences of this view. Del Giudice has focused his theory around the organized patterns of attachment while, by his own admission, excluding disorganized attachment. Because the life history theory of attachment is particularly concerned with the reproductive implications of early environmental adversity, disorganization could be considered, especially given that its prevalence is around 15% in low-risk middle-class families (Van IJzendoorn et al. 1999). Furthermore, the prevalence rates tend to increase in proportion to the degree of developmental adversity – parental depression, adolescent parenthood, unresolved loss or trauma and marital discord – reaching a high of up to 80% among maltreating and drug-abusing parents (Green & Goldwyn 2002; Lyons-Ruth & Jacobvitz 1999). Although in some cases a secondary organized attachment pattern can be discerned beyond the disorganized phenomena, at the extreme, cases of multiple attachment strategy in early childhood and unclassifiable adult attachment states of mind (coded as “Cannot Classify” [CC] on the Adult Attachment Interview), suggest a pervasive disorganization of the attachment system that has been found to be strongly associated with child and adult psychopathology (Bakermans-Kranenburg et al. 2005; Green & Goldwyn 2002). In introducing the formal classification of disoriented/disorganized attachment using the Strange Situation Procedure, Main BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

35

Commentary/Del Giudice: Sex, attachment, and the development of reproductive strategies and Solomon (1990) described its ethological and evolutionary framework as one of “fright without solution” (Hesse & Main 2000). The behavioural indices of disorganized attachment closely resemble phylogenetically conserved mammalian fear responses: flight, attack, and freezing behavior (Main & Solomon 1990). However, in this case, such responses are displayed toward the caregiver who ought to be a haven of safety. Thus, a degree of behavioural conflict occurs between a security-seeking attachment system and a fear-responding survival system. This is reflected in other common indices of disorganization such as simultaneous display of distress and avoidance, undirected or misdirected movements, and disrupted movements or gestures – all of which suggest the enactment of a conflict between approach and avoidance. Such an explanation draws heavily on Bowlby’s original assumption that the adaptive function of attachment is largely to do with seeking protection and, therefore, attachment behaviors in infants function to regulate fear via proximity seeking. Del Giudice cites several studies in which sex differences in the frequency and degree of severity of disorganization have been noted, with male infants tending to be the more frequently and severely disorganized. Males are more prone to aggressive pathology in middle childhood, as has been noted frequently in the conduct disorder literature. An organized response to parental insensitivity in the form of avoidant attachment may well produce an adaptive low-investment parenting strategy for males. However, the more severe modes of aggression associated with disorganization may well produce an antisocial behavioral pattern in males which, like many forms of severe psychopathology, would be maladaptive and reduce reproductive fitness. Given disorganization’s strong association with extremes of developmental adversity and dysregulation of stress responses in infancy and early childhood, why might it be more prevalent in males? Either males are more vulnerable to disorganization, or the primary caregivers of males – typically mothers – are more disorganizing in their caregiving toward male infants under some conditions. In addition to the investigation of proximate mechanisms, there may be an application of the TriversWillard hypothesis: namely, for polygynous species, parents in poor conditions are likely to invest more in females, who are more likely to bear them at least some grandchildren, whereas males raised in poor conditions will be unlikely to compete with other males and therefore would attract minimal investment (Trivers & Willard 1973). The evolutionary hypothesis that such fearful and frightening caregiving may be greater toward male offspring would be interesting to investigate. Evolutionary concepts are increasingly considered in definitions of psychopathology as an impairment of a biologically meaningful function (Wakefield 2005). The pathogenic effect of disorganization may be outside of the average expectable caregiving experiences for which humans are selected. In contrast, avoidant and preoccupied attachments are organized and strategic responses to parental – and most particularly maternal – caregiving, including sensitivity and attunement (see de Wolff & van IJzendoorn 1997). This distinction is an important one because the life history hypothesis serves to reinstate organized but insecure attachments as potentially adaptive in both the social and evolutionary senses, and in the latter, arguably serving to increase reproductive fitness in harsh rearing environments. This suggests that the attachment behavior system is both more robust and flexible than the normative assumptions of ideal security imply. However, beyond a certain threshold, highly adverse caregiving environments involving direct or implied dangers, consistent with maltreatment or caregiver absence (psychological or physical), lead to pathological outcomes and, although it remains to be established, reduced fitness. Implications for future directions include extensions of the research agenda to psychobiological dysregulation arising from disorganization. Implications of disorganization for pubertal timing and reproductive strategy, mate choice, and degree of

36

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

parental investment could be investigated. Finally, clarification of the classification of disordered attachment and enhanced clinical interventions can be derived from a better understanding of the biological function and dysfunction of the attachment system across the human life cycle.

Gender difference of insecure attachment: Universal or culture-specific? doi:10.1017/S0140525X09000181 Nanxin Li,a Jibo He,b and Tonggui Lic a

Department of Psychology, Yale University, New Haven, CT 06520; Department of Psychology, University of Illinois, Champaign, IL 61820; Department of Psychology, Peking University, Beijing, 100871, People’s Republic of China. [email protected] http://pantheon.yale.edu/~nl238/ [email protected] http://www.psych.uiuc.edu/people/showprofile.php?id ¼ 798 [email protected] http://www.psy.pku.edu.cn/en/litonggui.html b c

Abstract: Our research in China does not show gender differences in insecure attachment patterns. We believe that cultural differences between Chinese and Western societies may help to explain this phenomenon. Mating and parenting circumstances in China do not allow males to adopt a zero-investment strategy. In addition, attachment styles are transmitted across generations and last for the whole lifespan. Here, we argue that the influence of mating and parenting on the welldeveloped attachment patterns in childhood is relatively small.

In section 6 of the target article, Del Giudice reports a significant gender difference in insecure attachment: Whereas females are more likely to be ambivalent, males are more likely to be avoidant. However, gender differences have rarely been reported in prior studies. We believe that a cross-cultural perspective may help to reconcile this apparent contradiction. In particular, attachment studies in Asian cultural samples, such as China, should be taken into account for a more comprehensive analysis. Our recent studies in China suggest that there are no gender difference in insecure attachment styles (Li & Du 2005; Li & Kato 2006; Li et al. 2006a; 2006b; 2006c; 2007; 2008; Wan & Li, in press). Table 1 summarizes the results of our relationship questionnaire (RQ) among various samples, including middle school and college students, company employees, and inpatients. Pearson Chi-square tests showed that neither sample had significantly different attachment patterns between males and females. We also note that in the urban mother sample, anxious/ambivalent is most popular among three insecure patterns, with 60.8% secure, 18.5% dismissing, 10.8% preoccupied, and 9.9% fearful (Li 2005). Del Giudice argues that males and females strive to maximize their reproduction of genes. Gender differences in mating, reproduction, and parenting efforts lead to diverse attachment styles: insecure females tend to be anxious/ambivalent, while insecure males tend to be avoidant (sect. 6.3.1, para. 5). However, reproductive investment alone does not account for the total cost of reproduction and parenting. Females have the privilege of selecting the most suitable male to help with child-rearing (CluttonBrock 1991). Transitional China, since the 1980s, has been one such example, where parental investment is significantly higher than that in Western nations (Wang & Ollendick 2001). During the 1980s, the Chinese government began to implement a family planning (“one child”) policy to control population growth; this policy profoundly changed the demographic as well as cultural values in Chinese society (Arnold & Liu 1986; Xu et al. 2007). First, this policy does not allow males to have multiple children, which requires males to invest in the quality

Commentary/Del Giudice: Sex, attachment, and the development of reproductive strategies Table 1 (Li et al.). Gender-based comparison of attachment patterns from Chinese samples Sample Middle-school students (Chi-square [3] ¼ 0.717; P ¼ .869) Male Female College students (Chi-square [3] ¼ 5.901; P ¼ .117) Males Females Company employees (Chi-square [3] ¼ 4.136; P ¼ .247) Males Females Inpatients (Chi-square [3] ¼ 3.156; P ¼ .368) Males Females

of offspring, rather than in the quantity (Wang & Ollendick 2001). This greatly reduces the likelihood of males taking a zero-parenting strategy. Second, the traditional preference for sons was even exaggerated, and the “one child” policy often became a “one son” policy, creating an unbalanced gender ratio (Chan et al. 2006). In this case, males have to compete for a limited number of females. Finally, the women’s rights movement has been widespread since the communist liberation in the early 1950s, when the socio-economic status of women improved considerably. Recent studies have shown that during family purchase decisions, females now play a status role equal to that of males (Dong & Li 2007). Thus, for contemporary Chinese females, although they cannot shift the balance between parenting and mating effort as easily as men do, they do not need to develop an anxious/ambivalent attachment strategy to invite paternal investment. A gender difference in insecure attachment could also be explained from the perspective of intergeneration transmission. According to Bowlby (1980), people develop their mental representations of the environment and significant others on the basis of their experience with parents or other caregivers. Bowlby labeled this mental representation as an internal working model (IWM). Once formed, IWMs tend to remain stable for the person’s entire lifespan (Hu & Meng 2003). The stability of IWM produces similar attachment patterns from childhood to adulthood. This argument is supported by cross-sectional and longitudinal studies (Brennan et al. 1998; Durrett et al. 1984; Fraley & Spieker 2003; Hu & Meng 2003; Li & Kato 2006; Nakao & Kato 2003). Li (2006) summarized the distribution of attachment styles in infants and adults in Chinese and American samples. He found that the proportion of each attachment style was similar for both infants and adults. This result suggests that the attachment style may remain relatively stable across the lifespan. Longitudinal studies on attachment development also support the stability of attachment styles within generations (Emery et al. 2008; Shemmings 2006). The stability of attachment from infancy to adulthood suggests that the influence of mate selection and sex competition in early adulthood on attachment patterns is trivial. This may well explain the lack of gender difference in insecure attachment in Chinese samples. In conclusion, we propose that the gender differences of insecure attachment are not universal, but rather, depend on culture input. In China, parenting strategies and intergeneration transmission result in similar attachment patterns between males and females. ACKNOWLEDGMENTS This study was supported by the Natural Science Foundation of China (No. 70572007). We thank Mark Sheskin and Adam Pearson for useful comments and suggestions.

Secure

Dismissing

Preoccupied

Fearful

51 46

25 21

21 18

4 6

191 180

109 93

109 80

34 47

106 107

61 54

50 30

32 33

10 8

12 8

2 4

5 1

The contribution of comparative research to the development and testing of life history models of human attachment and reproductive strategies doi:10.1017/S0140525X09000193 Dario Maestripieri Department of Comparative Human Development, The University of Chicago, Chicago, IL 60637. [email protected] http://primate.uchicago.edu/dario.htm

Abstract: Research with nonhuman primates can make important contributions to life history models of human attachment and reproductive strategies, such as: including parental responsiveness into female reproductive strategies, testing the assumption that adult attachment is a reproductive adaptation, assessing genetic and environmental effects on attachment and reproduction, and investigating the mechanisms through which early stress results in accelerated reproductive maturation.

Life history theory is a branch of evolutionary biology that deals with the trade-offs in the allocation of time and resources over an organism’s lifespan, as Del Giudice discusses in the target article. Concepts and data from animal research played a central role in the development of life history theory. Animal research can also make an important contribution to the development and testing of life history models of human reproductive strategies. In particular, given the similarities in parenting, attachment, lifespan development, and reproduction between humans and other primates (Kappeler & Pereira 2003; Maestripieri 2003; 2005b), studies of nonhuman primates can make a significant contribution to our understanding of human attachment and reproductive strategies. In rhesus monkeys, infants possess an attachment system whose design features, ontogeny, and adaptive functions are very similar to those of the infant attachment system in humans (Maestripieri 2003; Maestripieri & Roney 2006). These similarities suggest that the attachment system is not a product of the modern human environment, but rather, an adaptation with a phylogenetic history that can be traced back to the common ancestor of humans and Old World monkeys. The infant attachment system in rhesus monkeys is best viewed as an ontogenetic adaptation with the specific function of increasing infant survival during a period of high vulnerability and dependence on a caregiver (Maestripieri & Roney 2006). Attachment theorists have hypothesized that the attachment relationship with a caregiver becomes a template for other relationships later in life, and especially for sexual and romantic relationships. In this view, attachment would be an adaptation not only for early survival, but also for reproduction. BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

37

Commentary/Del Giudice: Sex, attachment, and the development of reproductive strategies The hypotheses that human adult romantic attachment is an adaptation and that different attachment styles represent different reproductive strategies are currently not supported by strong empirical evidence. These hypotheses, however, could be supported by comparative and phylogenetic evidence showing that attachment serves reproductive functions in closely related primate species, and that humans and these primates are likely to share this reproductive adaptation by virtue of common descent (Maestripieri 2005b; Roney & Maestripieri 2002). Unfortunately, this comparative and phylogenetic evidence is currently lacking. Because the sexual and mating behavior of nonhuman primates appears to be fully accounted for by sexual selection theory (Kappeler & van Schaik 2004), the relationship between attachment and mating has not been investigated in primates. It is, of course, possible that attachment has acquired new reproductive functions in humans, which are not shared with nonhuman primates. As is generally the case with all negative evidence, the failure to find an association between adult attachment and reproduction in nonhuman primates would be difficult to interpret. However, evidence that variation in attachment is associated with variation in reproduction among closely related primates would provide important support for one of the crucial assumptions of the model presented in the target article. Another assumption of the model is that the timing of menarche is an important expression of female life history strategies in humans, and that variation in the timing of menarche can differentiate between individuals pursuing different reproductive strategies. Although this assumption is probably correct, research with rhesus monkeys and other primates has suggested that the onset of menstrual activity is only one of a suite of reproduction-related traits associated with different life history strategies. In mammals and birds, variation in offspring survival accounts for the largest fraction of variation in female reproductive success (Clutton-Brock 1988). Accordingly, in rhesus monkeys, the success of a female’s first reproductive attempt depends to a large extent on the amount of parenting experience the monkey acquired as a juvenile. Therefore, early menarche and sexual activity are accompanied, and most often also preceded, by early and intense interest in other females’ infants (Maestripieri & Roney 2006). Consistent with the predictions of Del Giudice’s life history model and its predecessors, research with rhesus monkeys has shown that females exposed to harsh and unpredictable parenting in infancy are more interested in infants early in life than females without this stressful experience (Maestripieri 2005a). By using cross-fostering experiments, we were able to disentangle the effects of early stress from genetic similarities between mothers and daughters, and by collecting physiological data we were able to show that the effects of early stress on the development of parental responsiveness are mediated by longterm changes in the activity of the hypothalamic-pituitaryadrenal axis (Maestripieri 2005a). Early and intense interest in infants was also observed among adolescent girls who grew up without their fathers at home and had early menarche (Maestripieri et al. 2004). Therefore, parental responsiveness is an important variable that should be included in life history models of human attachment and reproductive strategies. Studies of nonhuman primates can enhance our understanding of different components of human reproductive strategies, both conceptually and empirically. Lack of control for genetic effects on variation in attachment and reproduction, and lack of knowledge of the physiological mechanisms through which early stress affects reproductive maturation, are some of the limitations of life history models of human attachment and reproductive strategies. Experimental studies of nonhuman primates in which genetic and physiological variables can be manipulated provide opportunities to test some of the assumptions and predictions of models of reproductive

38

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

strategies in ways that would not be possible in humans. They also provide the opportunity to conduct longitudinal studies of lifespan development and reproduction in a relatively short period of time. Therefore, proponents of human life history models should explicitly encourage the testing of these models with comparative data, and acknowledge that these data can provide important evidence concerning the adaptive function, physiological regulation, ontogeny, and phylogeny of human social and reproductive behavior.

Adaptive developmental plasticity might not contribute much to the adaptiveness of reproductive strategies doi:10.1017/S0140525X0900020X Lars Penke MRC Center for Cognitive Ageing and Cognitive Epidemiology, Department of Psychology, University of Edinburgh, Edinburgh EH8 9JZ, Scotland, United Kingdom. [email protected] http://www.larspenke.eu

Abstract: Del Giudice’s model belongs among those that highlight the role of adaptive developmental plasticity in human reproductive strategies; but at least three other forms of evolutionary adaptation also influence reproductive behavior. Similar to earlier models, the existing evidence suggests that Del Giudice’s hypothesized effects are rather weak. In particular, adult attachment styles are hardly predictive of outcomes visible to natural selection.

Del Giudice presents a thoughtful overview, integration, and extension of the now copious literature on what is arguably the most influential developmental hypothesis in modern evolutionary psychology: Children infer environmental risk from cues within their families and adjust their development so that they are well adapted to the reproductive conditions they will face as adults. This is a case of adaptive phenotypic plasticity by conditional development, or adaptive developmental plasticity. Theoretically, adaptive developmental plasticity is a perfectly plausible form of evolutionary adaptation (Pigullici 2005; WestEberhardt 2003). However, there are at least three other forms that are equally plausible, and they can all be aligned along a dimension of spatiotemporal environmental stability (Fig. 1). When fitness-relevant environmental features are stable over tens of thousands of years or longer, organisms can evolve

Figure 1 (Penke). Four forms of evolutionary adaptation. They should be understood as distinguishable points along a continuum, not as distinct categories: Balanced genetic variants can get fixated in the population and thus contribute to evolved adaptations, or they can underlie individual differences in either of the two forms of phenotypic plasticity (Belsky 2005; Pigullici 2005), which themselves only differ in how quickly they react to the environment. Which mechanism governs adaptation depends on the spatiotemporal stability of the adaptively relevant environmental features. Different aspects of complex adaptations like life history strategies can be influenced by different mechanisms.

Commentary/Del Giudice: Sex, attachment, and the development of reproductive strategies universal adaptations that reliably develop every generation (Tooby & Cosmides 2005). Examples from the domain of human reproductive strategies include the romantic attachment system, which likely evolved in response to the high degree of parental care demanded by human offspring (Fraley et al. 2005), and sex differences in the desire for sexual variety, which are basically adaptive so long as women get pregnant and men do not (Schmitt et al. 2003b). When the environment is less stable and tends to fluctuate, balancing selection by environmental heterogeneity can maintain more adaptive genetic variants at higher frequencies in the population (Penke et al. 2007b). For example, it has been argued that the phenotypic effects of the seven-repeat allele of the DRD4 polymorphism were more adaptive in societies in which reproductive success is dependent on social competition, whereas the four-repeat allele was likely more advantageous when environmental harshness demanded biparental cooperation (Harpending & Cochran 2002). A similar logic might hold for the heritable components of traits related to reproductive strategies (e.g. Schaller & Murray 2008), including the polymorphisms affecting children’s sensitivity to rearing environments in Del Giudice’s model (Belsky 2005). However, it will likely not hold for the genetic foundations of the “K-factor,” which is far less plausible from an evolutionary genetic perspective (Penke et al. 2007a; 2007b). Even less stable and more heterogeneous environments favor the evolution of adaptive phenotypic plasticity (Hollander 2008), which includes developmental plasticity, as discussed by Del Giudice, and much faster adaptive conditional adjustments of life history strategies to the current environment. Examples of the latter include adjustments of strategic mating decisions to momentarily faced environmental harshness, quality of available mates, or sex ratio and competition on the local mating market (Gangestad & Simpson 2000; Penke et al. 2007c; Lenton et al., in press). Importantly, romantic attachment styles also show considerable plasticity during adulthood and might even be relationship-specific (Lehnart & Neyer 2006). These four different forms of adaptation are not mutually exclusive. I agree with Del Giudice that they will likely all contribute to individual differences in reproductive strategies in a probabilistic manner. However, the critical – and ultimately empirical – question is their relative importance. And this is where I find adaptive developmental plasticity hypotheses of reproductive strategies problematic. When the earlier models that predicted pathways from childhood stress to age of menarche in girls to adult reproductive strategy were empirically tested, hardly any evidence could be found (Ellis 2004; Hoier 2003; Neberich et al., in press). These results led some researchers to retract reproductive strategies altogether and to concentrate on the stress– menarche link (Ellis 2004). Del Giudice’s model, on the other hand, attempts to rescue the causal relationship between childhood stress and adult reproductive strategy by relying much more on attachment styles as the mediating factor and introducing some elegant theoretical refinements, including sex differences and children’s attachment styles as disposable phenotypes. However, although there is abundant evidence that adult attachment styles relate to the construal and experience of romantic relationships (Birnbaum et al. 2006; Feeney 1999), there seems to be surprisingly little evidence that romantic attachment styles actually relate to reproductive strategy-related consequential behavioral outcomes. This is a crucial point, because only consequential behaviors, not subjective experiences, are visible to natural selection and can thus be reasonably explained within an evolutionary framework. To give an example, sociosexuality shows almost no relationship with attachment styles (Schmitt 2005a). Strikingly, only restricted sociosexual attitudes, but not sociosexual behaviors, were related to attachment styles in a study by Jackson and Kirkpatrick (2007), but Penke and Asendorpf (in press) showed that attitudes were the only component of sociosexuality not related to a variety

of behavioral outcomes, much like self-reported mate preferences are unrelated to actual mate choices (Todd et al. 2007). As another example, attachment styles are not predictive of romantic relationship stability once relationship duration is taken into account (Lehnart & Neyer 2006), and avoidant men and anxious women can have as stable relationships as securely attached people, no matter how satisfied they are with it (Kirkpatrick & Davis 1994). Even the sex differences in insecure adult attachment styles, which enjoy a prominent role in Del Giudice’s model, are in fact quite modest in size (Schmitt 2005a; Schmitt et al. 2003a, being much smaller than in other mating-related dispositions (e.g., Schmitt 2005b; Schmitt et al. 2003b). Indeed, it could be argued that their size, even in harsher environments, is too small to be theoretically meaningful (Hyde 2005). To conclude, although the available evidence is clearly insufficient to fully evaluate Del Giudice’s complex model, it suggests that adaptive developmental plasticity might not account for much variance in reproductive strategies. The theoretical reason for this could be straightforward: During human evolution, environmental changes in reproductive conditions over a few generations were probably less important for successful propagation than changes over much longer or shorter time spans, which lead to universal adaptations, polymorphisms under balancing selection, and adaptive conditional adjustments related to reproductive strategies. Still, I am confident that the myriad of intriguing ideas in Del Giudice’s article will inspire future studies, which will hopefully confirm how big or small the contribution of adaptive developmental plasticity to reproductive strategies really is. ACKNOWLEDGMENTS The author is funded by the UK Medical Research Council. The UK Medical Research Council and the University of Edinburgh provide core funding for the MRC Centre for Cognitive Ageing and Cognitive Epidemiology, which supported this research. Thanks to Jaap Denissen and Michelle Luciano for helpful comments.

Modeling, simulating, and simplifying links between stress, attachment, and reproduction doi:10.1017/S0140525X09000211 Dean Pettersa and Everett Watersb a

Department of Psychology, Aston University, Birmingham, B48 7NR, United Kingdom; bDepartment of Psychology, State University of New York, Stony Brook, NY 11794-2500. [email protected] http://www.cs.bham.ac.uk/~ddp [email protected] http://www.johnbowlby.com

Abstract: John Bowlby’s use of evolutionary theory as a cornerstone of his attachment theory was innovative in its day and remains useful. Del Giudice’s target article extends Belsky et al.’s and Chisholm’s efforts to integrate attachment theory with more current thinking about evolution, ecology, and neuroscience. His analysis would be strengthened by (1) using computer simulation to clarify and simulate the effects of early environmental stress, (2) incorporating information about non-stress related sources of individual differences, (3) considering the possibility of adaptive behavior without specific evolutionary adaptations, and (4) considering whether the attachment construct is critical to his analysis.

One of the key innovations in Bowlby’s attachment theory was to replace Freud’s drive theory with a motivational model based on control systems theory. Concerned that this might seem like replacing one bit of magic with another, Bowlby turned to evolutionary theory to explain how an infant could be endowed with an attachment control system. Citing many examples, he argued that evolution shapes not only physical structures but also species learning abilities. Attachment is not an instinct or a preprogrammed blueprint ready to be activated by critical experiences; it is the capacity BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

39

Commentary/Del Giudice: Sex, attachment, and the development of reproductive strategies to construct a behavioral system through the interaction of speciesspecific learning abilities with information available in the organization of what Bowlby called an average expectable caregiving environment. This was an innovative solution to a difficult problem and continues to serve well despite the fact that Bowlby’s emphasis on the value of attachment behavior as an adaptation to predation pressure today seems more “classical” than modern (Waters 2002). 1. Modeling stress and adaptation. Early exposure to environmental stress is central to Del Giudice’s analysis. Yet he says little about what constitutes relevant stress and, more importantly, how its characteristics over time might bear on the costs and benefits of different reproductive strategies and patterns of parental investment. Simply put, the effectiveness of behavioral strategies depends on context. Imagine a behavioral ecologist examining avian foraging strategies without considering the implications of foraging for food that is plentiful or spare, concentrated or dispersed, consistently or intermittently available, and so on. In the case of early experience, attachment, and reproduction, it seems likely that the costs and benefits of any particular strategy would depend very much on the incidence, persistence, duration, mortality risk, and other parameters of the environmental stress, as well as on key features of a species’ life history strategy. Formalizing the properties of a stressful environment as computational models and using simulation to investigate the costs and benefits of different mating and parental strategies through a wide range of such parameters would enrich Del Giudice’s analysis, exploring the robustness of his hypotheses and predictions and possibly highlighting some interesting circumstances that deserve special attention. There are a number of existing approaches that might be adapted for this purpose; for example, simulations concerned with the evolution of protection periods (Bullinaria 2007), the growth of social complexity (Doran 1994), and simulation of secure-base behavior (Petters 2006a). 2. Attachment patterns without stress. In light of Del Giudice’s emphasis on the role of stress in shaping attachment patterns, it is useful to consider that individual differences can arise without stark ecological stressors. A dismissing or preoccupied adult or an avoidant or resistant baby is not necessarily one whose experience has been pathological. Temperament, garden-variety diversity in caregiving experiences, and idiosyncratic interpretations of personal experience guarantee, even in benign environments, a rich diversity of attachment patterns within individuals and across time. This point is underlined in computational experiments (e.g., Petters 2006b) in which distinct attachment styles can arise merely from the action of positive feedback loops acting upon small, random differences in the environment. Del Giudice’s analysis would be considerably strengthened if it were expanded to address conceptually and in terms of physiological mechanisms, how, in light of these non-stress related sources of individual differences, the attachment system’s sensitivity to stress could be tuned so it is neither insensitive to significant stressors nor overly responsive to minor perturbations. 3. The adaptationist fallacy. In Del Giudice’s analysis, the effects of early environmental stress on attachment include effects on attachment-related physiology. These are interesting and testable hypotheses. However, it is worth keeping in mind that adaptive behavior does not always imply an underlying “adaptation.” As Bowlby (1969/1982) argued, attachment relations can be represented at a variety of levels, from the organization of underlying physiological systems, to reactive fixed action patterns, to mental representations (internal working models) and natural language (Bowlby 1969/1982). Accordingly, adaptive behavior might also be explained by cognitive processes – allowing early experience to shape mental representations and social perception in ways that impact adaptively on courtship, mating, and parenting. Nor are these two routes to decision making independent or mutually exclusive. Designing computer simulations that integrate lower-level reactive mechanisms with simple reasoning processes (e.g., Petters 2006a, Ch. 4) seems a

40

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

promising approach to exploring the roles of evolved adaptations and rational adaptive behavior. 4. Is attachment critical to the argument? As Del Giudice notes, there are two distinct traditions in attachment assessment: one based on infant observation and adult interviews, and the other based on adult self-report questionnaires. Both traditions are inspired by Bowlby’s theoretical work, and both use similar terminology to describe the constructs they measure, describing individuals as secure versus anxious or insecure, and as anxious, avoidant, ambivalent, and so on. Both approaches have produced useful empirical results. However, measures from the two traditions do not yield correlated scores and have rather different patterns of stability and change, different behavioral correlates, and different relevance to courtship, marriage, and parenting (Waters et al. 2002). Although acknowledging some of the difficulties here, Del Giudice tends to treat similarly named measures as if they were interchangeable. This is too much to expect readers to track and undermines the conceptual and empirical foundations of key elements in his analysis. In addition, the correlations underlying links between types of insecure attachment and traits, such as aggressive/self-aggrandizing or fearful, passive, withdrawing, and so on, are typically very small or inflated by considerable method variance. Such low correlations, and the fact that they represent data from different measures and different age groups, substantially attenuate the empirical link between attachment and the patterns of mating and parental behavior at the heart of Del Giudice’s analysis. We wonder whether the link between early stress and later reproductive behavior might be better argued by having early experience directly affect approach and avoidance systems – without mediation through attachment. 5. Conclusion. Del Giudice’s analysis of links between early stressful environments, attachment, and reproduction illustrates some of the advantages – and some of the difficulties – of coordinating current attachment theory with current ideas about evolution, ecology, and life history strategies. There is no doubt that John Bowlby would have appreciated such efforts and looked forward, as we do, to further advances along these lines.

Predicting cross-cultural patterns in sexbiased parental investment and attachment doi:10.1017/S0140525X09000223 Robert J. Quinlan Department of Anthropology, Washington State University, Pullman, WA 99164-4910. [email protected] http://www.wsu.edu/~rquinlan/

Abstract: If parenting behavior influences attachment, then parental investment (PI) theory can predict sex differences and distributions of attachment styles across cultures. Trivers-Willard, local resource competition, and local resource enhancement models make distinct predictions for sex-biased parental responsiveness relevant to attachment. Parental investment and attachment probably vary across cultures in relation to “local fitness currencies” for status, wealth, and well-being.

Attachment may play a crucial role in the development of human mating and parenting behaviors. Hence, the evolutionary design of attachment could be revealed in relations among risk, resources, and parenting. Del Giudice presents a welcomed and state-of-the-art synthesis of attachment research in evolutionary context. The task ahead is to explore empirical avenues to test and refine predictive models. Data and hypotheses concerning associations between attachment and environmental conditions across populations are particularly scarce, suggesting a fruitful area for future research. Del Giudice’s

Commentary/Del Giudice: Sex, attachment, and the development of reproductive strategies analysis of sex differences in attachment suggests some intriguing questions: What environmental conditions promote boys’ and girls’ attachment styles, either to diverge into more pronounced patterns of avoidance and anxiety in some populations, or to converge, creating higher proportions of secure individuals in other populations? Is it possible for one sex to show a high proportion of security while the other sex experiences greater insecurity? Do environmental effects on attachment shape patterns of cultural diversity in reproductive behavior? Addressing these questions could enhance our understanding of attachment and its role in larger cultural patterns related to mating, parenting, family, and risk. Here, I focus on sex-biased parental investment and predicted associations with reproductive and family behaviors cross-culturally. Sex-biased parental care may be expressed as differential parental responsiveness to boys and girls, which could promote divergent attachment styles for males and females. Parental investment theory offers standard models for sex-biased investment (Clutton Brock 1991). Trivers and Willard’s (1973) model describes several conditions: (1) one sex has higher variance in reproductive success than the other; (2) offspring’s reproductive success is sensitive to the parental care they received; and (3) parental care is positively correlated with parental condition. Under these assumptions, parents in good condition could maximize long-term fitness by biasing their attention toward children of the sex with higher reproductive variance. Conversely, parents in poorer condition should bias attention toward the sex with less reproductive variance. Patterns of parental investment fit this model in some human populations (Cronk 2000). Under Trivers-Willard conditions, we expect to see more secure males and insecure females among relatively wealthy families, and more secure females and insecure males among relatively poor families. The population or culture-level implications of this attachment pattern are intriguing. For example, marriage patterns between the high-status Masai and low-status Mukogodo of Kenya include hypergamy, where Mukogodo families encourage daughters to marry into neighboring, higher-status Masai families. This marriage preference creates a problem for Mukugodo men, who sometimes have trouble finding mates; hence, Mukugodo parents tend to bias investment toward daughters. Daughter bias is evident early in infancy, when daughters are more likely to be seen suckling and being held than are sons (Cronk 2000). This is precisely the kind of sex bias in child care that could enhance divergences in attachment style. Trivers-Willard fails to predict parental investment in many human and other primate populations, probably because additional factors can affect sex-specific returns on parental investment. Local competition among same-sex siblings is a common complication that can weaken the Trivers-Willard effect. If one sex competes for parental resources, then competition can create lower offspring fitness returns per unit of parental investment. Local competition among brothers is not uncommon cross-culturally (Borgerhoff Mulder 1998; Quinlan et al. 2005), and it may promote female-biased investment, leading to higher levels of security among girls. Local competition can be quite strong in polygynous societies, in which brothers compete for access to parental resources crucial for accruing multiple wives. In that case, we expect relatively high levels of avoidance among males and higher levels of secure attachment among females. Note that local mate competition in polygynous societies is probably associated with high variance in male reproductive success – a key element in the Trivers-Willard effect – which can create a series of rather complex parental investment decisions. For example, among the agro-pastoralist Kipsigis of Kenya, polygyny, livestock payments made for brides, and patterns of sibling interaction create a mosaic of parental investment considerations that fit multiple models of parental investment (PI) (Borgerhoff Mulder 1998). Predicting attachment patterns among groups like the Kipsigis will require close attention to underlying

parental investment concerns related to “local fitness currencies” – local resources and relationships that are associated with cultural and reproductive success. A third pattern of sex-biased parental investment may present additional challenges for evolutionary theories of attachment. Local resource enhancement occurs when offspring of one sex help care for siblings or other relatives, which reduces the costs of parental effort. Parents tend to bias investment toward helpers because a portion of that investment is repaid through work that can enhance parental fitness. In that case, helpers can receive more direct parental care than non-helpers (Quinlan et al. 2005). This empirical pattern is at odds with the proposed attachment style promoting helping at the nest: Helpers are predicted to have insecure clingy styles (Chisholm 1996); yet biased parental care under local resource enhancement suggest that helpers could be securely attached in many circumstances. The role of anxious attachment style (if any) in promoting helping-at-the-nest requires verification in societies with substantial cooperative childrearing at the household level, which is common cross-culturally but relatively rare in many Western industrial populations. Sex-specific risks (and saturation points for parental effort) can further complicate parental investment decisions. In some populations, one sex can face greater extrinsic risk than the other. For example, in rural Dominica, boys and men consistently experience greater fitness risks and fewer opportunities than girls and women do. Asymmetry in risk may promote daughter-biased parental investment that includes later weaning (by 5 months in Dominica), greater direct parental care during childhood, and more investment in education in adolescence (Quinlan 2006). Attention to sex-specific risks will probably prove instructive in the years to come. Not all parental investment occurs during the sensitive period for attachment. Some parental resources may affect fitness, but they can have little impact on attachment. Family wealth could be channeled toward sons or daughters independent of parental responsiveness in early childhood. Parents in Germany, for example, invest substantial wealth in offspring, but parental care in infancy and early childhood promotes independence and a degree of physical and psychological distance that may seem distressing when viewed from other cultural models of parenting (Levine & Norman 2001). In sum, patterns of investment and their likely influence on attachment depend on local fitness “currencies” based on prevailing or anticipated economic, ecological, and social factors. Accurately predicting the distribution of attachment styles in and across populations will probably require close attention to multiple local dimensions of parental investment.

Neuroendocrine features of attachment in infants and nonhuman primates doi:10.1017/S0140525X09000235 Leslie J. Seltzera and Seth D. Pollakb a

Department of Anthropology and Waisman Center, University of Wisconsin– Madison, Madison, WI 53705; bDepartments of Psychology, Anthropology, Pediatrics, Psychiatry, and Waisman Center, University of Wisconsin-Madison, Madison, WI 53705. [email protected] [email protected] www.waisman.wisc.edu/childemotion

Abstract: The translation of research findings from other primates to humans, and from infants to adults within our own species, requires great care. If the many neurological, behavioral and adaptive distinctions between these groups are not precisely defined and considered, erroneous conclusions about evolutionary history and developmental processes may result.

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

41

Commentary/Del Giudice: Sex, attachment, and the development of reproductive strategies Research on primates reveals the myriad ways in which animals can alter their behavior to accommodate environmental change. Del Giudice’s argument that human sexual behavior may also be facultative represents a significant contribution to this larger literature, with important implications for both human evolutionary biology and developmental psychology. We applaud this effort to unite these fields; indeed, it is likely that many aspects of human emotional behavior are important targets for natural selection. Extrapolation of findings across age groups or species, however, is rarely straightforward, and the argument articulated by Del Giudice would be strengthened by greater sensitivity to the inherent limitations and challenges of such an enterprise. We illustrate this point in two ways. First, we underscore the need for precision when drawing analogies between humans and other primates by discussing how adaptive physiological mechanisms in other taxa may not act correspondingly in our own species. Second, we advocate caution when attempting to map specific biological systems onto relatively unspecified behaviors or feeling states. Emotions, such as those that mediate relationships in adult humans, are often difficult to assign clearly into those that are (or were) adaptive and those that are (or were) not. An example of our first point is highlighted by Del Giudice’s suggestion that human females practice facultative reproductive suppression by developing a lack of interest in sexual relationships when the social support systems necessary to help with childrearing are absent. While it is probable that humans are cooperative breeders, reproductive suppression in the context of cooperative breeding in other primates is accompanied by specific physiological, behavioral, and neuroendocrine mechanisms that have no parallel in humans. The best-known data concerning this phenomenon come from the marmosets and tamarins of the New World, who appear to utilize reproductive suppression in the context of kin selection (Hamilton 1964). Subordinate individuals in these species assist in the rearing of the offspring of a much smaller number of related dominants, who are typically the only ones who become pregnant (Carlson et al. 1997). This is caused by a pheromonal – behavioral mechanism that regulates the process whereby dominant females prevent subordinate females living in the same group from ovulating (Barrett et al. 1990). In fact, prevention of ovulation is the most widely used definition of reproductive suppression, with clear physiological correlates. For example, the ovaries of dominant females are 50% greater in volume than those of subordinates and have many more antral follicles, whereas subordinates lack corpus lutea and do not release sufficient lutenizing hormone for ovulation to occur (Abbott et al. 1998). When a subordinate female is removed from her natal social group, however, this constraint is released, which can lead to pregnancy as few as eight days later (Ziegler et al. 1987). In other words, it is the very presence of related females who could provide rearing support that causes reproductive suppression, not their absence. This is opposite to the scenario envisioned by Del Giudice. Although the term “reproductive suppression” does have other applications in the biological sciences, it is misleading to use this term to refer to mating avoidance in healthy females who are otherwise capable of reproducing successfully. In this case, Del Giudice’s argument suggests a parallel between primate physiology and human behavior that is incorrect: There is no evidence indicating that such a phenomenon occurs in human females. Our second and related point is that it is difficult to map clear, well-defined physiological mechanisms onto general behavioral constructs in humans. Del Giudice demonstrates creativity and thoughtfulness in attempting to link neurophysiology and evolution to a construct such as “attachment.” As he defines “attachment,” however (and indeed, how most psychologists now use the term), it is unlikely to be tied to any clear biological circuitry. We agree that discontinuity between infant and adult attachment is likely even though some of the same neural and endocrine systems, such as oxytocin-mediated social bonding, are involved.

42

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

The author makes this point as well, suggesting that adrenarche represents a hormonal disconnect of behaviors advantageous in infancy from those that may be advantageous in adulthood. What is lacking is a precise, neurologically plausible definition as to what attachment means in human adults, as well as reliable tools for measuring it. In infancy, attachment is characterized by a cessation of exploration, initiation of proximity to the caregiver, or distress if the caregiver is unavailable. The same behaviors are apparent in other primate infants as well, suggesting that intra-species comparisons of attachment may be justified in this specific case. In chimpanzees, for example, the mature caregiver provides a secure base from which the infant can explore and seek comfort, and one can interpret changes in hypothalamic-pituitary-adrenal axis activity as an index of the caregiver’s effectiveness in reducing infant stress (Miller et al. 1986). In human adults, however, attachment is construed broadly as a personality variable that is not amenable to intercross-species comparisons analyses. For example, adult attachment is often operationalized as the coherence of narrative responses to questionnaires or interviews. Such information is undoubtedly a rich source of data, but this type of data taps into cultural expectations, is difficult to relate to the behavioral and physiological phenomena observed in human and nonhuman infants, and is not directly related to the social behaviors observed in other species. For these reasons, it is difficult to reach firm conclusions about the evolutionary significance of attachment as it is construed in human adults. We admire Del Giudice’s thesis as a noteworthy effort toward a better understanding of the evolutionary underpinnings of modern human adult relationships. Empirical studies of human biobehavioral plasticity, and the adaptive advantages such plasticity may confer, require thoughtful integration across species and across the ontogenetic spectrum, with special attention paid to the role of species-typical and species-atypical contexts. When done appropriately, such research is likely to excavate the biobehavioral processes that promote social competencies and health.

Attachment styles within sexual relationships are strategic doi:10.1017/S0140525X09000247 Douglas K. Symons and Alicia L. Szielasko Psychology Department, Acadia University, Wolfville, Nova Scotia B4P 2R6, Canada. [email protected] [email protected] http://ace.acadiau.ca/science/psyc/dsymons/

Abstract: Del Giudice’s examination of sex differences in reproductive strategy within an attachment context is well taken. Sex has been studied as behavior within romantic relationships, but attachment styles should also be reflected in strategic behavior within relationships that are sexual. This seems particularly true within adolescence, and sex differences may be better reflected as differences in correlation patterns of process variables than as main effects models.

Attachment theory (e.g., Bowlby 1969/1982) has been used so widely in research and practice that it is easy to forget its theoretical roots in control systems, ethology, and the development of behavior within an evolutionary context. Del Giudice correctly directs attention on a neglected component of this theory: human reproductive strategies. Although attachment, caregiving, and sex have historically been argued to be three independent systems (e.g., Ainsworth 1989), this suggestion seems based on comparative data on the behavioral manifestations of mate attraction and the sexual response. Sex is a behavior that happens within romantic relationships; for example, Davis et al. (2004) portray sex as a behavior that serves attachment needs (see also

Commentary/Del Giudice: Sex, attachment, and the development of reproductive strategies Butzer & Campbell 2008). But the evolutionary context of relationships involves a combination of these three systems (see Mikulincer 2006), and Del Giudice’s proposal that sex differences in attachment have adaptive significance places an emphasis on this very issue. That is, attachment processes are an essential component and are thus reflected in sexual relationships, which follows on the respective work of Belsky (2007; Belsky et al. 1991) and others (e.g., Brumbaugh & Fraley 2006; Chisholm 1999). Del Giudice encourages an essential expansion of the evolutionary elements of attachment because adaptive significance includes infant survival, sexual relationships, and, then, parenting commitment to the young in the next generation. In general, attachment research has focused primarily on the first and last of these three components; but what Del Giudice refers to as the “double life” of the attachment system through care-eliciting and pair-bonding needs to expand on the reproductive strategy component. This should be particularly true in adolescence, when sex is a dominant focus of attention, teens typically have a series of sexual partners, and risky sexual behavior poses health risks that have immediate and long-term consequences for reproductive life history strategies. Three points are made in this commentary: (1) that sexual behavior needs to be examined within the attachment context of sexual relationships; (2) that this seems particularly true within adolescence, when sex is a dominant theme and goal; and (3) that sex differences need to be examined as differences in correlation patterns in addition to main effects models. Research on adult romantic attachment addresses intimacy, dependency, and trust issues within “emotionally intimate relationships” (Bartholomew & Horowitz 1991; Hazan & Shaver 1987; for review, see Hazan et al. 2006), which is related to some aspects of sexual behavior (see Brennan et al. 1998; Davis et al. 2004; Milkulincer & Goodman 2006). However, romantic attachment is not necessarily synonymous with sexual relationships. While romantic relationships usually contain a sexual element, not all sexual relationships are romantic ones (e.g., sex between friends, one-night stands), nor are all romantic relationships perceived as such after they have ended. Attachment insecurity may manifest itself in promiscuity, using sex to maintain a relationship, sexual coercion, and separation of sex from emotional caring for a partner, but these are domain-specific behaviors and motivations within sexual relationships (reviewed in Feeney & Noller 2004). As noted by Belsky (2007), these vary from opportunistic advantage-taking approaches to sex (i.e., avoidant) to dependent helper-at-nest kinds of approaches (e.g., ambivalent/ preoccupied). When attachment-relevant selfreport items focus on behavior with sexual partners, there is the potential to predict strategic sexual behavior above and beyond that predicted by romantic relationships (see Szielasko et al. 2007; under review). This distinction between sexual and romantic partners may be particularly relevant to teens who are very interested in sex, typically have a series of sexual partners during adolescence and early adulthood, and may be more motivated by physical attraction than an evaluation of partners in terms of future co-parenting investment. Del Giudice frames this issue within life history theory when he distinguishes between mating effort and parenting effort, wherein teens are usually invested in the former and not the latter. Downward extensions of adult-oriented attachment measures (e.g., Allen & Land 1999; O’Connor & Byrne 2007) may mask the significance that sexual behavior and misbehavior have with regard to later relationships and developmental processes for late teens (see Feeney et al. 1993; Kobak et al. 2007). Keep in mind that human evolution took place over a time frame when pregnancy was not controllable, so adolescence is the developmental period when pregnancies historically arose, not adulthood, when current-day pregnancies are often fit in with other life history strategies of investment.

Del Giudice addresses sex differences in attachment, but most of the literature reviewed uses a main effects model of sex that describes avoidant males and ambivalent/dependent females (i.e., investment-eliciting). Although these patterns fit his theory, no doubt there are avoidant females and preoccupied males whose sexual adaptations must also be considered. If ambivalence is related to sexual coercion and avoidance related to promiscuity (see Bartholomew & Allison 2006; Feeney & Noller, 2004; Szielasko et al. 2007), for example, it would be important to know whether this is equally true for both males and females, and how coercive behavior may vary between the sexes when a relationship feels threatened (e.g., physical force? trickery?). Such a finding would not necessarily pose a problem for the proposed theory, as ultimately attachment styles reflect adaptations to environmental circumstances, but we argue for a need to address potentially different patterns of correlations between attachment constructs and sexual behavior/misbehavior in males and females. Del Giudice hints at this when discussing different implications of insecurity for the reproductive strategies of men and women, but he could go even further with this thinking. Literature suggests that avoidance is related to a greater number of less committed sexual relationships, and ambivalence is related to dependency and possessiveness, but is this equally true for both genders, or are there sex-specific manifestations of sexual behavior that reflect these attachment styles? And may these behaviors have different longitudinal consequences for personality development and attachment styles? Bowlby (1969/1982) wrote that it may be that “attachment and sexual behavior share certain components and causal mechanisms” (p. 233), and considering the adaptive significance of both within the context of adult relationships is essential. This article by Del Giudice promises to invigorate thinking on this component of attachment theory in an evolutionary context, building on the inspiration of Belsky et al. (1991) that took place almost two decades ago.

Attachment and sexual strategies doi:10.1017/S0140525X09000259 Lane E. Volpe and Robert A. Barton Medical Anthropology and Evolutionary Anthropology Research Groups, Anthropology Department, Durham University, Durham, DH1 3HN, United Kingdom. [email protected] http://www.dur.ac.uk/anthropology/ [email protected] http://www.dur.ac.uk/anthropology/staff/profile/?id¼122

Abstract: Sexual behaviour and mate choice are key intervening variables between attachment and life histories. We propose a set of predictions relating attachment, reproductive strategies, and mate choice criteria.

Del Giudice presents a meaningful extension of evolutionary models of attachment. The target article builds on existing theory to account for the adaptive consequences of various attachment strategies beyond early childhood relationships with caregivers. The central tenet of this body of theory is that attachment functions as a mechanism for adaptively adjusting human life histories and reproductive strategies to the socio-ecological environment in which children grow up. Del Giudice’s major contribution is to focus the issue of sex differences in attachment processes through the lens of sexual selection theory. This provides a cogent interpretation of the different relationships between attachment processes and behavioural and life history outcomes in males and females. It also brings into sharper relief the relatively neglected but vital socio-developmental processes occurring in middle childhood. These are theoretical steps BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

43

Commentary/Del Giudice: Sex, attachment, and the development of reproductive strategies forward, and ones that are consistent with the data reviewed by Del Giudice. The major challenge now is to develop and test a range of more critical empirical predictions. We propose one particular set of predictions for future research. The hypothesized link between attachment and adaptive variation in reproductive strategies (e.g., mating vs. parenting effort) implies that sexual relationships should be a key area for study. Much evidence suggests that women’s sexual strategies correlate with mate preferences on certain dimensions of physical attractiveness, such as male facial and vocal masculinity (Gangestad & Simpson 2000; Gangestad & Thornhill 2008; Provost et al. 2008). In combination with the arguments of the target article, this leads to the prediction that attachment styles should correlate with females’ preference for masculine faces and voices. In populations characterized by low paternal investment and high levels of aggressive male–male competition, there is a significantly stronger average preference among females for masculinized facial features than in populations characterized by higher paternal investment (Penton-Voak et al. 2004). The evolutionary logic is that, where there is low paternal investment, women’s mate choices should be more strongly biased towards males of high competitive ability. Similarly, women seeking short-term sexual relationships have a stronger preference for masculine features than those seeking long-term relationships (Little et al. 2002), again indicating a link between likelihood of paternal investment and preference for masculinity. The theoretical link to attachment processes is that female dismissiveness should correlate positively with masculinity preferences. On the other side of the coin, males with more masculine faces tend to have more sexual partners and be less interested in long-term relationships (Boothroyd et al. 2008). Consequently, facial masculinity should correlate with attachment profiles in males. Sexual strategies also vary according to the individual’s perception of their own “market value” (Little & Mannion 2006). An intriguing corollary of the target article is that these perceptions may correlate not only with physical attractiveness, but also with the experience of attachment relationships in middle childhood. This would predict an association between attachment profile and self-rated attractiveness. Finally, female sexual strategies have been shown to be hormonally influenced. During the menstrual cycle, women become more interested in short-term relationships, are more likely to be unfaithful to their long-term partner, and are more attracted to masculine facial and vocal features around the time of ovulation, when the risk of conception is high (Penton-Voak et al. 1999; Feinberg et al. 2006; Gangestad & Thornhill 2008). This has been interpreted as reflecting a “dual” sexual strategy, promoting mating with competitive but low-investing males while also garnering paternal investment via a long-term relationship. A further prediction is therefore that attachment profiles should vary cyclically, directly mirroring the cyclicity in sexual behaviour. This would imply that attachment profiles, albeit perhaps largely fixed early in life, also have some previously unsuspected adaptive plasticity.

What love has to do with it: An attachment perspective on pair bonding and sexual behavior doi:10.1017/S0140525X09000260 Vivian Zayasa and Daphna Ramb a

Department of Psychology, Cornell University, Ithaca, NY 14853-7601; Department of Human Development, Cornell University, Ithaca, NY 14853-7601. [email protected] [email protected] http://people.psych.cornell.edu/~pac_lab/ b

Abstract: Del Giudice proposes that short-term mating strategies are adaptive for attachment-avoidant men. We argue that this model

44

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

(1) does not apply to the majority of avoidant men (fearful-avoidants); (2) is based on limited evidence that the remaining subset of avoidant men (dismissing-avoidants) engage in short-term mating strategies; and (3) disregards the importance of pair bonding even for dismissingavoidants.

A central assumption of Del Giudice’s model is that insecurely attached male infants develop an avoidant attachment strategy in adulthood that is associated with short-term mating strategies that maximize fitness. In this commentary, we first argue that Del Giudice’s model ignores the majority of men who engage in avoidant attachment strategies, namely, fearful-avoidant men, who constitute approximately 21% of the population (Bartholomew & Horowitz 1991). Moreover, the model does not accurately apply to dismissing-avoidant men, who make up approximately 18% of the population. Second, we question Del Giudice’s claim that the “link between avoidance and short-term mating style is well supported by research in adult attachment” (sect. 5.2, para. 3), as well as the idea that shortterm strategies increase fitness for dismissing-avoidant men. Finally, we discuss a point overlooked by Del Giudice: the benefits of pair bonding on, and in the absence of, direct reproductive outcomes. Del Giudice claims that an avoidant attachment strategy is related to short-term mating strategies (e.g., an earlier onset of first sexual experience and higher frequency of sexual activity with different partners). However, empirical work does not unequivocally support this claim. There is evidence that avoidant male adolescents are less likely to ever have had sex, to have engaged in fewer sexual behaviors before trying intercourse, and to have sex less frequently (Tracy et al. 2003). In a sample of 327 adult men between the ages of 19 and 35 years, attachment avoidance was unrelated to frequency of sexual intercourse (Bogaert & Sadava 2002). Even though there is some evidence that attachment avoidance is associated with a higher frequency of casual sex (i.e., sex in the absence of a committed relationship) (Brennan & Shaver 1995; Schachner & Shaver 2002), this is only true for a subset of avoidant males. One explanation for this discrepancy is a distinction that Del Giudice relatively ignores: the difference between dismissingavoidance and fearful-avoidance. According to adult attachment theory (Bartholomew & Horowitz 1991; Hazan & Shaver 1987), both dismissing-avoidant and fearful-avoidant people are uncomfortable with closeness and intimacy and avoid relying on attachment figures. A critical distinction between the two is in their level of attachment anxiety (i.e., concerns about rejection and abandonment by partners). Dismissing individuals avoid intimacy and closeness because, at least explicitly as part of an emotion-regulation strategy (Fraley & Shaver 1997), they devalue the importance of close others and emotional ties. In contrast, fearful individuals avoid intimacy and closeness because they strongly fear being rejected. Thus, dismissingavoidance is characterized by high avoidance and low anxiety, whereas fearful-avoidance is characterized by high avoidance and high anxiety. With respect to sexual behaviors, these two avoidant patterns share some similarities but also differ in important ways. Because both are uncomfortable with closeness, sexual activity is less satisfying, less pleasurable, and even aversive (Birnbaum et al. 2006), and consequently, is engaged in less frequently (e.g., Brassard et al. 2007). However, only dismissing-avoidance is associated with engaging in casual sex in lieu of intimate relationships (Schachner & Shaver 2004). Thus, Del Giudice’s model does not apply to a significant proportion of avoidant individuals, namely, fearful-avoidants. Further, even if one distinguishes between the two avoidant patterns, as proposed by attachment theory, Del Giudice’s model may still not be correct with respect to the short-term mating strategies of dismissing-avoidants. Although they engage in casual sex, there is no clear evidence that dismissing-avoidants have an earlier onset of sexual activity (Bogaert & Sadava

Response/Del Giudice: Sex, attachment, and the development of reproductive strategies 2002), or that they engage in other short-term mating strategies, such as extra-paired sexual affairs (Bogaert & Sadava 2002; Gangestad & Thornhill 1997). Thus, there is limited evidence supporting Del Giudice’s claim that dismissing-avoidant men engage in short-term mating strategies. As an alternative explanation, we propose that dismissing men who engage in casual sex do so to satisfy attachment-related needs, irrespective of direct reproductive outcomes. Research shows that avoidant individuals are more likely to cite external motivations for engaging in sex (“to fit in better,” “to be able to say you’ve done it”), which suggests that they engage in sex to seek social status and to avoid peer rejection (Schachner & Shaver 2004). Consistent with the idea that attachment-related needs underlie engaging in sexual behaviors for avoidant individuals, greater frequency of casual sex appears to be independent of sex drive (Schachner & Shaver 2002). Moreover, we question whether a short-term strategy is more beneficial than a long-term strategy for dismissing individuals. We consider this from both a sexual strategies perspective, in which individuals aim to maximize their fitness, as well as an attachment perspective (Bowlby 1982), according to which individuals aim to maintain close ties with others to promote feelings of security and well being. From a sexual strategies perspective, pair bonding in humans increases fitness (Hazan & Diamond 2000). It promotes parental investment in offspring, thereby increasing offspring survival. Further, it increases the chance of fertilization, given the covert ovulation cycle in women and that the frequency of women’s ovulation cycles increases within pair bonds (Veith et al. 1983). As a result, sex within a committed relationship is more likely to lead to reproductive success than promiscuous sex. Moreover, although individuals who engage in casual sex may have a greater number of partners per year, the frequency of engaging in sex is likely to still be higher for dismissing individuals within committed relationships. In short, it has not been established that a short-term mating strategy will increase fitness for dismissing men. From an attachment perspective, the physical, psychological, and social benefits of mother – child bonds (Harlow & Harlow 1965) and pair bonds (Uchino et al. 1996) are well-documented. The magnitude of attachment-related benefits on health and mortality are approximately the same as the detrimental effect of smoking on health (House et al. 1988). Moreover, the benefits of pair bonding and the costs associated with lack of a mate may be greater for men than women (Cramer & Neyedley 1998). Further, the nonrandom pairing of anxious women and avoidant men (Kirkpatrick & Davis 1994) may be extremely beneficial for dismissing-avoidant men. This pairing of dismissing men, who have low levels of caregiving, with anxious women, who are prone to compulsive caregiving, would enable dismissing men to benefit from pair bonding without the cost of having to provide care (Feeney & Collins 2001). Given these considerations, it is not surprising that a substantial proportion of dismissing-avoidant individuals still commit to relationship partners (Bogaert & Sadava 2002), instead of remaining unpaired (although less so than secure individuals). In summary, we argue that Del Giudice’s model ignores the majority of avoidant men, namely, fearful-avoidant men. Moreover, given the limited evidence that dismissing-avoidant men engage in short-term mating strategies, it may not even apply to the subset of men who are dismissing-avoidant. Finally, Del Giudice ignores the benefits of human pair bonding both on, and independent of, reproductive outcomes for dismissing men. ACKNOWLEDGMENTS We gratefully acknowledge Cindy Hazan, Melissa Ferguson, Barbara Finlay, Jeanne M. Robertson, and Amanda J. Carreiro for their helpful suggestions on earlier drafts of this commentary.

Author’s Response Human reproductive strategies: An emerging synthesis? doi:10.1017/S0140525X09000272 Marco Del Giudice Center for Cognitive Science, Department of Psychology, University of Turin, 10123 Torino, Italy. [email protected] http://www.psych.unito.it/csc/pers/delgiudice/delgiudice.html

Abstract: In the first part of this response, commentators’ critiques to the target article are reviewed and addressed. The main discussion topics are the role of attachment in developmental plasticity; the relationship between attachment, mating, and reproductive strategies; the existence, magnitude, and developmental timing of sex differences in attachment; the adaptiveness of insecure styles; and the neurobiology of attachment and reproduction. The model’s assumptions are clarified, and a number of methodological issues that can confound the interpretation of research findings are examined. In the second part, various proposals made by commentators are synthesized, and directions for future research and theoretical improvement are outlined. In addition, the issue of disorganized and fearful attachment is tentatively addressed. It is argued that different theoretical perspectives are converging toward a consistent and comprehensive theory of human reproductive strategies.

Receiving so much feedback is nothing short of thrilling, and the commentators did a great job dissecting my model, criticizing its assumptions, and suggesting novel ways to improve our understanding of human life histories – my thanks to all of them. I am excited at the new opportunities for interdisciplinary integration that are clearly emerging from this lively area of research. In the target article, I presented an updated evolutionary-developmental model of human reproductive strategies, integrating the life history framework of Belsky, Steinberg, and Draper (Belsky et al. 1991) and Chisholm (1999) with sexual selection and parental investment theory. In this model, insecure attachment in infancy and early childhood entrains reproductive strategies that are based on current reproduction and high mating effort; however, the optimal strategies (and the optimal balance of mating versus parenting effort) differ between males and females. This is reflected in a reorganization of the attachment system in middle childhood, leading to sex differences in insecure attachment styles. In particular, insecure males tend to adopt avoidant styles, whereas insecure females show higher levels of anxiety. In a parental investment/sexual selection framework, sex-specific attachment styles are adaptive both in adults, where they regulate commitment and investment in couple bonding, and in children, where they affect a suite of traits involved in same-sex peer competition. The reproductive strategy adopted by an adult individual is thought to reflect an interplay between early stress and attachment security, genotypic factors, and later experiences (see Fig. 2 of the target article). Finally, I proposed that adrenarche acts as an endocrine switch at the beginning of middle childhood, affecting the sex-specific development of attachment styles and of nascent reproductive strategies. BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

45

Response/Del Giudice: Sex, attachment, and the development of reproductive strategies I begin this response by addressing major criticism, clarifying my model’s assumptions, and discussing some methodological issues that had to be omitted from the target article but were often raised in the commentaries (sects. R1–R6). Then, I gather the new suggestions and ideas presented by commentators, and attempt to draw a map of the most intriguing directions for future research (sects. R7 and R8). R1. Developmental plasticity The idea that psychosocial stress in infancy and childhood affects the trajectory of reproductive strategies is crucially based on the concept of adaptive developmental plasticity. Penke proposes a balanced overview of the place of developmental plasticity among the many mechanisms of evolutionary adaptation. I agree with his presentation in almost every detail, and I concur that the essential question concerns the relative weight of the various mechanisms. Where we disagree is about the significance and weight of attachment, as I discuss further on (sect. R3.1). More radically, Ho¨nekopp presents an argument purportedly showing that the plasticity mechanism I proposed cannot work as an adaptation. There are, however, several problems with Ho¨nekopp’s argument, concerning both the theoretical reasoning and the evidence cited in its support. First, the small Dogon sample he cites as evidence that attachment is unrelated to mortality was drawn from an atypical, urbanized population, and even so, 23% of the 26 infants were disorganized in 4-way coding (True 1994). Second, the contention that attachment security is insensitive to environmental variation (based on a single review of cross-cultural studies, many of which took place in industrialized countries) is clearly inconsistent with the extreme insecurity proportions found in at-risk and low-socioeconomic status (SES) samples (see references in sects. 2.2 and 2.3 of the target article). Third, an evolutionary perspective challenges the idea that consistently sleeping away from parents is an “irrelevant” variable for infants; in our evolutionary past, losing physical proximity with parents was likely one of the surest signs of danger – not least, because of abandonment risk (Hrdy 1999). On the theoretical side, the “information chain” presented by Ho¨nekopp is based on a partial rendering of my model. In the target article, I argued that environmental risk during infancy and early childhood, strongly mediated by parental behavior, affects reproductive strategies in middle childhood and early adulthood, with the possibility of later revision. Strategic flexibility in adolescence and adulthood (sect. 7.1.3; Fig. 2) makes it unnecessary to accurately predict the environment two decades or so in advance, as implied by Ho¨nekopp; early information channels reproductive strategies in adaptive directions, but it does not “freeze” individuals onto fixed developmental courses. In addition, I tried to stress that parental behavior does not only mediate (macro-) environmental conditions, it also determines (micro-) environmental conditions (sect. 7.1.1), thus giving the child direct and valuable information about contingent risk, expected future investment (Ellis 2004; target article, sect. 5.4.1), and alloparental availability. There are two more reasons why Ho¨nekopp’s argument, while interesting, may be overstated. First of all, the information chain he presents is redundant, since the correlations between (1) parental behavior and attachment and 46

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

(2) attachment and adult reproductive strategy are both controlled by the child, and could well amount to 1.0 if that was in the child’s interest. Thus, contrary to the argument’s assumptions, there is no inherent information loss going from parental behavior to reproductive strategies. Second, acquiring direct and reliable information about environmental risks usually involves exposing oneself to those risks; and while parents are forced to do so, infants and young children should try to avoid direct exposure to danger as much as possible (letting parents do the work for them, so to speak). I agree that children can gain information by observing the environment on their own; but parental behavior (especially when sampled over many years) can still be a useful source of information at various levels, even if we still have a limited understanding of the exact ways in which such information is transmitted and encoded. Mathematical models and simulations could help immensely to clarify this issue (see sect. R7.2). It is nevertheless true that I did not emphasize flexibility enough (see Jackson & Ellis and Flinn, Muehlenbein, & Ponzi [Flinn et al.]). Reproductive strategies in a longlived species like ours are most likely implemented by a multi-stage process with a considerable degree of protracted plasticity. The middle childhood transition (for which I propose the label of juvenile transition; Del Giudice et al., in press) is only one of the possible “switch points” in human life history, and one of the functions of juvenility may precisely be that of allowing for social feedback about the viability and success of one’s strategy (Del Giudice et al., in press). As briefly anticipated in section 7.1, life history models should now move toward a more sophisticated, truly lifespan perspective, integrating the various time frames over which reproductive decisions are made. R1.1. The role of genetic factors

Developmental plasticity and genetic factors are, in fact, two sides of the same coin (West-Eberhard 2003). Although my present focus was on plasticity, I also stressed that genotypic differences are expected to play multiple and important roles in shaping reproductive strategies. The work on the K-Factor cited by Figueredo, Sefcek, & Olderbak (Figueredo et al.) is especially interesting, although it would really take too much space to discuss it in detail. I will limit myself to note that Figueredo et al.’s (2004) twin sample was composed of middle-aged people and that heritability could be substantially lower at younger ages, as it happens with other traits (e.g., general intelligence). There are also a number of issues to be considered in order to properly interpret heritability estimates; for example, high heritability may sometimes be compatible with strong causal effects of the environment on a trait (e.g., Turkheimer 2004). Further discussion of how genetic and environmental factors are integrated in the development of reproductive strategies can be found in Del Giudice et al. (in press). Although behavior genetics can give fundamental contributions to the evolutionary understanding of reproductive strategies, I find Harris’s arguments against life history models (and attachment theory in general) to be empirically weak and ultimately unconvincing. First of all, Harris equates non-shared variance in biometric models with extra-familial factors – a problematic interpretation that has been repeatedly challenged (for a thorough discussion,

Response/Del Giudice: Sex, attachment, and the development of reproductive strategies see Rutter et al. 2001; 2006). Moreover, she ignores the consistent findings of low heritability in infants’ and children’s attachment styles (sect. 2.3). Harris’s argument is also weakened by recent, genetically controlled studies linking family stress to age at menarche (Tither & Ellis 2008) and early initiation of sexual activity (D’Onofrio et al. 2006). Importantly, both Tither and Ellis (2008) and Ellis and Essex (2007) controlled for SES in their samples, and the latter also directly controlled for body mass index (BMI); thus, Harris’s hypothesis that SES and obesity fully account for the relationship between stress and early maturation doesn’t seem to enjoy much empirical support. R2. How big are sex differences in romantic attachment? Sex differences in romantic attachment are an essential feature of my model: They are predicted to arise in many human populations, and are thought to result from adaptive processes. In contrast, both Beckes & Simpson and Penke argue that sex differences in attachment are small, perhaps too small to be of evolutionary significance. This issue deserves a thorough discussion. First of all, my model does not predict generalized, context-independent sex differences in attachment patterns; rather, sex differences should be mostly apparent in insecurely attached individuals, and they should peak in moderately risky environments. Since most individuals in most populations are securely attached, overall sex differences in avoidance and anxiety can be expected to be moderate at best. Turning to the empirical size of sex differences in romantic attachment, I concur that there are many inconsistent results in the literature. However, we are likely to severely underestimate the magnitude of sex differences if we fail to take into account four confounding factors: (1) the unreliability of some attachment measures, (2) the restricted nature of most research samples, (3) the effect of age, and (4) the impact of scores distribution on the magnitude of effect sizes. Of course, to properly address these issues one would need a systematic meta-analysis (currently in preparation); bearing this in mind, in Table R1 I summarize the effect sizes from a number of recent studies. The unreliability of romantic attachment measures is a long-known problem (e.g., Baldwin & Fehr 1995). Unfortunately, many older studies (and some new ones) employ obsolete instruments such as the single-item scales of the Relationships Questionnaire (RQ) (Bartholomew & Horowitz 1991); because of score unreliability, such measures provide downwardbiased estimates of sex differences. In addition, the common practice of reducing continuous scores to categories is guaranteed to lose some sex-related variance in the process. At the moment, the best romantic attachment questionnaires are arguably the Experiences in Close Relationships (ECR; Brennan et al. 1998) and the Experiences in Close Relationships - Revised (ECR-R; Fraley et al. 2000); the studies included in Table R1 were selected because they employed one of these measures in the continuous form and reported some measure of association between attachment and sex (note that this is not intended as an exhaustive review, but only as a representative summary of recent studies). All effect sizes were converted to Cohen’s d.1

Table R1. Effect sizes of sex differences in romantic attachment (ECR scores) in community and college samples. Positive values indicate higher scores in males. Sex differences (d) Study

Anxiety

Community samples Watson et al. (2004) Brassard et al. (2007) Birnbaum (2007) Butzer & Campbell (2008) Godbout, Lussier & Sabourin (2006) College samples Crawford et al. (2006) Schwartz et al. (2007) Noftle & Shaver (2006) Picardi et al. (2005) Gentzler & Kerns (2004)

Avoidance

2.40 2.28 2.19 2.04 2.25

.24 .14 .21 .25 .14

2.15 .17 2.02 2.15 2.33 2.04

.36 .23 .00 .12 2.11 2.19

Even a casual review of the literature suggests that sex differences tend to be stronger and more consistent in community samples compared with samples of college students, which are likely to show restricted variability in many life-history-relevant characteristics (such as early family stress, present vs. future orientation, and so on).2 Note that some of the community samples (Birnbaum 2007; Butzer & Campbell 2008; Godbout et al. 2006) included a substantial proportion of middle-aged participants; and, as I tentatively showed in section 4.3, there seems to be an age-related decrease in the magnitude of sex differences, consistent with the proposed role of sex hormones. Most of the college samples were composed of psychology students; if my hypothesis is correct and gender-typicality is associated with attachment (sect. 7.2.2), then psychology students would provide an especially poor benchmark for estimating population sex differences. Psychology is a strongly female-biased faculty, and males enrolled in psychology courses are likely to represent a restricted segment of an already restricted subpopulation. Preliminary evidence supports this possibility: In a sample I just collected, composed of 200 Italian students from a wide range of faculties (e.g., psychology, humanities, engineering, mathematics, law, and so on), sex differences in ECR scores were d ¼ 2.37 (anxiety) and d ¼ .14 (avoidance) in the psychology subsample (N ¼ 74), and d ¼ 2.52 (anxiety) and d ¼ .43 (avoidance) in the rest of the sample.3 Another factor that may lead to underestimating sex differences is the non-normal distribution of attachment scores. With skewed or otherwise non-normal data, d (the standardized difference between means) can underestimate the size of group differences. For example, in the student sample described above the overall sex difference in avoidance was d ¼ .29, which some may automatically interpret as “small”; but, due to skewed distribution in females, 74% of males showed higher avoidance scores than the median female (hardly a trivial effect). Differences in score distributions can be effectively analyzed using specific statistical methods BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

47

Response/Del Giudice: Sex, attachment, and the development of reproductive strategies (e.g., Handcock & Morris 1999); these methods should be employed much more often, especially when dealing with group differences involving non-normal variables. To conclude, I concur with Symons & Szielasko that documenting sex differences in attachment is only the first step, and that sex-specific patterns of correlations should be systematically investigated. Unfortunately, too much attachment research still ignores or downplays the existence of sex differences, so that the relevant analyses are seldom performed (Kerns); I hope that the present work contributes to a wider appreciation of this crucial issue. R2.1. Cultural variation

Although many studies (mostly carried out in Western countries) show evidence of sex differences in romantic attachment, life history models do not predict universal, context-independent sex differences (see Schmitt 2008). As discussed in section 7.1.3, the magnitude of sex differences is expected to be contingent on environmental risk; moreover, Quinlan suggests that sex-specific patterns of risk and parental bias could also have important effects. In the cross-cultural study by Schmitt et al. (2003a; 2004), East Asian samples showed small sex differences in dismissiveness and high preoccupation scores; this pattern appears to be mirrored in the Chinese data presented by Li, He, & Li (Li et al.). However, reanalysis of continuous ECR scores from their original data sets (Li, personal communication, July 7, 2008) shows that, in Chinese college students (from mixed faculties), sex differences go in the same direction as of those observed in Western samples, with higher male scores in avoidance (d ¼ .26) and somewhat higher female scores in anxiety (d ¼ 2 .12). In Li et al.’s data on white-collar workers, Chinese men report higher scores in both avoidance (d ¼ .39) and anxiety (d ¼ .24; details available from the author). Overall, it is possible that sex differences in attachment anxiety are reduced in Chinese people, although it would take a targeted meta-analysis (and more community samples) to put this hypothesis on firmer ground. Provisionally taking these data at face value, the really interesting question is why East Asian samples should show this kind of pattern, which is not accounted for by my model. Some of the explanations proposed by Li et al. (e.g., government-imposed limitations on the number of children) are dubious from an evolutionary point of view. For example, the one-child policy has been implemented only from 1979, and if male reproductive strategies have evolved over phylogenetic time, such recent limitations on reproductive rate are not expected to have a strong effect on male psychology. In addition, similar distributions of attachment scores are found in nations (e.g., Taiwan, South Korea) where there is no equivalent policy (Schmitt et al. 2003a; 2004). These authors may be right, however, when they point to the high sex ratio (more males than females) resulting from preferences for male children.4 High sex ratios (and preferences for male offspring) are found in many East Asian countries (e.g., South Korea, China, Taiwan; see Coale & Banister 1996; Schmitt 2005b). When reanalyzing the data by Schmitt (2005b; Schmitt et al. 2003a; 2004), 48

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

I found that high sex ratio in a cultural region predicted higher dismissing (partial b ¼ .29: P ¼ .048, controlling for age) and (especially) preoccupied attachment (partial b ¼ .36; P ¼ .011, controlling for age). In high-sex-ratio populations, males compete to marry the few available females, the mating system tends towards high relationship stability, and sociosexuality is restricted (Schmitt 2005b); one could hypothesize that, in this context, increased anxiety in males might be adaptive by helping maintaining an exclusive bond with females. Although I did not explicitly include the sex ratio in my model, it could be that it affects adult attachment in yet unknown ways; the mechanisms linking sex ratios to attachment and their interaction with environmental risk constitute a fascinating topic for future research. From a more speculative perspective, it is also possible that population-level differences in attachment are influenced by local differences in neurophysiological parameters and genetic background. The dopamine receptor D4 7-repeat allele, for example, has been implicated in the development of disorganized attachment (see Belsky et al. 2007a; van IJzendoorn & BakermansKranenburg 2006). Strikingly, this genetic variant, relatively common in Europe and Africa, and even more so in South America, is virtually absent in Asian populations (Chang et al. 1996). Other studies indicate that the 7-repeat allele has a recent evolutionary origin (about 50,000 years ago) and that it has been subject to intense positive selection (Ding et al. 2002; Wang et al. 2004); moreover, some authors have proposed that selection on the 7-repeat allele could have been driven by population differences in mating systems and reproductive strategies (Harpending & Cochran 2002). Given that, in my model, sex hormones have a crucial role in the development of sex differences in attachment, it is also of interest that Asian populations tend to have relatively low androgen levels (for references, see Wang et al. 2007). This effect seems to depend, at least in part, on dietary factors (e.g., Santner et al. 1998), though social factors may also play a role; for example, through a low emphasis on competition in collectivistic societies (see Mazur & Booth 1998). However, recent studies have also documented ethnic variation in a number of genes involved in the sex steroids pathway (Kardia et al. 2006). More research is needed to evaluate this intriguing possibility and to fully work out its evolutionary implications; interestingly, both the distribution of DRD4 alleles and the data on androgens could point to stronger selection for long-term, biparental care in Asian populations, which would be consistent with less sexually dimorphic attachment patterns (see Harpending & Cochran 2002). R2.2. Are there sex differences in middle childhood?

Bakermans-Kranenburg & van IJzendoorn present a meta-analysis of sex differences in middle childhood and early adolescence. I especially welcome their hands-on approach and the new data they present; however, their analysis suffers from a number of limitations that make it a poor test of my specific predictions. On the other hand, it also provides a perfect illustration of the subtle issues involved in testing the idea of a hormonally driven reorganization in attachment patterns.

Response/Del Giudice: Sex, attachment, and the development of reproductive strategies The first issue is that of age. The studies showing significant sex differences presented in section 4.2 all involved children aged 7 years or older; I reported only one study of 6-year-olds (Toth et al. 2006), but I did so in a tentative way and stressing that the effect was smaller, however suggestive. In section 4.1, it was explicitly stated that studies with children as old as six usually fail to detect sex differences. Thus, the fact that three of the studies included in Bakermans-Kranenburg & van IJzendoorn’s meta-analysis involve 6-year-old children (Toth et al. 2006; Bureau et al. 2006; Gloger-Tippelt et al. 2007) predictably reduces the overall effect size. However, age itself is not the whole story, because determining when attachment patterns become sex-biased requires careful consideration of the hormonal mechanism of adrenarche. In my model, I hypothesized that adrenarche drives the re-organization of attachment patterns. Indeed, adrenarche is the biological marker of juvenility (Bogin 1999; Locke & Bogin 2006): this implies that, just as with puberty, the juvenile transition should not be tied to a specific age, but rather, be considered as a dynamic and variable process (Del Giudice et al., in press). Adrenarche starts around 6 years of age for early-maturing children and, at age 7 years, about 50% of children have actually reached adrenarcheal state (Ellis & Essex 2007).5 To further complicate things, the timing of adrenarche is likely to be geographically and ethnically variable, just as that of puberty. In particular, sexual maturation is slower in northern Europe compared with southern Europe and the United States: while the mean age of menarche is 12.0 years in Italy and 12.5 years in the United States, it is 13.2 years in the Netherlands and 13.5 years in Germany, about one year later (Parent et al. 2003). Assuming (as is reasonable in absence of targeted studies) that adrenarche shows a similar time lag, sex biases in attachment patterns may not be noticeable in Dutch and German samples until about age 8 years or so. Since three of the samples in Bakermans-Kranenburg & van IJzendoorn’s meta-analysis are from the Netherlands (Gilissen et al., in press: 7 year-olds; Pannebakker 2007: 7 year-olds) and Germany (Gloger-Tippelt & Konig 2007: 6 year-olds), it is quite possible that most of these children were in fact pre-adrenarcheal. In summary, any metaanalysis targeted at properly testing the developmental hypothesis proposed in the target article should explicitly take into account three key factors: (1) children’s age, (2) geographic variation in maturational timing, and, when appropriate, (3) local variation in environmental risk (sect. 7.1.3). The last issue is that of differences between measurement instruments. As acknowledged in section 4.3, sex differences in adults are only apparent with romantic attachment measures, and (as also shown by Bakermans-Kranenburg & van IJzendoorn) do not emerge with state-of-mind interviews like the Adult Attachment Interview (AAI). Explaining why this is so is a fascinating puzzle, and I hope that future research will provide a fully satisfying answer. However, many attachment researchers agree that interviews and questionnaires measure distinct attachment-related constructs (e.g., Bartholomew & Shaver 1998; Bernier & Dozier 2002; Carnelly & Brennan 2002); they also tend to predict different outcomes, both in couple relationships (e.g., Roisman et al. 2007) and in psychopathology (e.g., Fortuna & Roisman 2008). I want to stress once more that AAI-like

interviews focus on past attachment relationships and are mostly rated according to narrative qualities (e.g., coherence); that is, they assess the way people talk about their past experiences. The study by Ammaniti et al. (2000) included in Bakermans-Kranenburg & van IJzendoorn’s meta-analysis used the Attachment Interview for Childhood and Adolescence (AICA), which is based on the same principles and asks questions such as: “Tell me about your relationship with your parents as a little child.” It is quite possible that the way children and adults talk about past relationships does not show sex differences, whereas their attitude towards current attachment figures does. Unfortunately, since Bakermans-Kranenburg & van IJzendoorn take the AAI to be the “gold standard” of adult attachment measures, they choose to ignore the data from questionnaire studies (in children as well as in adults). They also speculate that sex differences in doll-play tasks may be an artifact caused by differences in verbal abilities; however, it is difficult to see how this would account for the observed sex differences in forced-choice questionnaires such as the Coping Strategies Questionnaire (CSQ) (sect. 4.2). R3. Attachment and reproductive strategies In the target article, attachment patterns are presented as an integral component of adaptive, sex-specific reproductive strategies. This crucial relationship is the focus of many commentaries and of some critical appraisals. Penke and Petters & Waters question the general relationship between attachment and fitness-relevant traits and behaviors; Zayas & Ram, Campbell, and Chen & Li raise doubts on the adaptive nature of specific attachment patterns. I am not surprised that the link between attachment and reproductive strategies has turned out to be the most controversial part of my article. Attachment theorists have largely drifted away from Bowlby’s original emphasis on evolutionary biology, thus making it difficult to reintegrate many attachmentrelated concepts and models with modern evolutionary theory. For the same reason, empirical research on attachment often focuses on variables that lack clear biological relevance, or (even more often) concentrates on psychological well-being while neglecting potentially fitness-relevant behaviors. This state of affairs demands special care in properly evaluating the relevant empirical findings. Although I argue that the extant data support the essential lines of my model, there is clearly need for more (and more biologically meaningful) evidence to settle some of the most burning empirical issues. R3.1. The relationship between attachment, sociosexuality, and behavior

This critical issue easily lends itself to misunderstandings and confusions. Attachment has two separate roles in my model: In infants and children, it “encodes” environmental risk and parental investment, thus contributing to direct the development of reproductive strategies; in older children and adults, it is one of the traits that implement a given reproductive strategy at the behavioral level. In particular, attachment styles contribute to some aspects of behavior (e.g., aggression, self-esteem) in juveniles, and BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

49

Response/Del Giudice: Sex, attachment, and the development of reproductive strategies mediate the regulation of intimate couple relationships in adults (e.g., emotional commitment, investment, and requests for investment; sect. 6.3.1). Two important implications follow: (1) attachment styles are not the “cause” of all the behavioral manifestations of life history strategies; (2) the specific effects of attachment in adults primarily concern the formation, maintenance, and regulation of long-term relationships, and only secondarily the regulation of short-term sexual relationships (unfortunately, the distinction was somewhat blurred in sect. 5.2). Although the standard approach to sociosexuality sees long- and short-term relationships on a bipolar continuum (see Simpson et al. 2004), things are probably much more complex. As Jackson and Kirkpatrick (2007) showed, mating strategies (already a subset of reproductive strategies) can be mapped on at least two distinct dimensions, with possible overlap and tactical admixture between the pursuit of long-term relationships and short-term sexual affairs (see also sect. 6.4.1). For example, “sexually unrestricted” females are nevertheless highly sensitive to partners’ willingness to invest, tend to feel emotionally vulnerable after intercourse, and feel distressed if their partner fails to show involvement (Townsend 1995; 2005). Until recently, most researchers have focused exclusively on short-term mating, usually measured with the Sociosexual Orientation Inventory (SOI; Simpson & Gangestad 1991). Based on the evidence reviewed in my article, however, attachment can be expected to prove much more predictive of attitudes, desires, and behaviors related to intimate, committed relationships (see Jackson & Kirkpatrick 2007 for promising results). There is absolute need of instruments assessing “long-term reproductive behaviors,” that is, actions that promote couple stability and channel investment within the monogamous couple (a good example is the Partner-Specific Investment Inventory by Ellis [1998]). Natural selection is not only a matter of individual fertility: offspring quantity and quality both contribute to long-term fitness (Ackerman & Kenrick; Houston & McNamara 1999), and tracking the dynamics of parental investment is just as important as tracking the number of one’s mates. That said, I still think that Penke’s contention that “there seems to be surprisingly little evidence that romantic attachment styles actually relate to reproductive strategy-related consequential behavioral outcomes” is overstated, even with respect to short-term mating. First, he ignores the available evidence on casual sex, age of intercourse, infidelity, and sexual coercion (see sect. 5.2 in the target article and earlier here). Then, he challenges the significance of attachment on the grounds that attachment styles only predict “sociosexual attitudes,” which do not uniquely correlate with (short-term) behavioral outcomes. However, in their own research, Penke and Asendorpf (in press) found that attachment also correlated with “sociosexual desire,” which in turn did predict a number of interesting reproduction-related behaviors. Considering that they employed the single-item RQ, Penke’s dismissal of attachment may be premature. Finally, when evaluating the size of correlations one must be wary of some possible statistical artifacts. As discussed earlier, attachment scores may present with skewed distributions; the same is true of sociosexual measures (e.g., Jackson & Kirkpatrick 2007) and of reports of relatively infrequent behaviors (e.g., 50

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

infidelity, sexual coercion). Pearson’s correlation assumes normally distributed variables, and correlation coefficients can be deflated when one or both variables are skewed. If researchers do not take the appropriate statistical steps (e.g., transforming the data or computing rank correlations), they can end up underestimating the size of associations. In a similar vein, Petters & Waters argue that correlations between attachment styles and behavioral traits like aggression or withdrawal are “very small or inflated by considerable method variance.” It is true that correlations are sometimes small (e.g., consistently around .2 in Finnegan et al. 1996), but this is not always the case; for example, Granot and Mayseless (2001) reported a threefold increase in externalizing symptom scores between the secure and avoidant attachment groups (together with other effects of similar magnitude). In both of these studies, shared method variance was not an issue, as attachment was assessed by self-report or doll-play and the other measures came from teacher and peer reports. Skewed distributions (see earlier) may also contribute to deflating effect sizes in some of the studies. At a more fundamental level, I propose that employing more ecologically relevant variables and categories (e.g., dominance-seeking and relational aggression instead of “externalizing behaviors”) is likely to yield stronger and more meaningful correlations. To conclude this section, I want to address Petters & Waters’s suggestion that attachment may not be crucial to my model, and that I should focus on the direct relationship between early stress and later reproductive behavior. In short, I don’t think that such a direct relationship exists: although my focus is on reproductive strategies, attachment relationships are the primary source, filter, and buffer of stress for the child (e.g., Flinn 2006). Thus, it is difficult even to conceptualize early stress without considering attachment. At the same time, attachment is crucially involved in human pairbonding, so that a full understanding of the dynamics of parental investment cannot sidestep adult attachment styles. Any theory of human reproductive strategies must deal with attachment, one way or the other. R3.2. Are insecure patterns adaptive?

While accepting the general lines of the target article’s evolutionary framework, some commentators disagree about the specific adaptive function of insecure attachment patterns. Zayas & Ram argue that neither avoidance nor short-term mating are adaptive for males; Campbell doubts that attachment anxiety can help females in eliciting male investment, and Chen & Li believe that avoidance should be more adaptive for females than anxiety. Zayas & Ram take issue with the view that male avoidant attachment is part of an adaptive reproductive strategy, and cite a number of findings concerning the lack of correlations between avoidance and frequency of intercourse, age at first intercourse, and extra-pair affairs. The first study they cite is Tracy et al. (2003); this study, however, was based on the same data set of Cooper et al. (1998), which was already criticized in section 5.2 for employing an unreliable measure of attachment, likely to misclassify as dismissing many fearful subjects, who indeed show a peculiar pattern of sexual behavior

Response/Del Giudice: Sex, attachment, and the development of reproductive strategies (see sect. R8). Second, they cite Bogaert and Sadava (2002), who found no relationship between avoidant attachment and infidelity; however, in another study, Allen and Beaucom (2004) reported that dismissing males had had the highest number of extra-dyadic partners. Avoidance also predicts reduced sexual fantasies about the current partner (Brassard et al. 2007) and greater romantic attraction to potential alternative partners (Overall & Sibley 2008). Taken as a whole, the evidence is not as negative as implied by Zayas & Ram; of course, the relevant studies are still a handful, and more data are needed to firmly settle this issue. Third, I agree that avoidance predicts lower frequency of intercourse, but intercourse frequency has little to do with short-term mating; if anything, one should expect a negative correlation between partner variety and intercourse frequency, as already discussed by Brody and Breiterstein (2000) and Simpson et al. (2004). Looking for new partners and getting to have sex with them takes time, with a resulting trade-off between variety and frequency; and if frequent intercourse with a partner can strengthen a couple’s relationship and increase intimacy (e.g., Costa & Brody 2007; Mellen 1981), people engaging in low-commitment strategies should have sex with their partners less often in order to maintain a low level of intimacy. Reduced intercourse frequency may, ironically, reflect an adaptive aspect of lowinvestment strategies. Another argument used by these authors to challenge the adaptiveness of casual sex is that frequent intercourse within a stable pair-bond increases the likelihood of fertilization and, therefore, reproductive success. This fails to take into account the evidence that women tend to engage in extra-pair sex precisely in their phase of maximum fertility (see Volpe & Barton). Concerning their argument that pair-bonding increases reproductive success because it promotes parental investment in offspring, I wholeheartedly agree: but my model predicts that males will shift to avoidant strategies precisely when there are cues of risk, that is, when parental investment is less likely to benefit offspring (sect. 6.3.1). While Zayas & Ram may be mixing proximate and ultimate causation when they refer to the “psychological and social benefits” of pair-bonding, they make an interesting point when they cite evidence of health-related benefits, which could have nontrivial fitness effects. Also in this case, however, their logic ultimately turns out to support my model: for if intimate pair-bonds have beneficial health effects, the latter are likely to be reaped in the long term and only when the environment is not too risky. In risky environments, the optimal behavior is to trade long-term fitness benefits (including the healthpromoting effects of close relationships) for current reproduction, which is precisely what avoidant males are doing in the model. Finally, Zayas & Ram are correct in writing that I neglected fearful attachment in my article. In many respects, fearfulness (high avoidance plus high anxiety) is a puzzle similar to that of disorganized attachment, and the two may actually be related (Simpson & Rholes 2002). Further on, I address this issue in a preliminary way (sect. R8). The hypothesis that anxious attachment in females is part of an adaptive, investment-eliciting strategy is challenged by Campbell, mostly on the basis of research

showing correlations between attachment anxiety and violence toward partners. Unfortunately, nearly all the evidence she cites is from studies of borderline personality disorder (BPD). But attachment anxiety is not, as Campbell states, “a short step [. . .] to the clinical condition of borderline personality disorder”; the latter is a serious condition with an incidence of about 2% (see Diagnostic and Statistical Manual of Mental Disorders, 4th edition, Text Revision, American Psychiatric Association 2000), whereas anxious attachment is a non-clinical trait with high frequency in normal samples. In addition, BPD appears to be strongly related to fearful attachment (see sect. R8), rather than just anxiety (Brennan & Shaver 1998; Goldenson et al. 2007). The only non-clinical study of violence cited by Campbell is that by Orcutt et al. (2005), who reported an association between anxious attachment and female violence against partners. However, a closer look at their results reveals that the great majority of reported violent acts was composed of minor assaults such as “grabbed partner,” “slapped partner,” and “threw something at partner that could hurt him/her,” and engaging even once in any of these behaviors had participants classified as “violence perpetrators.” Despite these behaviors’ unpleasant nature, they don’t seem likely to end a romantic relationship. Indeed, they could help in maintaining a relationship if triggered by the feeling of being neglected by one’s partner; and, if triggered by jealousy, in dissuading him from engaging in extra-pair activities (i.e., securing exclusive investment). Anxious people require high levels of emotional support by their partners, over-attribute rejection during conflicts, and tend to escalate conflictual episodes (Campbell et al. 2005); in this context, aggression and escalation may be directed precisely at preventing rejection. Thus, although I don’t think that Campbell’s critique undermines the logic of my model, I thank her for pointing to an omission in my account: namely, the role of anger and aggression in the relational style of attachment-anxious people. Ambivalent infants and children often alternate submission and dependency with bouts of anger toward their attachment figures, and the expression of anger is considered as an integral component of their attachment pattern (Ainswort et al. 1978; Cassidy & Berlin 1994); data like those by Orcutt et al. (2005) suggest that anger may have a similar role in adult relationships as well, though more focused research is clearly needed. Campbell then points out that, if ambivalent girls are preparing to compete for access to male investment, they could be predicted to show higher “direct or indirect” aggression toward peers. This is a well-taken point; and although extant evidence indicates that ambivalent girls engage in less physical/direct aggression with peers (Corby 2006; Finnegan et al. 1996; Granot & Mayseless 2001), I am aware of no specific data on relational aggression. Because commonly employed measures of externalizing symptoms are heavily biased toward direct aggression, it may well be that researchers have been missing a piece of the puzzle, and that ambivalent girls do engage in more relational aggression with their friends. The higher level of anxious/depressive symptoms they experience could partly follow from the resulting stress in peer relationships. Distinguishing between direct and relational aggression may also be the key to answering the argument put BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

51

Response/Del Giudice: Sex, attachment, and the development of reproductive strategies forward by Chen & Li, who reason that avoidance should benefit girls and boys alike because (1) in high-risk conditions aggression can be adaptive, and (2) avoidance, but not anxiety, predicts aggression in children. Indeed, all the studies they cite measured direct and/or physical aggression, which is less typical of girls and probably less adaptive in female peer groups under most conditions. I also think that Chen & Li’s position is actually more similar to mine than they acknowledge: They state that “the behavioral correlates of avoidant attachment [. . .] are more adaptive than traits associated with ambivalent attachment [. . .] under extreme high-risk environmental conditions.” But that is fully consistent with my model: I also predict that, under extreme risk, females will shift toward avoidance, while ambivalence should be preferred in conditions of moderate risk (sects. 6.3.1 and 7.1.3). In any event, there are differences between boys and girls in the key resources they need and in the ways they acquire them (Jackson & Ellis); we need further research to understand the possible roles of ambivalence in the context of girls’ developmental trajectories. While Goetz, Perilloux, & Buss (Goetz et al.) make the plausible point that the primary evolutionary targets of ambivalence in middle childhood are parents (see also sect. 6.3.1), their statement that girls’ social networks are too volatile to last into adulthood may not have been true in ancestral environments (or even in smallscale traditional societies, where choice of friends is severely limited by the small number of same-aged peers). Campbell also questions the adaptive nature of avoidant attachment in females by suggesting that avoidance may be a forced option in absence of men able or willing to invest, and may not involve devaluing stable relationships. Research suggests that avoidant females, at least in college samples, do tend to devalue stable relationships and to deny that they are interested in them (Jackson & Kirkpatrick 2007). The gang girls described by Campbell sound as if they are fearfully attached, rather than just avoidant; nevertheless, reproductive strategies can change following environmental feedback (sect. 7.1.3), and the commentary by Kang & Glassman anecdotally suggests that some homeless girls may shift from “hopeful” anxiety to avoidance when they repeatedly fail to establish a viable relationship. Whether this represents temporary reproductive suppression or an escalation in perceived environmental risk is a question for future systematic research. In the end, the observations by Campbell and Kang & Glassmann reinforce my hypothesis that female avoidance is a heterogeneous construct, and that it may reflect different kinds of underlying strategies (sect. 6.3.1); I think we are still a long way from fully understanding the development of avoidance in women. R4. Attachment, reproduction, and neurobiology In section 6.3.1, I speculated that women may sometimes experience reproductive suppression, and that this might correlate with shifts in attachment styles (e.g. temporary increases in avoidance). Seltzer & Pollack argue that I made a careless comparison with primates, among which suppression is induced in helpers by the presence of reproducing dominants; they then contend that humans lack specialized physiological mechanism capable of suppressing fertility. Wasser and Barash (1983) were 52

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

the first to propose that human females may undergo reproductive suppression following lack of social support and/or aggression by other females. Leaving aside the purely terminological issues, I don’t think their hypothesis ignores the complexities of comparative biology; in a way, Seltzer & Pollack may themselves be drawing too quick a parallel between cooperative breeding in humans and in other primates. Remarkably, in the human breeding system, helpers are frequently older and more dominant individuals (e.g., grandparents, aunts); thus, the common pattern of dominant-induced reproductive suppression may rightly fail to apply to humans, and a differently tuned mechanism may be present. Moreover, Seltzer & Pollack’s contention that humans lack the necessary fertility-reducing mechanisms is not warranted: there is evidence that social stress can affect reproduction in human females by reducing ovarian function, delaying ovulation, inducing spontaneous abortions and premature births (and, not least, by increasing the likelihood of abandonment or infanticide). For references, see Wasser and Barash (1983), Wasser (1994), Wasser and Place (2001), Hrdy (1999), Arck et al. (2001), and Sanders and Bruce (1997); see Ellison (2001) and Vitzthum (2008) for a discussion of “graded” fertility regulation in women. I am also more optimistic than Seltzer & Pollack about the possibility of linking adult attachment to a relatively clear neurobiological basis. While human romantic attachment is in many respects a unique trait, it is not “seemingly unrelated to the physiological mechanisms of social behavior studied in other species.” For example, romantic attachment anxiety has been recently linked to increased oxytocin (OT) levels (Marazziti et al. 2006) and heightened cortisol reactivity (Quirin et al. 2008) – two variables with clear relevance to comparative models of pair-bonding. Speaking of oxytocin, I appreciate the suggestion by Chen & Li that this hormone may be crucially involved in the development of female (and, I think, male) attachment styles. I also agree that avoidant attachment is probably linked to low OT levels, and believe that the regulation of OT by sex hormones may be one of the causal mechanisms through which adrenarche affects attachment behavior (see also Campbell 2008a; Goldstein Ferber’s commentary). They probably go too far, however, when they argue that OT has little or no effect on the adult brain: As recent experimental studies demonstrate, OT administration in adult humans can have remarkable effects on social perception and behavior (e.g., Domes et al. 2007; Guastella et al. 2008; Kosfeld et al. 2005; Zak et al. 2005). In addition, it is often difficult to know for sure whether or not OT receptors are expressed in a given brain region (see Gimpl & Fahrenholz 2001). The physiology of stress is also at the heart of Goldstein Ferber’s commentary, where the proposal is made that prenatal mother–fetus interactions may prime the subsequent development of attachment and sex differences. I basically agree with Goldstein Ferber on this point (see Fig. 2), although her evolutionary reasoning appears somewhat obscure when she discusses costly signaling and the handicap principle. Finally, Petters & Waters remark that, while physiological factors can partly mediate the effects of the environment on attachment, adaptive processes can be mediated by cognitive processes as well. I totally agree with them, but different processes require different types of mechanisms: and when it comes to

Response/Del Giudice: Sex, attachment, and the development of reproductive strategies coordinating sexual maturation, reproduction, motivation, and social behavior (as life history strategies imply), there really is no alternative to hormones and their powerful integrative function (Flinn et al.). See Del Giudice et al. (in press) for an extended discussion of this topic. R5. Sex differences in mating versus parenting effort I now address a problematic aspect of the target article that was not singled out by commentators. In order to derive sex-specific predictions about life histories, I relied on Trivers’ Parental Investment Theory (Trivers 1972). I was unaware that, a few years before, Trivers’ famous model had been shown to embed a number of fallacies, one of which I imported straight into my own article. In short, the assumption that males enjoy (on average) a higher fitness gain than females for a given investment in mating effort (sect. 6.3.1) is inaccurate, because, with an even sex ratio, the average reproductive success of males and that of females have to be equal by arithmetic necessity (see Kokko & Jennions 2003; 2008). This fallacy is not fatal to the model, but it requires further specification and a more complex set of assumptions. For example, investing in mating instead of parenting can be an optimal strategy for males, provided that they can reliably assess their own mating potential and adjust their strategy accordingly (Kokko & Jennions 2008; Kokko, personal communication, April 9, 2008). What are the consequences for my theory? First, this is one more reason to call for a multi-stage model with repeated sequences of strategic decision-assessmentadjustment (sect. R1). My hypothesis is that juvenility and adolescence (Symons & Szielasko; Jackson & Ellis) are especially useful as self-assessment phases (Del Giudice et al., in press). Second, the importance of self-assessment broadens the theoretical rationale for including phenotypic quality and social feedback in lifehistory models (Jackson & Ellis). Third, for some males (those of uncertain mating potential) who adopt avoidant strategies, the optimal tactic would not be that of eschewing parenting effort altogether; rather, these males should be more willing to cut on parenting effort and invest in mating when mating opportunities arise. The evidence that avoidant people are less involved in current relationships and more easily “tempted” by potential alternative partners (e.g., Brassard et al. 2007; Overall & Sibley 2008) is intriguing when viewed in this light. Of course, these considerations apply all else being equal; for example, unbalanced sex ratios can considerably alter the costs and benefits of different strategies, and so on. R6. Miscellaneous topics A number of specific comments and critiques were made by Figueredo et al.; I address them in turn here. With respect to the assessment of attachment, I also side with the continuous/dimensional measurement camp. I am not sure, however, that two dimensions are enough: in particular, fearful attachment may possess some qualitatively different properties, and they may be better captured by a specific scale (see sect. R8). As for the misstated sentence

in section 3.1, its intended meaning is that the fitness contribution of a given trait must be weighted – costs and benefits – over the whole lifetime of an organism. Figueredo et al. are also puzzled by my discussion of age-related shifts toward paternal investment and by my (schematic) analysis of polygyny. Concerning the former, I see no particular contradiction: as discussed above (see sect. R3.1), short-term mating is only a single facet of reproductive strategies, and (depending on various social and personal factors) it can coexist with variable degrees of investment in long-term relationships and in offspring care. With respect to polygyny, the literature on human mating systems is complex and, alas, sometimes contradictory; however, in contrast to what Figueredo et al. argue, there is evidence that in polygynous systems, paternal involvement is reduced (Quinlan 2007) and males contribute in smaller proportions to family subsistence (Marlowe 2000; 2003). R7. Emerging perspectives Many commentators proposed ways to advance, extend, and specify the model presented in the target article; I found their comments to be exceptionally inspiring and stimulating. Such a high degree of convergence does not happen by chance: It seems to me that a new, comprehensive theoretical synthesis on human reproductive strategies is just within reach (Ackerman & Kenrick; Jackson & Ellis; Beckes & Simpson; Goetz et al.). Here I shall try to show how a number of empirical and theoretical threads are coming together, and how this could help in shaping future research in the field. R7.1. Theoretical integration

Reproductive strategies are not just another facet of human behavior; they lie at the very heart of developmental processes and have implications for a wide range of psychological phenomena, including some that may at first sight seem unrelated (Ackerman & Kenrick). It is not surprising, then, that understanding them requires multiple levels of analysis and a confluence of different evolutionary approaches. In the target article, I tried to work on the life-history side of the topic, and I am glad that commentators responded by making explicit some of the links I hinted at in section 6.4. I find Jackson & Ellis’s complementary approach exciting, and I believe that their focus on social competition and phenotypic quality hits the target. Indeed, I reached a parallel conclusion when adjusting my model to avoid unrealistic assumptions about reproductive success (see sect. R5); this strongly suggests that the theory can be made fully consistent once we reach a sufficient level of detail. My view (in agreement with Jackson & Ellis, Flinn et al., and Goetz et al.) is that we need a multi-stage theory, with multiple decision-assessment-adjustment phases focused on different cues at different points in development. Of course, more research (theoretical as well as empirical) is needed to gain a fuller appreciation of which cues are important and how they are conveyed – a point beautifully made by Beckes & Simpson. I agree with them that different risk factors call for different strategies, and differentiating pathogen stress from interpersonal BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

53

Response/Del Giudice: Sex, attachment, and the development of reproductive strategies conflict is a good place to start; I have only a few minor reservations on their present account. First, a non-obvious possibility is that the levels of attachment-related stress already contain some implicit information on the likely sources of risk, especially when both parents are considered (sect. 6.2.1). Second and related, the specific consequences of pathogen load may depend on its severity. In a lifehistory perspective, the crucial question is whether parental effort is repaid by offspring fitness or not; at moderate levels of pathogen stress, increasing parental care may work better than withholding it (see Quinlan 2007). Third, different sources of risk are likely to covary, as suggested for example by the association of polygyny with pathogen stress (e.g., Quinlan 2007), thus making tests of independent effects more difficult. Beckes & Simpson and Volpe & Barton point to another issue neglected in the target article – that of mate preferences and mate choice. This is a bridge that needs to be crossed in order to integrate life-history and strategic-pluralism models. Fortunately, there is some promising work in this direction, investigating how attachment styles in women relates to mate preferences (Cohen & Belsky 2008; Kruger 2008). Also in this case, actual behaviors should be assessed in addition to self-reported preferences, but I suspect that the data will ultimately reveal consistency rather than conflict between different theoretical perspectives. A crucial issue in testing hypotheses about mate preferences will be careful consideration of how reproductive strategies relate to mating styles (see sect. R3.1); for example, it is important not to conflate short-term mating (e.g., unrestrictedness at the SOI) with low-investment, present-oriented life histories. Even with this caveat in mind, Volpe & Barton’s fascinating suggestion that some components of romantic attachment may fluctuate in parallel with mate preferences during the menstrual cycle is definitely worth investigating.

R7.2. A call for interdisciplinarity

A common theme in many commentaries is the need for a multidisciplinary approach to the development of reproductive strategies (e.g., Kruger). One of the beauties of evolutionary theory is that it naturally leads to interdisciplinarity, as masterfully illustrated by Wilson (1998). In addition, a serious interdisciplinary perspective provides abundant constraints on hypothesis-building, thus protecting against just-so-storytelling. Despite what some critics of evolutionary psychology believe, figuring out an adaptive explanation that is simultaneously consistent with the available psychological, anthropological, and neurobiological data (in addition to general evolutionary theory) is anything but easy. Of course, theory construction also becomes exponentially harder, and some commentators have rightly pointed at some underdeveloped aspects of my account. I agree with Maestripieri, Seltzer & Pollack, and Kruger on the need for more comparative evidence from other primates, although our reproductive system is probably unique in various respects. Maestripieri provides an excellent rationale for the use of primate experiments in investigating attachment, and his data on girls’ interest in infants suggest how to assess life-history hypotheses going beyond the usual focus on mating styles and menarche timing. 54

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

The importance of anthropological and cross-cultural data cannot be understated, as I hope to have shown in the target article. Quinlan’s brilliant analysis of sexspecific risk and investment patterns illustrates the complex interplay among ecological factors that anthropologists are familiar with, and suggests some directions to formulate new, testable cross-cultural predictions (see also Lewis & Tooley for another application of the TriversWillard hypothesis). Then, when moving from the functional to the mechanistic level of analysis, it is extremely useful to investigate the neurobiological substrates of behavior (Maestripieri; Goldstein Ferber; Seltzer & Pollak; Chen & Li; Flinn et al.). True integration of human (and nonhuman) neurobiology with modern evolutionary biology still has a long way to go, although there are signs of accelerating progress. Clearly, both sides are going to greatly benefit from increased interdisciplinarity. I don’t want to convey the impression that I am neglecting the contribution of developmental psychology and, in particular, of attachment research. Despite occasional criticism, I believe that attachment researchers are in an especially favorable position to leave behind psychology’s biophobia and contribute to the cutting edge of interdisciplinary research. In this context, Kerns’s commentary provides a much needed developmental perspective on attachment dynamics in middle childhood. It is true that the target article concentrated on the functional and neurobiological levels at the expense of the psychological one (see also Petters & Waters). But what happens to parenting and parent –child relationships with the transition to middle childhood? And what happens to “internal working models”? Do peers have a proximate role in shaping children’s attachment patterns? I am grateful to Kerns for asking these questions, and hope that research will soon provide some answers. The study of attachment development in middle childhood is still at the beginning (but see Kerns & Richardson 2005); hopefully, increased attention to evolutionary questions and models will help researchers to frame their studies in a biologically relevant manner, thus contributing to testing and refining the most interesting evolutionary hypotheses on development. Finally, I am very positive about the modeling approach proposed by Petters & Waters. If anything is lacking from current evolutionary psychology, it is formal mathematical modeling in addition to verbal theorizing. Here, the lag behind evolutionary biology at large is apparent; this forces researchers to rely on general middle-level theories (with loss of specificity), or on models originally developed for other species (with the risk of making unrealistic assumptions). My feeling is that our verbal models (mine included) are already too complex to be fully worked out without the benefit of a formal approach; and if we start adding the many layers of complexity required for realistic and detailed prediction (Quinlan; Beckes & Simpson; Goetz et al.), things are going to quickly become unmanageable. In this context, simulations can provide a useful tool to assess the consistency of theories, discover their most robust predictions, and possibly simplify them in important ways. Most likely, they will also help in understanding what aspects of the environment are really important, and how they can be effectively encoded by the developing child and his/her family (see sect. R1).

Response/Del Giudice: Sex, attachment, and the development of reproductive strategies R8. The puzzle of disorganized and fearful attachment One of the main limitations of the target article is that it leaves out of the picture both disorganization and adult fearful attachment (i.e., high levels of both anxiety and avoidance), and some commentators correctly noted this omission (Figueredo et al., Zayas & Ram, Kerns). The first crucial question about disorganization is whether or not it is adaptive. Apparently, the current consensus lies on the “no” side; and Lewis & Tooley present a reasoned argument to the effect that disorganization falls outside the expected caregiving range of humans, and may therefore lead to maladaptive and pathological behavioral outcomes. On one hand, it is clearly possible that some forms of disorganization may be non-adaptive results of a disruption of the attachment system. On the other hand, there are at least two reasons to look for possible adaptive explanations. First, the resemblance (also noted by Lewis & Tooley) of some disorganized behaviors with mammalian defense reactions is probably not a coincidence; and second, the implicit assumption that abuse and severe neglect were more or less unknown in ancestral human families is probably incorrect. For example, Hrdy (1999) provides a sobering evolutionary account of abandonment and filicide throughout human evolution and across cultures; furthermore, step-fathering has probably been frequent in our phylogenetic history (see Miller 2000), and this should have led to recurrent risk of violence and abuse (e.g., Daly & Wilson 1996). For all these reasons (and without any ethical implication), I object to the widespread idea that severe neglect or abuse are outside our species-typical parenting patterns (e.g., Scarr 1992). If the above is correct, it may be that disorganization represents (at least in some cases) an adaptive response to grave and persisting danger, and that its developmental correlates are tuned to the task of surviving through extreme high-risk conditions. The fact that attachment disorganization seems to represent a transitory stage for most children (e.g., Main 2005; Moss et al. 2005) could be regarded as consistent with this position. What about fearful attachment in adults? The least that can be said is that fearfulness is not a well-understood category. I am sympathetic to Simpson and Rholes’s (2002) argument that adult fearfulness may be functionally similar to infant disorganization. However, this would imply that present two-dimensional models of attachment (which do not assess anything resembling disorganization) are probably inadequate to fully capture the functional meaning of fearfulness. Indeed, it can be argued (as Feeney [2002] has done) that we need more than two dimensions to adequately describe adult attachment styles; and evolutionary theory may actually provide the best guidelines for generating new items and scales. At present, it is likely that romantic attachment questionnaires lump together phenotypically similar but functionally different patterns: for example, if my speculative hypothesis were correct, avoidance in women (with or without high anxiety levels) could sometimes represent a behavioral correlate of reproductive suppression (sect. 6.3.1). Another pertinent example is provided by developmental shifts in attachment styles (Kang & Glassman); what happens to anxious girls when they shift toward avoidance following repeated failures to establish a

couple relationship? Do they become dismissing (i.e., their anxiety drops) or fearful (i.e., they maintain high anxiety coupled with increased avoidance)? For some individuals, fearfulness may represent a transitory stage, whereas for others it may be a stable condition. It quickly becomes apparent that fearfulness (as presently defined) could be a mixed category, including persons with very different histories and reproductive strategies. This may also account for the bizarre pattern of sexual behavior associated with fearfulness: for example, a close look at the results by Gentzler and Kerns (2004) shows that high avoidance and high anxiety are found in the group of students reporting early intercourse, but also in that of students reporting no intercourse at all. In conclusion, it seems to me that in order to reach a satisfactory understanding of adult romantic attachment, we need (1) more longitudinal studies, both in the long and in the short term, and (2) more sophisticated measurement models that tap on biologically significant attachment-related constructs. R9. Conclusion I am glad that the target article has stimulated discussion of so many important themes. My top list of issues worthy of further investigation includes the developmental course of sex differences in attachment; the reasons for cross-cultural variation; the nature of the information transmitted from parents to children; the role and weight of genotypic factors; the flexibility of individual reproductive strategies; the differences between alternative measures of attachment and their developmental correlates; and the meaning of disorganized and fearful attachment. I will be proud if this article contributes even a little to (re)integrating attachment theory with evolutionary psychology; but, as the commentaries make evident, the study of human reproductive strategies already transcends parochial academic boundaries and involves a wide array of interdisciplinary competences. The main message I gather from this exchange is that, despite the multitude of perspectives that bear on this topic, there are striking opportunities for synthesis and complementarity. If the future delivers what the present promises, the study of human development is in for a really exciting time. NOTES 1. An advantage of converting from r to d is that imbalances in the frequency of males versus females in a sample (extremely common in the attachment literature) lead to smaller point-biserial correlations, thus encouraging downward-biased estimation of sex differences. In contrast, d is unbiased by the relative frequencies of the two sexes in the sample (see McGrath & Meyer 2006). 2. In Crawford et al. (2006), 90 participants were males and 217 were females (Crawford, personal communication, July 10, 2008). These frequencies are needed to calculate d from r. 3. It should also be noted that effect sizes such as Cohen’s d must be interpreted case-by-case, depending on the theoretical meaning of the effect, the scale’s reliability, and the amount of measurement error. Relying on “canned” effect sizes (e.g., d ¼ .8 is “large,” and so on) is bad statistical practice and was strongly discouraged (alas, to little effect) by Cohen himself (Cohen 1988; see also Breaugh 2003). BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

55

References/Del Giudice: Sex, attachment, and the development of reproductive strategies 4. The increasing proportion of male children in the last thirty years is probably explained by selective abortion; however, neglect of female children and selective adoption have also been suggested as potential factors, and parents may sometimes omit to register the birth of a girl; see Coale and Banister (1996) and Ding and Hesketh (2006). 5. There are some discrepancies in the literature about the exact timing of adrenarche, as reflected, for example, in Flinn et al.’s commentary. I find most convincing the data showing initial adrenal activity at about 6 years (at least in some children), and smaller sex differences in age of onset compared with gonadarche. See section 7.2.1 for references.

References Letters “a” and “r” appearing before authors’ initials stand for target article and response references, respectively. Abbott, D. H., Saltzman, W., Schultz-Darken, N. J. & Tannenbaum, P. L. (1998) Adaptations to subordinate status in female marmoset monkeys. Comparative Biochemistry and Physiology, Part C: Pharmacology, Toxicology and Endocrinology 119(3):261 – 74. [LJS] Ackerman, J. M. & Kenrick, D. T. (2008) The costs of benefits: Help-refusals highlight key trade-offs of social life. Personality and Social Psychology Review 12:118 – 40. [JMA] (under review) Raising and razing romantic relationship barriers: How men and women cooperate in courtship. [JMA] Ackerman, J. M., Kenrick, D. T. & Schaller, M. (2007) Is friendship akin to kinship? Evolution and Human Behavior 28:365 – 74. [JMA] Adkins-Regan, E. (2005) Hormones and animal social behavior. Princeton University Press. [aMDG] Adolphs, R. (2003) Cognitive neuroscience of human social behavior. Nature Reviews/Neuroscience 4:165– 79. [MVF] Ahnert, L. (2005) Parenting and alloparenting: The impact on attachment in humans. In: Attachment and bonding: A new synthesis, ed. C.S. Carter, L. Ahnert, K. E. Grossmann, S. B. Hrdy, M. E. Lamb, S.W. Porges & N. Sachser, pp. 229 – 44. MIT Press. [aMDG] Ainsworth, M. D. S. (1989) Attachments beyond infancy. American Psychologist 44:709 – 16. [DKS] Ainsworth, M. D. S., Blehar, M. C., Waters, E. & Wall, S. (1978) Patterns of attachment: A psychological study of the Strange Situation. Erlbaum. [arMDG] Allen, E. S. & Baucom, D. H. (2004) Adult attachment and patterns of extradyadic involvement. Family Process 43:467– 88. [arMDG] Allen, J. P. & Land, D. (1999) Attachment in adolescence. In: Handbook of attachment: Theory, research and clinical applications, ed. J. Cassidy & P. R. Shaver, pp. 355 – 77. Guilford. [aMDG, DKS] Als, H. (1986) A synactive model of neonatal behavioral organization: Framework for assessment of neurobehavioral development in the premature infant and for support of infants and parents in the neonatal intensive care environment. Physical and Occupational Therapy in Pediatrics 6(Special issue):3– 55. [SGF] Als, H., Duffy, F. H., McAnulty, G. B., Rivkin, M. J., Vajapeyam, S., Mulkern, R. V., Warfield, S. K., Huppi, P. S., Butler, S. C., Conneman, N., Fischer, C. & Eichenwald, E. C. (2004) Early experience alters brain function and structure. Pediatrics 113(4):846 – 57. [SGF] Als, H., Lawhon, G., Duffy, F. H., McAnulty, G. B., Gibes-Grossman, R. & Blickman, J. G. (1994) Individualized developmental care for the very low-birth-weight preterm infant: Medical and neurofunctional effects. Journal of the American Medical Association (JAMA) 272(11):853 –58. [SGF] American Psychiatric Association (2000) Diagnostic and statistical manual of mental disorders, 4th edition [Text revision]. American Psychiatric Association. [AC, rMDG] Ammaniti, M., Van IJzendoorn, M. H., Speranza, A. M. & Tambelli, R. (2000) Internal working models of attachment during late childhood and early adolescence: An exploration of stability and change. Attachment and Human Development 2:328 – 46. [rMDG, KAK, MJB-K] Anderson, P. J. & Doyle, L. W. (2008) Cognitive and educational deficits in children born extremely preterm. Seminars in Perinatology 32:51– 58. [SGF] Andersson, M. (1994) Sexual selection. Princeton University Press. [aMDG] Andrews, P. A., Gangestad, S. W. & Mathews, D. (2002) Adaptationism – How to carry out an exaptationist program. Behavioral and Brain Sciences 25:489 – 553. [JH]

56

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

Archer, J. (1996) Sex differences in social behavior – Are the social role and evolutionary explanations compatible? American Psychologist 51(9):909– 17. [SGF] Archer, J. & Mehdikhani, M. (2000) Strategic pluralism: Men and women start from a different point. Behavioral and Brain Sciences 23:588. [aMDG] Arck, P. C., Rose, M., Hertwig, K., Hagen, E., Hildebrandt, M. & Klapp, B.F. (2001) Stress and immune mediators in miscarriage. Human Reproduction 16:1505 –11. [rMDG] Arlt, W., Callies, F. & Allolio, B. (2000) DHEA replacement in women with adrenal insufficiency – pharmokinetics, bioconversion, and clinical effects on well-being, sexuality and cognition. Endocrine Research 26:505 – 11. [MVF] Arnold, F. & Liu, Z. (1986) Sex preference, fertility, and family planning in China. Population and Development Review 12:221 –46. [NL] Auchus, R. J. & Rainey, W. E. (2004) Adrenarche – physiology, biochemistry and human disease. Clinical Endocrinology 60:288 – 96. [aMDG] Aviezer, O., Sagi, A., Resnick, G. & Gini, M. (2002) School competence in young adolescence: Links to early attachment relationships beyond concurrent self-perceived competence and representations of relationships. International Journal of Behavioral Development 26:397 – 409. [aMDG] Bakermans-Kranenburg, M. J., van IJzendoorn, M. H., Bokhorst, C. L. & Schuengel, C. (2004) The importance of shared environment in infant-father attachment: A behavioral genetic study of the Attachment Q-sort. Journal of Family Psychology 18:545– 49. [aMDG] Bakermans-Kranenburg, M. J., van IJzendoorn, M. H. & Juffer, F. (2005) Disorganized infant attachment and preventive interventions: A review and meta-analysis. Infant Mental Health Journal 26:191 – 216. [AJL] Baldwin, M. W. & Fehr, B. (1995) On the instability of attachment style ratings. Personal Relationships 2:247 – 61. [arMDG] Barber, N. (2000) On the relationship between country sex ratios and teen pregnancy rates: A replication. Cross-Cultural Research 34:26– 37. [aMDG] Barker, D. J. (1998) In utero programming of chronic disease. Clinical Science 95:115 – 28. [MVF] Barrett, J., Abbott, D. H. & George, L. M. (1990) Extension of reproductive suppression by pheromonal cues in subordinate female marmoset monkeys, Callithrix jacchus. Journal of Reproduction and Fertility 90(2):411 – 18. [LJS] Bartholomew, K. (1990) Avoidance of intimacy: An attachment perspective. Journal of Social and Personal Relationships 7:147 – 78. [aMDG] Bartholomew, K. & Allison, C. J. (2006) An attachment perspective on abusive dynamics in intimate relationships. In: Dynamics of romantic love: Attachment, caregiving, and sex, ed. M. Mikulincer & G. S. Goodman, pp. 71 – 101. Guilford Press. [DKS] Bartholomew, K. & Horowitz, L. M. (1991) Attachment styles in young adults: A test of a four-category model. Journal of Personality and Social Psychology 61:226 – 44. [arMDG, DKS, VZ] Bartholomew, K. & Shaver, P. R. (1998) Methods of assessing adult attachment: Do they converge? In: Attachment theory and close relationships, ed. J.A. Simpson & W. S. Rholes, pp. 25– 45. Guilford Press. [rMDG] Bateson, P. (2005) The return of the whole organism. Journal of Biosciences 30:101 – 109. [aMDG] Beijersbergen, M. D., Bakermans-Kranenburg, M. J., Van IJzendoorn, M. H. & Juffer, F. (in press) Stress regulation in adolescents: Physiological reactivity during the adult attachment interview and conflict interaction. Child Development. [MJB-K] Belsky, J. (1997a) Attachment, mating and parenting: An evolutionary interpretation. Human Nature 8:361 – 81. [CDG, aMDG] (1997b) Variation in susceptibility to environmental influence: An evolutionary argument. Psychological Inquiry 8:82 – 186. [aMDG] (1999) Modern evolutionary theory and patterns of attachment. In: Handbook of Attachment: Theory, research and clinical applications, ed. J. Cassidy & P. R. Shaver, pp. 141 – 61. Guilford. [CDG, aMDG] (2000) Fixed versus flexible strategies: Individual differences in facultative responsiveness? Behavioral and Brain Sciences 23:591 – 92. [aMDG] (2002) Developmental origins of attachment styles. Attachment and Human Development 4:166– 70. [MVF, aMDG] (2005) The developmental and evolutionary psychology of intergenerational transmission of attachment. In: Attachment and bonding: A new synthesis, ed. C.S. Carter, L. Ahnert, K. E. Grossmann, S. B. Hrdy, M. E. Lamb, S. W. Porges & N. Sachser, pp. 169 – 98. MIT Press. [aMDG, LP] (2007) Experience in childhood and the development of reproductive strategies. Acta Psychologica Scienica 39:454 – 68. [DKS] Belsky, J., Bakermans-Kranenburg, M. J. & van IJzendoorn, M. H. (2007a) For better and for worse: Differential susceptibility to environmental influences. Current Directions in Psychological Science 16:300 – 304. [arMDG] Belsky, J., Steinberg, L. D., Houts, R. M., Friedman, S. L., DeHart, G., Cauffman, E., Roisman, G. I., Halpern-Felsher, B. L., Susman, E. & The NICHD Early Child Care Research Network (2007b) Family rearing antecedents of pubertal timing. Child Development 78:1302 – 21. [aMDG]

References/Del Giudice: Sex, attachment, and the development of reproductive strategies Belsky, J., Steinberg, L. & Draper, P. (1991) Childhood experience, interpersonal development, and reproductive strategy: An evolutionary theory of socialization. Child Development 62:647 – 70. [arMDG, JH, JJJ, DJK, DKS] Benenson, J. & Christakos, A. (2003) The greater fragility of females’ versus males’ closest same-sex friendships. Child Development 74:1123– 29. [CDG] Benoit, D. & Parker, K. C. H. (1994) Stability and transmission of attachment across three generations. Child Development 65:1444– 56. [aMDG] Bereczkei, T. & Csanaky, A. (2001) Stressful family environment, mortality, and child socialisation: Life-history strategies among adolescents and adults from unfavourable social circumstances. International Journal of Behavioral Development 25:501 – 508. [aMDG] Bernier, A. & Dozier, M. (2002) Assessing adult attachment: Empirical sophistication and conceptual bases. Attachment and Human Development 4:171– 79. [arMDG] Bernroider, G., Holztrattner, M. & Rottner, K. (1996) Sex steroid-opioid interactions associated with the temporal component of avian calling patterns. Hormones and Behavior 30:583 – 89. [aMDG] Betzig, L. (1989) Causes of conjugal dissolution: A cross-cultural study. Current Anthropology 30:654 – 76. [aMDG] Birnbaum, G. E. (2007) Beyond the borders of reality: Attachment orientations and sexual fantasies. Personal Relationships 14:321 – 42. [arMDG] Birnbaum, G. E., Reis, H. T., Mikulincer, M., Gillath, O. & Orpaz, A. (2006) When sex is more than just sex: Attachment orientations, sexual experience, and relationship quality. Journal of Personality and Social Psychology 91:929 – 43. [LP, VZ] Bjorklund, D. F. (1997) The role of immaturity in human development. Psychological Bulletin 122:153– 69. [aMDG] Bjorklund, D. F. & Pellegrini, A. D. (2002) The origins of human nature: Evolutionary developmental psychology. American Psychological Association Press. [MVF] Bjorklund, D. F. & Rosenberg, J. S. (2005) The role of developmental plasticity in the evolution of human cognition: evidence from enculturated, juvenile great apes. In: Origins of the social mind: Evolutionary psychology and child development, ed. B. J. Ellis & D. F. Bjorklund, pp. 45 – 75. Guilford Press. [aMDG] Blurton Jones, N. (1989) The cost of children and the adaptive scheduling of births: Towards a sociobiological perspective on demography. In: Sociobiology of sexual and reproductive strategies, ed. A. Rasa, C. Vogel & E. Voland. Chapman and Hall. [aMDG] Blurton Jones, N. G., Marlowe, F. W., Hawkes, K. & O’Connell, J. F. (2000) Paternal investment and hunter-gatherer divorce rates. In: Adaptation and human behavior: An anthropological perspective, ed. L. Cronk, N. Chagnon & W. Irons, pp. 69– 90. Aldine De Gruyter. [aMDG] Bogaert, A. F. & Sadava, S. (2002) Adult attachment and sexual behavior. Personal Relationships 9:191– 204. [arMDG, VZ] Bogin, B. (1999) Evolutionary perspective on human growth. Annual Review of Anthropology 28:109 – 53. [MVF, arMDG] Bokhorst, C. L., Bakermans-Kranenburg, M. J., Fearon, R. M. P., van IJzendoorn, M. H., Fonagy, P. & Schuengel, C. (2003) The importance of shared environment in mother-infant attachment security: A behavioral genetic study. Child Development 74:1769 – 82. [aMDG] Booth-Laforce, C., Oh, W., Hayoung Kim, A., Rubin, K. H., Rose-.Krasnor, L. & Burgess, K. (2006) Attachment, self-worth, and peer-group functioning in middle childhood. Attachment and Human Development 8:309 – 25. [B-BC, aMDG] Boothroyd, L. G., Jones, B. C., Burt, D. M., DeBruine, L. M. & Perrett, D. I. (2008) Facial correlates of sociosexuality. Evolution and Human Behavior 29:211 – 18. [LEV] Borgerhoff Mulder, M. (1989) Menarche, menopause, and reproduction in the Kipsigis of Kenia. Journal of Biosocial Science 21:179 –92. [aMDG] (1998) Brothers and sisters: How sibling interactions affect optimal parental allocations. Human Nature 9(2):119 – 62. [RJQ] Bowlby, J. (1969/1982) Attachment and loss. Vol. 1. Attachment, 2nd edition. Basic Books. (Original work published 1969). [CDG, aMDG, DP, DKS, VZ] (1973) Attachment and loss. Vol. 2. Separation. Basic Books. [aMDG] (1980) Attachment and loss. Vol. 3. Loss. Basic Books. [aMDG, NL] Boyce, W. T. & Ellis, B. J. (2005) Biological sensitivity to context: I. An evolutionary– developmental theory of the origins and functions of stress reactivity. Development and Psychopathoogy 17:271 – 301. [aMDG] Brassard, A., Shaver, P. R. & Lussier, Y. (2007) Attachment, sexual experience, and sexual pleasure in romantic relationships: A dyadic approach. Personal Relationships 14:475 – 93. [arMDG, VZ] Breaugh, J. A. (2003) Effect size estimation: Factors to consider and mistakes to avoid. Journal of Management 29:79 – 97. [rMDG] Brennan, K. A., Clark, C. L. & Shaver, P. R. (1998) Self-report measurement of adult attachment: An integrative overview. In: Attachment theory and close relationships, ed. J.A. Simpson & W. S. Rholes, pp. 46 – 76. Guilford Press. [arMDG, NL, DKS]

Brennan, K. A. & Shaver, P. R. (1995) Dimensions of adult attachment, affect regulation, and romantic relationship functioning. Personality and Social Psychology Bulletin 21:267 – 83. [VZ] Bretherton, I. & Munholland, K. A. (1999) Internal working models in attachment relationships: A construct revisited. In: Handbook of Attachment: Theory, research and clinical applications, ed. J. Cassidy & P. R. Shaver. Guilford. [aMDG] Bretherton, I., Ridgeway, D. & Cassidy, J. (1990) Assessing internal working models of the attachment relationship: An attachment story completion task. In: Attachment in the preschool years: Theory, research, and intervention, ed. M. T. Greenberg, D. Cicchetti, & E. M. Cummings, pp. 273 – 308. University of Chicago Press. [aMDG, MJB-K] Brody, S. & Breitenstein, C. (2000) The trade-off between frequency of intercourse and sexual partner accumulation may reflect evolutionary adaptations. Behavioral and Brain Sciences 23:594. [rMDG] Brumbaugh, C. C. & Fraley, R. C. (2006) The evolution of attachment in romantic relationships. In: Dynamics of romantic love: Attachment, caregiving, and sex, ed. M. Mikulincer & G. S. Goodman, pp. 71– 101. Guilford Press. [DKS] Brussoni, M. J., Lang, K., Livesley, W. J. & Macbeth, T. M. (2000) Genetic and environmental influences on adult attachment styles. Personal Relationships 7:283– 89. [aMDG] Bukowski, W. M., Sippola, L. A. & Newcomb, A. F. (2000) Variations in patterns of attraction to same- and other-sex peers during early adolescence. Developmental Psychology 36:147 – 54. [aMDG] Bullinaria, J. A. (2007) The effect of learning on life history evolution. In: Proceedings of the Genetic and Evolutionary Computation Conference (GECCO 2007), pp. 222 – 29. ed. H. Lipson. Association for Computing Machinery. ACM. Available at: http://www.cs.bham.ac.uk/jxb/PUBS/ LELHE.pdf. [DP] Bureau, J. F., Beliveau, M. J., Moss, E. & Lepine, S. (2006) Association between the mother-child bond and the stories of attachment in the school age. Canadian Journal of Behavioural Science (Revue Canadienne des Sciences du Comportement) 38:50– 62. [rMDG, MJB-K] Burnstein, E., Crandall, C. & Kitayama, S. (1994) Some neo-Darwinian decision rules for altruism. Journal of Personality and Social Psychology 67:773– 89. [JMA] Buss, D. M. & Greiling, H. (1999) Adaptive individual differences. Journal of Personality 67:209 – 43. [CDG] Buss, D. M. & Schmitt, D. P. (1993) Sexual strategies theory: An evolutionary perspective on human mating. Psychological Review 100:204 – 32. [aMDG] Butzer, B. & Campbell, L. (2008) Adult attachment, sexual satisfaction, and relationship satisfaction: A study of married couples. Personal Relationships 15:141 – 54. [rMDG, DKS] Campbell, A. (1992) The girls in the gang, 2nd edition. Basil Blackwell. [AC] (1995) A few good men: The evolutionary psychology of female adolescent aggression. Ethology and Sociobiology 16:99 – 123. [AC] (1999) Staying alive: Evolution, culture and intrasexual aggression. Behavioral and Brain Sciences 22:203 – 52. [B-BC] (2000) Putting people before parasites and places. Behavioral and Brain Sciences 23:596 – 97. [aMDG] (2008a) Attachment, aggression and affiliation: The role of oxytocin in female social behavior. Biological Psychology 77:1 – 10. [rMDG] (2008b) The morning after the night before: Affective reactions to one-night stands among mated and unmated women and men. Human Nature 19:157 – 73. [AC] Campbell, B. & Udry, J. R. (1995) Stress and age at menarche of mothers and daughters. Journal of Biosocial Science 27:127 – 34. [aMDG] Campbell, B. C. (2006) Adrenarche and the evolution of human life history. American Journal of Human Biology 18:569 – 89. [MVF] Campbell, L., Simpson, J. A., Boldry, J. & Kashy, D. A. (2005) Perceptions of conflict and support in romantic relationships: The role of attachment anxiety. Journal of Personality and Social Psychology 88:510 – 31. [rMDG] Card, N. A. & Hodges, V. E. (2003) Parent-child relationships and enmity with peers: The role of avoidant and preoccupied attachment. New Directions for Child and Adolescent Development 102:23 – 37. [aMDG] Carlson, A. A., Ziegler, T. E. & Snowdon C. T. (1997) Ovarian function of pygmy marmoset daughters (Cebuella pygmaea) in intact and motherless families. American Journal of Primatology 43(4):347 – 55. [LJS] Carlson, V., Cicchetti, D., Barnett, D. & Braunwald, K. (1989) Disorganized/ disoriented attachment relationships in maltreated infants. Developmental Psychology 25:525 – 31. [aMDG] Carnelly, K. B. & Brennan, K. A. (2002) Building bridges. Attachment and Human Development 4:189– 92. [rMDG] Carter, C. S. (1998) Neuroendocrine perspectives on social attachment and love. Psychoneuroendocrinology 23:779 – 818. [aMDG] Cashdan, E. (1996) Women’s mating strategies. Evolutionary Anthropology 5:134– 43. [aMDG]

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

57

References/Del Giudice: Sex, attachment, and the development of reproductive strategies Caspi, A. & Roberts, B. W. (2001) Personality development across the life course: The argument for change and continuity. Psychological Inquiry 12:49–66. [JRH] Cassidy, J. (1988) Child-mother attachment and the self in six-year-olds. Child Development 59:121 – 34. [aMDG, MJB-K] (1999) The nature of the child’s ties. In: Handbook of attachment: Theory, research and clinical applications, ed. J. Cassidy & P. R. Shaver. Guilford. [aMDG] (2003) Continuity and change in the measurement of infant attachment: Comment on Fraley and Spieker (2003). Developmental Psychology 39:409–12. [aMDG] Cassidy, J. & Berlin, L. (1994) The insecure/ambivalent pattern of attachment: Theory and research. Child Development 65:971 – 91. [arMDG] Cassidy, J. & Shaver, P. R., eds. (1999) Handbook of attachment: Theory, research and clinical applications. Guilford. [aMDG] Cellerino, A. & Jannini, E. A. (2005) Male reproductive physiology as a sexually selected handicap? Erectile dysfunction is correlated with general health and health prognosis and may have evolved as a marker of poor phenotypic quality. Medical Hypotheses 65:179 – 84. [SGF] Cesario S. K. & Hughes, L. A. (2007) Precocious puberty: A comprehensive review of literature. Journal of Obstetric, Gynecologic, and Neonatal Nursing 36:263 – 74. [JRH] Chan, C., Eric, B. & Chan C. (2006) Attitudes to and practices regarding sex selection in China. Prenatal Diagnosis 26:610 – 13. [NL] Chang, F., Kidd, J. R., Livak, K. J., Pakstis, A. J. & Kidd, K. K. (1996) The worldwide distribution of allele frequencies at the human dopamine D4 receptor locus. Human Genetics 98:91 – 101. [rMDG] Chasiotis, A., Scheffer, D., Restemeier, R. & Keller, H. (1998) Intergenerational context discontinuity affects the onset of puberty: A comparison of parent – child dyads in West and East Germany. Human Nature 9:321– 39. [aMDG] Chen, C. C. G. & Parker, C. R. Jr. (2004) Adrenal androgens and the immune system. Seminars in Reproductive Medicine 22:369 –77. [MVF] Chisholm, J. S. (1993) Death, hope, and sex: Life-history theory and the development of reproductive strategies. Current Anthropology 34:1 – 24. [aMDG, JH] (1996) The evolutionary ecology of attachment organization. Human Nature 7:1– 38. [aMDG, JH, RJQ] (1999) Death, hope and sex: Steps to an evolutionary ecology of mind and morality. Cambridge University Press. [arMDG, JH, DKS] Chisholm, J. S., Burbank, V. K., Coall, D. A. & Gemmiti, F. (2005a) Early stress: Perspectives from developmental evolutionary ecology. In: Origins of the social mind: Evolutionary psychology and child development, ed. B.J. Ellis & D. F. Bjorklund, pp. 76 – 107. Guilford. [aMDG] Chisholm, J. S., Quinlivan, J. A., Petersen, R. W. & Coall, D. A. (2005b) Early stress predicts age at menarche and first birth, adult attachment, and expected lifespan. Human Nature 16:233 – 65. [aMDG] Cho, J., Holditch-Davis, D. & Belyea, M. (2007) Gender and racial differences in the looking and talking behaviors of mothers and their 3-year-old prematurely born children. Journal of Pediatric Nursing 22(5):356 – 67. [SGF] Chugani, H. T. (1998) Critical period of brain development: Studies of cerebral glucose utilization with PET. Preventive Medicine 27:184 – 88. [MVF] Clark, A. & Daly, M. (2005) What is the significance of cross-national variability in sociosexuality? Behavioral and Brain Sciences 28:280. [aMDG] Clutton-Brock, T. H. (1991) The evolution of parental care. Princeton University Press. [aMDG, NL, RJQ] (2002) Breeding together: Kin selection and mutualism in cooperative vertebrates. Science 296:69 – 72. [aMDG] Clutton-Brock, T. H., ed. (1988) Reproductive success: Studies of individual variation in contrasting breeding systems. The University of Chicago Press. [DM] Coale, A. J. & Banister, J. (1996) Five decades of missing females in China. Proceedings of the American Philosophical Society 140:421– 50. [rMDG] Cohen, D. L. & Belsky, J. (2008) Individual differences in female mate preferences as a function of attachment and hypothetical ecological conditions. Journal of Evolutionary Psychology 6:25 – 42. [rMDG] Cohen, J. (1988) Statistical power analysis for the behavioral sciences. Erlbaum. [rMDG] Collins, N. L. & Read, S. J. (1990) Adult attachment, working models and relationship quality in dating couples. Journal of Personality and Social Psychology 58:644 – 63. [aMDG] Comings, D. E., Muhlemann, D., Johnson, J. P. & Mac Murray, J. P. (2002) Parent-daughter transmission of the androgen receptor gene as an explanation of the effect of father absence on age of menarche. Child Development 73:1046 – 51. [aMDG] Compagnone, N. A. & Mellon, S. H. (1998) Dehydroepiandrosterone: A potential signaling molecule for neocortical organization during development. Proceedings of the National Academy of Sciences USA 95:4678 – 83. [MVF]

58

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

Cooper, E. (2006) The runaway and homeless youth program: Administrative, funding, and legislative action. Congressional Research Service, Library of Congress. Available at: http://digital.library.unt.edu/govdocs/crs/permalink/ meta-crs-9962:1. [MJK] Cooper, M. L., Shaver, P. R. & Collins, N. L. (1998) Attachment styles, emotion regulation, and adjustment in adolescence. Journal of Personality and Social Psychology 74:1380– 97. [arMDG] Corby, B. C. (2006) Attachment and adjustment in preadolescence. Unpublished doctoral dissertation, Psychology Department, Florida Atlantic University. [arMDG] Cosmides, L. & Tooby, J. (1994) Origins of domain specificity: The evolution of functional organization. In: Mapping the mind: Domain specificity in cognition and culture, ed. L. A. Hirschfeld & S. A. Gelman, pp. 85 –116. Cambridge University Press. [DJK] Costa, R. M. & Brody, S. (2007) Women’s relationship quality is associated with specifically penile-vaginal intercourse orgasm and frequency. Journal of Sex and Marital Therapy 33:319 – 27. [rMDG] Cramer, K. M. & Neyedley, K. A. (1998) Sex differences in loneliness: The role of masculinity and femininity. Sex Roles 38:645 –50. [VZ] Crawford, C. & Anderson, J. (1989) Sociobiology: An environmentalist discipline? American Psychologist 44:1449 –60. [DJK] Crawford, T. N., Shaver, P. R., Cohen, P., Pilkonis, P. A., Gillath, O. & Kasen, S. (2006) Self-reported attachment, interpersonal aggression, and personality disorder in a prospective community sample of adolescents and adults. Journal of Personality Disorders 20:331 – 51. [rMDG] Critchfield, K. L., Levy, K. N., Clarkin, J. F. & Kernberg, O. F. (2008) The relational context of aggression in Borderline Personality Disorder: Using adult attachment style to predict forms of hostility. Journal of Clinical Psychology 64:67– 82. [AC] Cronin, H. (1991) The ant and the peacock. Cambridge University Press. [LB] Cronk, L. (2000) Female-biased parental investment and growth performance among the Mukogodo. In: Adaptation and human behavior: An anthropological perspective, ed. L. Cronk, N. Chagnon, & W. Irons, pp. 203–21. [RJQ] Crowell, J. A., Fraley, R. C. & Shaver, P. R. (1999/2008) Measurement of individual differences in adolescent and adult attachment. In: Handbook of attachment: Theory, research and clinical applications, ed. J. Cassidy & P. R. Shaver. Guilford (2nd Edition). [LB, aMDG] Crowell, J. A., Treboux, D. & Waters, E. (2002) Stability of attachment representations: The transition to marriage. Developmental Psychology 38:467 – 79. [aMDG] Cummings, F. M. (2003) Toward assessing attachment on an emotional security continuum: Comment on Fraley and Spieker (2003). Developmental Psychology 39:405 – 408. [aMDG] Cushing, B. S. & Kramer, K. M. (2005) Mechanisms underlying epigenetic effects of early social experience: The role of neuropeptides and steroids. Neuroscience and Biobehavioral Reviews 29:1089 – 105. [B-BC] Daly, M. & Wilson, M. (1996) Violence against stepchildren. Current Directions in Psychological Science 5:77 – 81. [rMDG] David, D. H. & Lyons-Ruth, K. (2005) Differential attachment responses of male and female infants to frightening maternal behavior: Tend or befriend versus fight or flight? Infant Mental Health Journal 26:1 – 18. [aMDG] Davies, P. T. & Forman, E. M. (2002) Children’s patterns of preserving emotional security in the interparental subsystem. Child Development 73:1880– 903. [B-BC] Davis, D., Shaver, P. R. & Vernon, M. L. (2004) Attachment style and subjective motivations for sex. Personality and Social Psychology Bulletin 30:1076– 90. [DKS] Davis, M. & Emory, E. (1995) Sex differences in neonatal stress reactivity. Child Development 66:14– 27. [SGF] Dawkins, R. (2006) The selfish gene. Oxford University Press. [JMA] DeKlyen, M., Speltz, M. & Greenberg, M. T. (1999) Fathering and early onset conduct problems: Positive and negative parenting, father–son attachment, and the marital context. Clinical Child and Family Psychology Review 1:3–21. [aMDG] Del Giudice, M. (2008) Sex-biased ratio of avoidant/ambivalent attachment in middle childhood. British Journal of Developmental Psychology. 26:369 – 79. [aMDG, MJB-K] Del Giudice, M., Angeleri, R. & Manera, V. (in press) The juvenile transition: A developmental switch point in human life history. Developmental Review. doi:10.1016/j.dr.2008.09.001 [rMDG] Delton, A. W., Robertson, T. E. & Kenrick, D. T. (2006) The mating game isn’t over: A reply to Buller’s critique of the evolutionary psychology of mating. Evolutionary Psychology 4:262– 73. [aMDG] Deulofeut, R., Dudell, G. & Sola, A. (2007) Treatment-by-gender effect when aiming to avoid hyperoxia in preterm infants in the NICU. Acta Paediatrica 96:990 – 94. [SGF] De Wolff, M.S. & van IJzendoorn, M. H. (1997) Sensitivity and attachment: A meta-analysis on parental antecedents of infant attachment. Child Development 68:571 – 91. [JH, AJL]

References/Del Giudice: Sex, attachment, and the development of reproductive strategies Dhom, G. (1973) The prepubertal and pubertal growth of the adrenal (adrenarche). Beitrage Pathologie 150:357– 77. [MVF] Ding, Q. J. & Hesketh, T. (2006) Family size, fertility preferences, and sex ratio in China in the era of the one child family policy: Results from national family planning and reproductive health survey. British Medical Journal 333:371 – 73. [rMDG] Ding, Y. C., Chi, H. C., Grady, D. L., Morishima, A., Kidd, J. R., Kidd, K. K., Foldman, P., Spence, M. A., Schuck, S., Swanson, J. M., Zhang, Y. P. & Moyzis, R. K. (2002) Evidence of positive selection at the human dopamine receptor D4 gene locus. Proceedings of the National Academy of Sciences USA 99:309 – 14. [rMDG] Dishion, T. J., Duncan, T. E., Eddy, J. M., Fagot, B. I. & Fetrow, R. (1994) The world of parents and peers: Coercive exchanges and children’s social adaptation. Social Development 3:255– 68. [JRH] Doligez, B., Danchin, E. & Clobert, J. (2002) Public information and breeding habitat selection in a wild bird population. Science 297:1168 –70. [JRH] Domes, G., Heinrichs, M., Michel, A., Berger, C. & Herpetz, S. C. (2007) Oxytocin improves “mind-reading” in humans. Biological Psychiatry 61:731 – 33. [rMDG] Dong, M. & Li, S. (2007) Conflict resolution in Chinese family purchase decisions: The impact of changing female roles and marriage duration. International Journal of Conflict Management 18:308 – 24. [NL] D’Onofrio, B. M., Turkheimer, E., Emery, R. E., Slutske, W. S., Heath, A. C., Madden, P. A. & Martin, N. G. (2006) A genetically informed study of the processes underlying the association between parental marital instability and offspring adjustment. Developmental Psychology 42:486 – 99. [rMDG] Doran, J. E. (1994) Modelling early human societies. Artificial Intelligence and Simulation of Behaviour Quarterly 90(5):31 – 33. Available at: http:// citeseerx.ist.psu.edu/viewdoc/summary?doi¼10.1.1.55.2287. [DP] Draper, P. (1989) African marriage systems: Perspectives from evolutionary ecology. Ethology and Sociobiology 10:145 – 69. [aMDG] Draper, P. & Harpending, H. (1982) Father absence and reproductive strategy: An evolutionary perspective. Journal of Anthropological Research 38:255 –73. [MVF, aMDG] (1988) A sociobiological perspective on the development of human reproductive strategies. In: Sociobiological perspectives on human development, ed. K. B. MacDonald, pp. 340 – 72. Springer-Verlag. [aMDG] Ducharme, J., Doyle, A. B. & Markiewicz, D. (2002) Attachment security with mother and father: Associations with adolescents’ reports of interpersonal behavior with parents and peers. Journal of Social and Personal Relationships 19:203 – 31. [aMDG] Dunbar, R. I. M. & Schultz, S. (2007) Evolution in the social brain. Science 317:1344 – 47. [aMDG] Durrett, M. E., Otaki, M. & Richards, P. (1984) Attachment and the mother’s perception of support from the father. International Journal of Behavioral Development 7:167– 76. [NL] Easterbrooks, A., Biesecker, G. & Lyons-Ruth, K. (2000) Infancy predictors of emotional availability in middle childhood: The role of attachment security and maternal depressive symptomatology. Attachment and Human Development 2(2):170 – 87. [MJK] Egan, V. & Angus, S. (2004) Is social dominance a sex-specific strategy for infidelity? Personality and Individual Differences 36:575 –86. [aMDG] Eibl-Eibesfeldt, I. (1989) Human ethology. Aldine de Gruyter. [aMDG] Ellis, B. J. (1998) The Partner-Specific Investment Inventory: An evolutionary approach to individual differences in investment. Journal of Personality 66:383 – 441. [rMDG] (2004) Timing of pubertal maturation in girls: An integrated life history approach. Psychological Bulletin 130:920– 58. [arMDG, JJJ. LP] (2005) Determinants of pubertal timing: An evolutionary developmental approach. In: Origins of the social mind: Evolutionary psychology and child development, ed. B.J. Ellis & D. F. Bjorklund, pp. 164 –88. Guilford. [aMDG] Ellis, B. J. & Essex, M. J. (2007) Family environments, adrenarche, and sexual maturation: A longitudinal test of a life history model. Child Development 78:1799 – 817. [MVF, arMDG] Ellis, B. J. & Garber, J. (2000) Psychosocial antecedents of variation in girls’ pubertal timing: Maternal depression, stepfather presence, and marital and family stress. Child Development 71:485 – 501. [JJJ] Ellis, B. J., Jackson, J. J. & Boyce, W. T. (2006) The stress response system: Universality and adaptive individual differences. Developmental Review 26:175 – 212. [aMDG] Ellison, P. T. (2001) On fertile ground: A natural history of human reproduction. Harvard University Press. [arMDG] Emery, J., Paguette, D. & Bigras, M. (2008) Factors predicting attachment patterns in infants of adolescent mothers. Journal of Family Studies 14:65 – 90. [NL] Emlen, D. (1997) Alternative reproductive tactics and male dimorphism in the horned beetle Onthophagus acuminatus (Coleoptera: Scarabaeidae). Behavioral Ecology and Sociobiology 41:335 – 41. [aMDG]

Erickson, M. F., Sroufe, L. A. & Egeland, B. (1985) The relationship between quality of attachment and behavior problems in preschool in a high-risk sample. Monographs of the Society for Research in Child Development 50:147 – 66. [aMDG] Feeney, B. C. & Collins, N. L. (2001) Predictors of caregiving in adult intimate relationships: An attachment theoretical perspective. Journal of Personality and Social Psychology 80:972 – 94. [VZ] Feeney, J. A. (1999) Adult romantic attachment and couple relationships. In: Handbook of attachment: Theory, research and clinical applications, ed. J. Cassidy & P. R. Shaver, pp. 355–77. Guilford Press. [B-BC, aMDG, LP] (2002) Attachment-related dynamics: What can we learn from self-reports of avoidance and anxiety? Attachment and Human Development 4:193– 200. [rMDG] Feeney, J. A. & Noller, P. (1990) Attachment style as a predictor of adult romantic relationships. Journal of Personality and Social Psychology 58:281 – 91. [aMDG] (2004) Attachment and sexuality in close relationships. In: Handbook of sexuality in close relationships, ed. J. H. Harvey, A. Wenzel, and S. Sprecher, pp. 183 – 201. Erlbaum. [DKS] Feeney, J. A., Noller, P. & Patty, J. (1993) Adolescents’ interactions with the opposite sex: Influence of attachment style and gender. Journal of Adolescence 16:169 – 89. [DKS] Feinberg, D. R., Jones, B. C., Law-Smith, M. J., Moore, F. R., DeBruine, L. M., Cornwell, R. E., Hillier, S. G. & Perrett, D. I. (2006) Menstrual cycle, trait estrogen level, and masculinity preferences in the human voice. Hormones and Behavior 49:215 –22. [LEV] Figueredo, A. J. & Jacobs, W. J. (2000) Strategic pluralism through a Brunswikian lens. Behavioral and Brain Sciences 23:603– 604. [JJJ] Figueredo, A. J., Va´squez, G., Brumbach, B. H. & Schneider, S. M. R. (2004) The heritability of life history strategy: The K-factor, covitality, and personality. Social Biology 51:121 – 43. [AJF, arMDG] (2007) The K-factor, covitality, and personality: A psychometric test of life history theory. Human Nature 18(1):47 – 73. [AJF] Figueredo, A. J., Va´squez, G., Brumbach, B., Schneider, S. M. R., Sefcek, J. A., Tal, I. R., Hill, D., Wenner, C. J. & Jacobs, W. J. (2006) Consilience and life history theory: From genes to brain to reproductive strategy. Developmental Review 26:243 – 75. [AJF, aMDG] Finkel, D., Wille, D. E. & Matheny, A. P. (1998) Preliminary results from a twin study of infant-caregiver attachment. Behavior Genetics 28:1 – 8. [aMDG] Finnegan, R. A., Hodges, E. & Perry, D. (1996) Preoccupied and avoidant coping during middle childhood. Child Development 67:1318– 28. [B-BC, arMDG] Flinn, M. V. (2006) Evolution and ontogeny of stress response to social challenges in the human child. Developmental Review 26:138 – 74. [MVF, rMDG] Flinn, M. V., Geary, D. C. & Ward, C. V. (2005) Ecological dominance, social competition, and coalitionary arms races: Why humans evolved extraordinary intelligence. Evolution and Human Behavior 26(1):10 – 46. [MVF] Flinn, M. V. & Ward, C. V. (2005) Ontogeny and evolution of the social child. In: Origins of the social mind: Evolutionary psychology and child development, ed. B. J. Ellis & D. F. Bjorklund, pp. 19 – 44. Guilford Press. [aMDG] Fortuna, K. & Roisman, G. I. (2008) Insecurity, stress, and symptoms of psychopathology: Contrasting results from self-reports versus interviews of adult attachment. Attachment and Human Development 10:11 – 28. [rMDG] Fraley, R., Brumbaugh, C. C. & Marks, M. J. (2005) The evolution and function of adult attachment: A comparative and phylogenetic analysis. Journal of Personality and Social Psychology 89:731 – 46. [LP] Fraley, R. C. (2002) Attachment stability from infancy to adulthood: Meta-analysis and dynamic modeling of developmental mechanisms. Personality and Social Psychology Review 6:123– 51. [aMDG] Fraley, R. C. & Shaver, P. R. (1997) Adult attachment and the suppression of unwanted thoughts. Journal of Personality and Social Psychology 73:1080– 91. [VZ] Fraley, R. C. & Spieker, S. J. (2003) Are infant attachment patterns continuously or categorically distributed? A taxometric analysis of Strange Situation behavior. Developmental Psychology 39:387 – 404. [aMDG, NL] Fraley, R. C., Waller, N. G. & Brennan, K. A. (2000) An item response theory analysis of self-report measures of adult attachment. Journal of Personality and Social Psychology 78:350 – 65. [rMDG] Francis, D. D., Young, L. J., Meaney, M. J. & Insel, T. R. (2002) Naturally occurring differences in maternal care are associated with the expression of oxytocin and vasopressin (V1a) receptors: Gender differences. Journal of Neuroendocrinology 14:349 – 53. [B-BC, SGF] Frisch, R. E. (1988) Fatness and fertility. Scientific American 258(3):88 – 95. [JRH] Furstenberg, F. F. & Nord, C. W. (1985) Parenting apart: Patterns of childrearing after marital disruption. Journal of Marriage and the Family 47:893– 904. [aMDG]

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

59

References/Del Giudice: Sex, attachment, and the development of reproductive strategies Furstenberg, F. F., Peterson, J. L., Nord, C. W. & Zill, N. (1983) The life course of children of divorce: Marital disruption and marital contact. American Sociological Review 48:656 –68. [aMDG] Gangestad, S. W. & Simpson, J. A. (1990) Towards an evolutionary history of female sociosexual variation. Journal of Personality 58:69 – 96. [LB, aMDG] (2000) The evolution of human mating: Trade-offs and strategic pluralism. Behavioral and Brain Sciences 23:573 – 87; discussion 597 – 644. [aMDG, JH, JJJ, LP. LEV] Gangestad, S. W. & Thornhill, R. (1997) The evolutionary psychology of extrapair sex: The role of fluctuating asymmetry. Evolution and Human Behavior 18:69 –88. [VZ] (2008) Human oestrus. Proceedings of the Royal Society B (Biological Sciences) 275:991 – 1000. [LEV] Geary, D. C. (1998) Male, female. The evolution of human sex differences. American Psychological Association. [aMDG] (2000) Evolution and proximal expression of human paternal investment. Psychological Bulletin 126:55 – 77. [aMDG] (2002) Sexual selection and human life history. Advances in Child Development and Behavior 30:41 –101. [aMDG] (2005a) Evolution of paternal investment. In: The evolutionary psychology handbook, ed. D. M. Buss, pp. 483 – 505. Wiley & Sons. [aMDG] (2005b) The origin of mind. American Psychological Association. [MVF] Geary, D. C. & Bjorklund, D. F. (2000) Evolutionary developmental psychology. Child Development 71:57 – 65. [MVF, aMDG] Geary, D. C. & Flinn, M. V. (2001) Evolution of human parental behavior and the human family. Parenting: Science and Practice 1:5 –61. [aMDG] (2002) Sex differences in behavioral and hormonal response to social threat. Psychological Review 109(4):745 – 50. [MVF] Gentzler, A. L. & Kerns, K. A. (2004) Associations between insecure attachment and sexual experiences. Personal relationships 11:249 – 65. [B-BC, arMDG] George, C. & West, M. (1999) Developmental vs. social personality models of adult attachment and mental ill health. British Journal of Medical Psychology 72:285 – 303. [aMDG] Gilissen, R., Bakermans-Kranenburg, M. J., Van IJzendoorn, M. H. & Linting, M. (2008) Electrodermal reactivity during the Trier Social Stress Test for Children: Interaction between the serotonin transporter polymorphism and children’s attachment representation. Developmental Psychobiology 50(6):615 – 25. [rMDG, MJB-K] Gimpl, G. & Fahrenholz, F. (2001) The oxytocin receptor system: Structure, function and regulation. Physiological Reviews 81:629 – 83. [rMDG] Gloger-Tippelt, G. & Konig, L. (2007) Attachment representations in 6-year-old children from one and two parent families in Germany. School Psychology International 28:313 – 30. [rMDG, MJB-K] Gluckman, P. D. & Hanson, M. A. (2006) Evolution, development and timing of puberty. Trends in Endocrinology and Metabolism 17:7 – 12. [MVF] Godbout, N., Lussier, Y. & Sabourin, S. (2006) Early abuse experiences and subsequent gender differences in couple adjustment. Violence and Victims 21:744 – 60. [rMDG] Gogtay, N., Giedd, J. N., Lusk, L., Hayashi, K. M., Greenstein, D., Vaituzis, A. C., Nugent, T. F., Herman, D. H., Clasen, L. S., Toga, A. W., Rapoport, J. L. & Thompson, P. M. (2004) Dynamic mapping of human cortical development during childhood through early adulthood. Proceedings of the National Academy of Sciences USA 101:817– 19. [MVF] Goldenson, J., Geffner, R., Foster, S. L. & Clipson, C. R. (2007) Female domestic violence offenders: Their attachment security, trauma symptoms, and personality organization. Violence and Victims 22:532 – 45. [AC, rMDG] Goldstein Ferber, S. (2008) The concept of co-regulation between neurobehavioral sub-systems: The logic interplay between excitatory and inhibitory ends. Behavioral and Brain Sciences 31(3):337 – 38. [SGF] Gottesmann, I. I. (1966) Genetic variance in an adaptive personality trait. Journal of Child Psychology and Psychiatry 7:199– 208. [aMDG] Graham, C. A. & Easterbrooks, M. A. (2000) School-aged children’s vulnerability to depressive symptomatology: The role of attachment security, maternal depressive symptomatology, and economic risk. Development and Psychopathology 12:201 – 13. [KAK] Granot, D. & Mayseless, O. (2001) Attachment security and adjustment to school in middle childhood. International Journal of Behavioral Development 25:530 – 41. [B-BC, arMDG, MJB-K] Green, J., Stanley, C., Smith, V. & Goldwyn, R. (2000) A new method of evaluating attachment representations in school-age children: The Manchester Child Attachment Story Task. Attachment and Human Development 2:48 – 70. [aMDG] Green, J. M. & Goldwyn, R. (2002) Attachment disorganization and psychopathology: New findings in attachment research and their potential implications for developmental psychopathology in childhood. Journal of Child Psychology and Psychiatry 43:835 – 46. [AJL]

60

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

Greenberg, M. T. (1999) Attachment and psychopathology in childhood. In: Handbook of Attachment: Theory, research and clinical applications, ed. J. Cassidy & P. R. Shaver. Guilford. [aMDG] Gross, M. R. (1996) Alternative reproductive strategies and tactics: Diversity within sexes. Trends in Ecology and Evolution 11:92 – 98. [aMDG, JJJ] Grossmann, K. E., Grossmann, K. & Waters, E., eds. (2005) Attachment from infancy to adulthood: The major longitudinal studies. Guilford Press. [aMDG] Grossmann, K. E., Grossmann, K. & Zimmermann, P. (1999) A wider view of attachment and exploration: stability and change during the years of immaturity. In: Handbook of Attachment: Theory, research and clinical applications, ed. J. Cassidy & P. R. Shaver. Guilford. [aMDG] Guastella, A. J., Mitchell, P. B. & Dadds, M. R. (2008) Oxytocin increases gaze to the eye region of human faces. Biological Psychiatry 63:3– 5. [rMDG] Gurven, M. & Walker, R. (2006) Energetic demand of multiple dependents and the evolution of slow human growth. Proceedings of the Royal Society of London B 273:835 –41. [aMDG] Guttmann-Steinmetz, S. & Crowell, J. A. (2006) Attachment and externalizing disorders: A developmental psychopathology perspective. Journal of the American Academy of Child and Adolescent Psychiatry 45:440– 51. [B-BC] Hagen, E. H. & Hammerstein, P. (2005) Evolutionary biology and the strategic view of ontogeny: Genetic strategies provide robustness and flexibility in the life course. Research in Human Development 2:87 – 101. [aMDG] Hamilton, C. E. (2000) Continuity and discontinuity of attachment from infancy through adolescence. Child Development 71:690 – 94. [aMDG] Hamilton, W. D. (1964) The genetical evolution of social behaviour, Parts I and II. Journal of Theoretical Biology 7:1 – 52. [aMDG, LJS] Handcock, M. S. & Morris, M. (1999) Relative distribution methods in the social sciences. Springer-Verlag. [rMDG] Harel, J. & Scher, A. (2003) Insufficient responsiveness in ambivalent mother-infant relationships: Contextual and affective aspects. Infant Behavior and Development 26:371 – 83. [aMDG] Harlow, H. & Harlow, M. K. (1965) The affectional systems. In: Behavior of nonhuman primates, vol. 2, ed. A. M. Schrier, H. F. Harlow & F. Stollnitz, pp. 287 – 334. Academic Press. [VZ] Harlow, H. F. (1964) Early social deprivation and later behavior in the monkey. In: Unfinished tasks in the behavioral sciences, ed. A. Abrams, H. H. Gurner, & J. E. P. Tomal, pp. 154 –73. Williams & Wilkins. [DJK] Harpending, H. & Cochran, G. (2002) In our genes. Proceedings of the National Academy of Sciences USA 99:10 – 12. [rMDG, LP] Harris, J. R. (1995) Where is the child’s environment? A group socialization theory of development. Psychological Review 102:458 – 89. [aMDG] (1998) The nurture assumption. Free Press. [JRH] (2005) Social behavior and personality development. In: Origins of the social mind: Evolutionary psychology and child development, ed. B. J. Ellis & D. F. Bjorklund, pp. 245 – 70. Guilford. [aMDG] (2006) No two alike: Human nature and human individuality. Norton. [JRH] Hawley, P. H. (2007) Social dominance in childhood and adolescence: Why social competence and aggression may go hand in hand. In: Aggression and adaptation: The bright side to bad behavior, ed. P. H. Hawley, T. D. Little & P. C. Rodkin, pp. 1 – 29. Erlbaum. [B-BC] Hazan, C., Campa, M. & Gur-Yaish, N. (2006) What is adult attachment? In: Dynamics of romantic love: Attachment, caregiving, and sex, ed. M. Mikulincer & G. S. Goodman, pp. 47– 70. Guilford Press. [DKS] Hazan, C. & Diamond, L. (2000) The place of attachment in human mating. Review of General Psychology 4:186 – 204. [VZ] Hazan, C. & Shaver, P. R. (1987) Romantic love conceptualized as an attachment process. Journal of Personality and Social Psychology 52:511 – 24. [aMDG, DKS, VZ] Hazan, C. & Zeifman, D. (1999) Pair bonds as attachments: Evaluating the evidence. In: Handbook of attachment: Theory, research and clinical applications, ed. J. Cassidy & P. R. Shaver. Guilford Press. [aMDG] Heath, K. & Hadley, C. (1998) Dichotomous male reproductive strategies in a polygynous human society: Mating versus parental effort. Current Anthropology 39:369 – 74. [DJK] Heckmann, M., Hartmann, F. M., Kampschulte, B., Gack, H., Bodeker, R. H., Gorther, L. & Wudy, S. A. (2005) Cortisol production rates in preterm infants in relation to growth and illness: A noninvasive prospective study using gas chromatography-mass spectrometry. Journal of Clinical Endocrinology and Metabolism 90(10):5737– 42. [SGF] Henderson, A. J. Z., Bartholomew, K. & Dutton, D. G. (1997) He loves me, he loves me not: Attachment and separation resolution of abused women. Journal of Family Violence 12:169 – 91. [aMDG] Hennebert, O., Chalbot, S., Alran, S. & Morfin, R. (2007) Dehydroepiandrosterone 7alpha-hydroxylation in human tissues: Possible interference with type 1 11beta-hydroxysteroid dehydrogenase-mediated processes. Journal of Steroid Biochemistry and Molecular Biology 104:326– 33. [MVF] Herdt, G. & McClintock, M. (2000) The magical age of 10. Archives of Sexual Behavior 29:587 –606. [aMDG]

References/Del Giudice: Sex, attachment, and the development of reproductive strategies Hesse, E. (1999) The Adult Attachment Interview: Historical and current perspectives. In: Handbook of attachment: Theory, research and clinical applications, ed. J. Cassidy & P. R. Shaver. Guilford. [aMDG, MJB-K] Hesse, E. & Main, M. (2000) Disorganized infant, child, and adult attachment: Collapse in behavioral and attentional strategies. Journal of the American Psychoanalytic Association 48:1097– 1127. [AJL] (2006) Frightening, threatening, and dissociative parental behavior in low-risk samples: Description, discussion, and interpretations. Development and Psychopathology 18:309 – 43. [aMDG] Hewlett, B. S., Lamb, M. E., Leyendecker, B. & Scho¨lmerich, A. (2000) Parental investment strategies among Aka foragers, Ngandu farmers, and Euro American urban-industrialists. In: Adaptation and human behavior: An anthropological perspective, ed. L. Cronk, N. Chagnon & W. Irons, pp. 155 – 78. Aldine De Gruyter. [aMDG] Hill, K. (1993) Life history theory and evolutionary anthropology. Evolutionary Anthropology 2:78 – 88. [aMDG] Hill, K. & Hurtado, A. M. (1996) Ache life history: The ecology and demography of a foraging people. Aldine De Gruyter. [aMDG] Hill, K. & Kaplan, K. (1988) Tradeoffs in male and female reproductive strategies among the Ache. In: Human reproductive behavior: A Darwinian perspective, ed. L. Betzig, M. Borgerhoff Mulder & P. Turke. Cambridge University Press. [aMDG] Hinde, R. A. (1982) Attachment: Some conceptual and biological issues. In: The place of attachment in human behavior, ed. C.M. Parkes & J. Stevenson-Hinde, pp. 60 – 76. Basic Books. [aMDG] Hofer, M. A. (1994) Early relationships as regulators of infant physiology and behavior. Acta Paediatrica (Suppl.) 397:9– 18. [SGF] Ho¨glund, J. & Sheldon, B. C. (1998) The cost of reproduction and sexual selection. Oikos 83:478 –83. [aMDG] Hoier, S. (2003) Father absence and age at menarche: A test of four evolutionary models. Human Nature 14:209 – 33. [CDG, LP] Hollander, J. (2008) Testing the grain-size model for the evolution of phenotypic plasticity. Evolution 62:1381– 89. [LP] Holtzworth-Munroe, A. & Stuart, G. (1994) Typologies of male batterers: Three subtypes and the differences among them. Psychological Bulletin 116:476 – 97. [AC] Ho¨nekopp, J., Rudolph, U., Beier, L., Liebert, A. & Mu¨ller, C. (2007) Physical attractiveness of face and body as indicators of physical fitness. Evolution and Human Behavior 28:106 – 11. [JH] House, J. S., Landis, K. R. & Umberson, D. (1988) Social relationships and health. Science 241:540– 45. [VZ] Houston, A. I. & McNamara, J. M. (1999) Models of adaptive behaviour. An approach based on state. Cambridge University Press. [arMDG] Hrdy, S. B. (1981) The woman that never evolved. Harvard University Press. [B-BC] (1999) Mother nature. Pantheon Books. [arMDG] (2000) The optimal number of fathers: evolution, demography, and history in the shaping of female mate preferences. Annals of the New York Academy of Sciences 907:75 – 96. [aMDG] (2005a) Comes the child before the man: Cooperative breeding and the evolution of prolonged post-weaning dependence. In: Hunter gatherer childhoods: Evolutionary, developmental and cultural perspectives, ed. B. Hewlett & M. Lamb. Aldine/Transaction. [aMDG] (2005b) Evolutionary context of human development: The cooperative breeding model. In: Attachment and bonding: A new synthesis, ed. C. S. Carter, L. Ahnert, K. E. Grossmann, S. B. Hrdy, M. E. Lamb, S.W. Porges & N. Sachser, pp. 9 – 32. MIT Press. [aMDG] Hu, P. & Meng, Z. (2003) Research on discrimination of mother-infant attachment type. Acta Psychologica Sinica 35:201 – 208. [NL] Huber, D., Veinante, P. & Stoop, R. (2005) Vasopressinand oxytocin excite distinct neuronal populations in the central amygdala. Science 308:245 –48. [SGF] Hyde, J. S. (2005) The gender similarities hypothesis. American Psychologist 60:581 – 92. [LP] Iba´n˜ez, L., Dimartino-Nardi, J., Potau, N. & Saenger, P. (2000) Premature adrenarche – normal variant or forerunner of adult disease. Endocrine Reviews 21:671 – 96. [aMDG] Impett, E. A. & Peplau, L. A. (2002) Why some women consent to unwanted sex with a dating partner: Insights from attachment theory. Psychology of Women Quarterly 26:360 – 70. [aMDG] Ingemarsson, I. (2003) Gender aspects of preterm birth. British Journal of Obstetrics and Gynaecology 110(20):34– 38. [SGF] Insel, T. R. (2000) Toward a neurobiology of attachment. Review of General Psychology 4:176– 85. [aMDG] Insel, T. R. & Young, L. J. (2001) The neurobiology of attachment. Nature Neuroscience 2:129– 36. [aMDG] Jackson, J. J. & Ellis, B. J. (submitted) The development of human reproductive strategies: toward a comprehensive model. [JJJ]

Jackson, J. J., Ellis, B. J. & Garber, J. (submitted) Sex differences in the developmental antecedents of life history strategies: Tests of an integrated theory. [JJJ] Jackson, J. J. & Kirkpatrick, L. A. (2007) The structure and measurement of human mating strategies: Towards a multidimensional model of sociosexuality. Evolution and Human Behavior 28:382 –91. [arMDG, LP] Jang, S. A., Smith, S. W. & Levine, T. R. (2002) To stay or to leave? The role of attachment styles in communication patterns and potential termination of romantic relationships following discovery of deception. Communication Monographs 69:236 – 52. [aMDG] Jezova, D., Jurankova, E., Mosnarova, A., Kriska, M. & Skultetyova, I. (1996) Neuroendocrine response during stress with relation to gender differences. Acta Neurobiologae Experimentalis 56:779 – 85. [aMDG] Jones, H. P., Karuri, S., Cronin, C. M. G., Ohlsson, A., Peliowski, A., Synnes, A., Lee, S. K. & the Canadian Neonatal Network (2005) Actuarial survival of a large Canadian cohort of preterm infants. BMC Pediatrics 5:40. [SGF] Jorm, A. F., Christensen, H., Rodgers, B., Jacomb, P. A. & Easteal, S. (2004) Association of adverse childhood experiences, age of menarche, and adult reproductive behavior: Does the androgen receptor gene play a role? American Journal of Medical Genetics, B 125:105 – 11. [aMDG] Juvonen, J., Graham, S. & Schuster, M. A. (2003) Bullying among young adolescents: The strong, the weak, and the troubled. Pediatrics 112:1231 – 37. [aMDG] Kanazawa, S. (2001) Why father absence might precipitate early menarche: The role of polygyny. Evolution and Human Behavior 22:329 – 34. [aMDG] Kaplan, H., Hill, K., Lancaster, J. & Hurtado, A. M. (2000) A theory of human life history evolution: Diet, intelligence, and longevity. Evolutionary Anthropology 9:156– 85. [aMDG] Kaplan, H. & Robson, A.J. (2002) The emergence of humans: The coevolution of intelligence and longevity with intergenerational transfers. Proceedings of the National Academy of Sciences 99:10221– 226. [aMDG] Kaplan, H. S. & Gangestad, S. W. (2005) Life history theory and evolutionary psychology. In: Handbook of evolutionary psychology, ed. D. M. Buss, pp. 68– 95. Wiley. [LB, aMDG] Kappeler, P. M. & Pereira, M. E., eds. (2003) Primate life histories and socioecology. The University of Chicago Press. [DM] Karavasilis, L., Doyle, A. B. & Markiewicz, D. (2003) Associations between parenting style and attachment to mother in middle childhood and adolescence. International Journal of Behavioral Development 27:153–64. [aMDG] Kardia, S. R., Chu, J. & Sowers, M. R. (2006) Characterizing variation in sex steroid hormone pathway genes in women of 4 races/ethnicities: The study of women’s health across the nation (SWAN). American Journal of Medicine 119:s3– s15. [rMDG] Karishma, K. K. & Herbert, J. (2002) Dehydroepiandrosterone (DHEA) stimulates neurogenesis in the hippocampus of the rat, promotes survival of newly formed neurons and prevents corticosterone-induced suppression. European Journal of Neuroscience 16:445 – 53. [MVF] Kenrick, D. T., Ackerman, J. M. & Ledlow, S. (2003) Evolutionary social psychology: Adaptive predispositions and human culture. In: Handbook of social psychology, ed. J. DeLamater, pp. 103 – 22. Kluwer Academic/Plenum Press. [JMA] Kenrick, D. T., Sundie, J. M. & Kurzban, R. (2008) Cooperation and conflict between kith, kin, and strangers: Game theory by domains. In: Foundations of evolutionary psychology, ed. C. Crawford & D. Krebs, pp. 353 – 70. Erlbaum. [JMA] Kerns, K. A. (2008) Attachment in middle childhood. In: Handbook of attachment: Theory, research, and clinical applications, ed. J. Cassidy & P. R. Shaver. Guilford. [LB, KAK] Kerns, K. A., Abraham, M. M., Schlegelmilch, A. & Morgan, T. A. (2007) Motherchild attachment in later middle childhood: Assessment approaches and associations with mood and emotion regulation. Attachment and Human Development 9:33– 53. [aMDG, MJB-K] Kerns, K. A. & Richardson, R. A., eds. (2005) Attachment in middle childhood. Guilford Press. [rMDG] Kerns, K. A., Schlegelmilch, A., Morgan, T. A. & Abraham, M. M. (2005) Assessing attachment in middle childhood. In: Attachment in middle childhood, ed. L. A. Kerns & R. A. Richardson, pp. 46– 70. Guilford Press. [aMDG, KAK] Kerns, K. A., Tomich, P. L., Aspelmeier, J. E. & Contreras, J. M. (2000) Attachment-based assessments of parent-child relationships in middle childhood. Developmental Psychology 36:614 – 26. [B-BC, aMDG] Kerns, K. A., Tomich, P. L. & Kim, P. (2006) Normative trends in perceptions of availability and utilization of attachment figures in middle childhood. Social Development 15:1– 22. [KAK] Keverne, E. B., Nevison, C. M. & Martel, F. L. (1999) Early learning and the social bond. In: The integrative neurobiology of affiliation, ed. C. S. Carter & I. I. Lederhendler, pp. 263 – 74. MIT Press. [aMDG]

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

61

References/Del Giudice: Sex, attachment, and the development of reproductive strategies Kimonides, V. G., Spillantini, M. G., Sofroniew, M. V., Fawcett, J. W. & Herbert, J. (1999) Dehydroepiandrosterone antagonizes the neurotoxic effects of corticosterone and translocation of stress-activated protein kinase 3 in hippocampal primary cultures. Neuroscience 89:429 – 36. [MVF] Kirk, K. M., Blomberg, S. P., Duffy, D. L., Heath, A. C., Owens, I. P. F. & Martin, N. G. (2001) Natural selection and quantitative genetics of life-history traits in western women: A twin study. Evolution 55:423 – 35. [aMDG] Kirkpatrick, L. A. (1998) Evolution, pair bonding, and reproductive strategies: A reconceptualization of adult attachment. In: Attachment theory and close relationships, ed. J. A. Simpson & W. S. Rholes, pp. 353 – 93. Guilford Press. [aMDG] Kirkpatrick, L. A. & Davis, K. E. (1994) Attachment style, gender and relationship stability: A longitudinal analysis. Journal of Personality and Social Psychology 66:502 – 12. [aMDG, LP, VZ] Kobak, R., Rosenthal, N., Zajac, K. & Madsen, S. D. (2007) Adolescent attachment hierarchies and the search for an adult pair bond. New Directions for Child and Adolescent Development 117:57 –72. [DKS] Kokko, H. (1997) Evolutionary stable strategies of age-dependent sexual advertisement. Behavioral Ecology and Sociobiology 41:99– 107. [aMDG] Kokko, H. & Jennions, M. (2003) It takes two to tango. Trends in Ecology and Evolution 18:103 – 104. [rMDG] (2008) Parental investment, sexual selection and sex ratios. Journal of Evolutionary Biology 21:919 – 48. [rMDG] Korte, S. M., Koolhaas, J. M., Wingfield, J. C. & McEwen, B. S. (2005) The Darwinian concept of stress: Benefits of allostasis and costs of allostatic load and the trade-offs in health and disease. Neuroscience and Biobehavioral Reviews 29:3 – 38. [aMDG] Kosfeld, M., Heinrichs, M., Zak, P. J., Fischbacher, U. & Fehr, E. (2005) Oxytocin increases trust in humans. Nature 435:673– 76. [rMDG] Kramer, K. L. (2005) Children’s help and the pace of reproduction: Cooperative breeding in humans. Evolutionary Anthropology 14:224 – 37. [aMDG] Kruger, D. J. (2008) Women’s life history attributes are associated with preferences in mating relationships. Evolutionary Psychology 6:245 – 58. [rMDG] Labrie, F., Belanger, A., Luu-The, V., Labrie, C., Simard, J., Cusan, L., Gomez, J. L. & Candas, B. (1998) DHEA and the intracrine formation of androgens and estrogens in peripheral target tissues: Its role during aging. Steroids 63: 322 – 28. [MVF] Labrie, F., Luu-The, V., Belanger, A., Lin, S. G., Simard, J., Pelletier, G. & Labrie, C. (2005) Is dehydroepiandrosterone a hormone? Journal of Endocrinology 187:169 – 96. [aMDG] Lamb, M. E., Frodi, A. M., Hwang, C. P. & Frodi, M. (1982) Varying degrees of paternal involvement in infant care: Attitudinal and behavioral correlates. In: Nontraditional families: Parenting and child development, ed. M. E. Lamb, pp. 117 – 37. Erlbaum. [aMDG] Lampert, A. & Friedman, A. (1992) Sex differences in vulnerability and maladjustment as a function of parental investment: An evolutionary approach. Social Biology 39:65– 81. [aMDG] Lancaster, J. B. (1989) Evolutionary and cross-cultural perspectives on singleparenthood. In: Sociobiology and the social sciences, ed. R. Bell & N. Bell. Texas Tech University Press. [aMDG] Landolt, M. A., Lalumiere, M. L. & Quinsey, V. L. (1995) Sex differences in intra-sex variation in human mating tactics: An evolutionary approach. Ethology and Sociobiology 16:3 – 23. [aMDG] LeBlanc, A. N. (2004) Random family: Love, drugs, trouble and coming of age in the Bronx. HarperPerennial. [AC] Leckman, J. F., Carter, C. S., Hennessy, M. B., Hrdy, S. B., Keverne, E. B., Klann-Delius, G., Schradin, C., Todt, D. & von Holst, D. (2005) Biobehavioral processes in attachment and bonding. In: Attachment and bonding: A new synthesis, ed. C.S. Carter, L. Ahnert, K. E. Grossmann, S. B. Hrdy, M. E. Lamb, S. W. Porges & N. Sachser, pp. 301 – 48. MIT Press. [aMDG] Lehnart, J. & Neyer, F. J. (2006) Should I stay or should I go? Attachment and personality in romantic relationships across young adulthood. European Journal of Personality 20:475 – 95. [LP] Lenton, A., Penke, L., Todd, P. M. & Fasolo, B. (in press) The heart does have its reasons: Social rationality in mate choice. In: Social rationality, ed. R. Hertwig, U. Hoffrage & the ABC Research Group. Oxford University Press. [LP] Lewis, M., Feiring, C. & Rosenthal, S. (2000) Attachment over time. Child Development 71:707 – 20. [aMDG] Levine, R. & Norman, K. (2001) The infant’s acquisition of culture: Early attachment reexamined in anthropological perspective. In: The psychology of cultural experience, ed. C. Moore & H. Mathews, pp. 83 –104. Cambridge University Press. [RJQ] Li, N. P. & Kenrick, D. T. (2006) Sex similarities and differences in preferences for short-term mates: What, whether, and why. Journal of Personality and Social Psychology 90:468 – 89. [JMA] Li, T. (2005) Patterns of attachment in adulthood and rearing style. Chinese Journal of Behavioral Medical Science 15:149 – 50. [NL]

62

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

(2006) Some hot topics in attachment theory: Comment on Hu & Meng. Acta Scientiarum Naturalium Universitatis Pekinensis 42:18 – 25. [NL] Li, T. & Du, S. (2005) Analysis of adult attachment style of 50 surgery patients. Chinese Journal of Clinical Psychology 13:417 – 19. [NL] Li, T., He, J., Guo, X. & Lu, X. (2006a) Adult attachment and social support of self-learning students. Chinese Journal of Behavioral Medical Science 15:1019 –20. [NL] Li, T. & Kato, K. (2006) Measuring adult attachment: validation of ECR in Chinese sample. Acta Psychologica Sinica 38:399– 406. [NL] Li, T., Li, J., Qin, H., Guo, X., Wang, X. & Liu, Z. (2008) Adult attachment and mental health in Chinese college students. Chinese Mental Health Journal 22:631 – 35. [NL] Li, T., Li, N. & Li, M. (2006b) Correlation of adult attachment with social support and subjective well-being. Chinese Journal of Clinical Rehabilitation 10:47– 49. [NL] Li, T., Li, N. & Zhu, Y. (2007) Adult attachment and subjective well-being of Chinese college students. Chinese Journal of Behavioral Medical Science 16:54– 56. [NL] Li, T., Wang, X. & Guo, X. (2006c) Attachment style and dyadic heterosexual interaction behavior among junior high school students. Chinese Journal of Behavioral Medical Science 15:644 – 46. [NL] Lieberman, M., Doyle, A. & Markiewicz, D. (1999) Developmental patterns in security of attachment to mother and father in late childhood and early adolescence: Associations with peer relations. Child Development 70:202 – 13. [aMDG] Little, A. C., Jones, B. C., Penton-Voak, I. S., Burt, D. M. & Perrett, D. I. (2002) Partnership status and the temporal context of relationships influence human female preferences for sexual dimorphism in male face shape. Proceedings of the Royal Society B (Biological Sciences) 269:1095 – 1100. [LEV] Little, A. C. & Mannion, H. (2006) Viewing attractive or unattractive same-sex individuals changes self-rated attractiveness and face preferences in women. Animal Behaviour 72:981 –87. [LEV] Liu, Y. (2007) An examination of three models of the relationships between parental attachment and adolescents’ social functioning and depressive symptoms. Journal of Youth and Adolescence. 37:941 – 52. [aMDG] Locke, J. L., & Bogin, B. (2006) Language and life history: A new perspective on the development and evolution of human language. Behavioral and Brain Sciences 29:259 – 80. [rMDG] Low, B. (2000) Sex, wealth, and fertility: Old rules, new environments. In: Adaptation and human behavior: An anthropological perspective, ed. L. Cronk, N. Chagnon & W. Irons, pp. 323–44. Aldine De Gruyter. [aMDG] Lynch, M. (1987) Evolution of intrafamilial interactions. Proceedings of the National Academy of Sciences USA 84:8507– 11. [aMDG] Lyons-Ruth, K., Bronfman, E. & Parsons, E. (1999) Frightened, frightening, and atypical maternal behavior and disorganized infant attachment strategies. In: Atypical patterns of infant attachment, ed. J. Vondra & D. Barnett. Monographs of the Society for Research in Child Development, No. 64, pp. 67 –96. [aMDG] Lyons-Ruth, K. & Jacobvitz, D. (1999) Attachment disorganization: Unresolved loss, relational violence, and lapses in behavioral and attentional strategies. In: Handbook of attachment: Theory, research and clinical applications, ed. J. Cassidy & P. R. Shaver, pp. 520 –54. Guilford. [aMDG, AJL] Maccoby, E. E. (1991) Different reproductive strategies in males and females. Child Development 62:676 – 81. [aMDG] Mace, R. (2000a) An adaptive model of human reproductive rate where wealth is inherited: Why people have small families. In: Adaptation and human behavior: An anthropological perspective, ed. L. Cronk, N. Chagnon & W. Irons, pp. 283 –322. Aldine De Gruyter. [aMDG] (2000b) Evolutionary ecology of human life history. Animal Behaviour 59:1 – 10. [aMDG] Mace, R. & Sear, R. (2005) Are humans cooperative breeders? In: Grandmotherhood. The evolutionary significance of the second half of female life, ed. E. Voland, A. Chasiotis & W. Schiefenhoevel, pp. 143 – 59. Rutgers University Press. [aMDG] Maestripieri, D. (2003) Attachment. In: Primate psychology, ed. D. Maestripieri, pp. 108 – 43. Harvard University Press. [DM] (2005a) Effects of early experience on female behavioural and reproductive development in rhesus macaques. Proceedings of the Royal Society of London B 272:1243 – 48. [DM] (2005b) On the importance of comparative research for the understanding of human behavior and development: A reply to Gottlieb & Lickliter (2004). Social Development 14:181 – 86. [DM] Maestripieri, D. & Roney, J. R. (2006) Evolutionary developmental psychology: Contributions from comparative research with nonhuman primates. Developmental Review 26:120 – 37. [DM]

References/Del Giudice: Sex, attachment, and the development of reproductive strategies Maestripieri, D., Roney, J. R., DeBias, N., Durante, K. M. & Spaepen, G. M. (2004) Father absence, menarche, and interest in infants among adolescent girls. Developmental Science 7:560 – 66. [DM] Main, M. (1981) Avoidance in the service of attachment: a working paper. In: Behavioral development: The Bielefeld Interdisciplinary Project, ed. K. Immelmann, G. Barlow, M. Main & L. Petrinovich, pp. 651 – 93. Cambridge University Press. [aMDG] (1990) Cross-cultural strategies of attachment and attachment organization: Recent studies, changing methodologies, and the concept of conditional strategies. Human Development 33:48– 61. [aMDG] Main, M. & Cassidy, J. (1988) Categories of response to reunion with the parent at age six: Predictable from infant attachment classification and stable over a 1-month period. Developmental Psychology 24:415 – 526. [MJB-K] Main, M. & Goldwyn, R. (1998) Adult attachment scoring and classification system. Unpublished manuscript, University of California at Berkeley. [aMDG] Main, M. & Hesse, E. (1990) Parent’s unresolved traumatic experiences are related to infant disorganized attachment status: Is frightened and/or frightening parental behavior the linking mechanism? In: Attachment in the preschool years: Theory, research, and intervention, ed. M. Greenberg, D. Cicchetti & E. M. Cummings. The University of Chicago Press. [aMDG] Main, M., Hesse, E. & Kaplan, N. (2005) Predictability of attachment behavior and representational processes at 1, 6, and 19 years of age. In: Attachment from infancy to adulthood. The major longitudinal studies, ed. K. E. Grossmann, K. Grossmann & E. Waters, pp. 245 – 304. Guilford Press. [rMDG] Main, M., Kaplan, N. & Cassidy, J. (1985) Security in infancy, childhood and adulthood: A move to the level of representation. In: Growing points in attachment theory and research, ed. I. Bretherton & E. Waters. Monographs of the Society for Research in Child Development 50(1 – 2, Serial No. 209), pp. 66 – 106. [MJB-K] Main, M. & Solomon, J. (1990) Procedures for identifying infants as disorganized/ disoriented during Ainsworth Strange Situations. In: Attachment in the pre-school years: Theory, research and intervention, ed. M. T. Greenberg, D. Cicchietti, & E. M. Cummings, pp. 121–60. Chicago University Press. [AJL] Majewska, M., Demigoren, S., Spivak, C. E. & London, E. D. (1990) The neurosteroid dehydroepiandrosterone sulfate is an allosteric antagonist of the GABAa receptor. Brain Research 526:143– 46. [MVF] Marazziti, D., Dell’Osso, B., Baroni, S., Mungai, F., Catena, M. & Rucci, P. (2006) A relationship between oxytocin and anxiety of romantic attachment. Clinical Practice and Epidemiology in Mental Health 2:28. [rMDG] Marlowe, F. (2000) Paternal investment and the human mating system. Behavioural Processes 51:45 – 61. [arMDG] (2003) The mating system of foragers in the Standard Cross-Cultural Sample. Cross-Cultural Research 37:282 –306. [arMDG] Mazur, A. & Booth, A. (1998) Testosterone and dominance in men. Behavioral and Brain Sciences 21:353 – 97. [rMDG] McCarthy, M. M. (1995) Estrogen modulation of oxytocin and its relation to behavior. In: Oxytocin: Cellular and molecular approaches in medicine and research, ed. R. Ivell & J. Russell, pp. 235 – 42. Plenum Press. [aMDG] McCarthy, M. M., McDonald, C. H., Brooks, P. J. & Goldman, D. (1996) An anxiolytic action of oxytocin is enhanced by estrogen in the mouse. Physiology and Behavior 60:1209 – 15. [aMDG] McClintock, M. & Herdt, G. (1996) Rethinking puberty: The development of sexual attraction. Current Directions in Psychological Science 5:178 – 83. [aMDG] McNamara, J. M. & Houston, A. I. (1996) State-dependent life histories. Nature 380:215 – 21. [aMDG] Meadows-Oliver, M., Lois, S. S., Swartz, M. K. & Ryan-Krause, P. (2007) Sources of stress and support and maternal resources of homeless teenage mothers. Journal of Child and Adolescent Psychiatric Nursing 20(2):116– 25. [MJK] Meaney, M. J. & Szyf, M. (2005) Environmental programming of stress responses through DNA methylation: Life at the interface between a dynamic environment and a fixed genome. Dialogues in Clinical NeuroSciences 7(2):103 – 23. [SGF] Mellen, S. L. W. (1981) The evolution of love. Freeman. [rMDG] Mendle, J., Turkheimer, E., D’Onofrio, B. M., Lynch, S. K., Emery, R. E., Slutske, W. S. & Martin, N. G. (2006) Family structure and age at menarche: a children-of-twins approach. Developmental Psychology 42:533 – 42. [JRH] Micheal, A., Jenaway, A., Paykel, E. S. & Herbert, J. (2000) Altered salivary dehydroepiandrosterone levels in major depression in adults. Biological Psychiatry 48:989 – 95. [MVF] Mikulincer, M. (2006) Attachment, caregiving, and sex within romantic relationships: A behavioral systems perspective. In: Dynamics of romantic love: Attachment, caregiving, and sex, ed. M. Mikulincer & G. S. Goodman, pp. 23 – 46. Guilford Press. [DKS] Mikulincer, M. & Goodman, G. S., eds. (2006) Dynamics of romantic love: Attachment, caregiving, and sex. Guilford Press. [DKS] Miller, E. M. (1994) Paternal provisioning versus mate seeking in human populations. Personality and Individual Differences 17: 227 – 55. [aMDG]

Miller, G. (2000) The mating mind. How sexual choice shaped the evolution of human nature. Heinemann. [rMDG] Miller, L. C., Bard, K. A., Juno, C. J. & Nadler, R. D. (1986) Behavioral responsiveness of young chimpanzees (Pan troglodytes) to a novel environment. Folia Primatologica (Basel) 47(2 – 3):128– 42. [LJS] Mock, D. & Parker, G. (1997) The evolution of sibling rivalry. Oxford University Press. [aMDG] Moffitt, T. E., Caspi, A. & Belsky, J. (1992) Childhood experience and the onset of menarche: A test of a sociobiological model. Child Development 63:47– 58. [aMDG] Morison, P. & Masten, A. S. (1991) Peer reputation in middle childhood as a predictor of adaptation in adolescence: A seven-year follow-up. Child Development 62:991 – 1007. [aMDG] Moss, E., Cyr, C. & Dubois-Comtois, K. (2004) Attachment at early school age and developmental risk: Examining family contexts and behavior problems of controlling-caregiving, controlling-punitive, and behaviorally disorganized children. Developmental Psychology 40:519 – 32. [KAK] Moss, E., Rousseau, D., Parent, S., St-Laurent, D. & Saintonge, J. (1998) Correlates of attachment at school age: Maternal reported stress, mother-child interaction, and behavior problems. Child Development 69:1390–1405. [aMDG] Moss, E., St-Laurent, D., Dubois-Comtois, K. & Cyr, C. (2005) Quality of attachment at school age. In: Attachment in middle childhood, ed. K. A. Kerns & R. A. Richardson, pp. 189 – 211. Guilford. [rMDG] Muehlenbein, M. P. & Bribiescas, R. G. (2005) Testosterone-mediated immune functions and male life histories. American Journal of Human Biology 17:527 – 58. [MVF] Muehlenbein, M. P., Richards, R. J., Campbell, B. C., Phillippi, K. M., Murchison, M. A., Myers, L. & Svec, F. (2003) Dehydroepiandrosterone-sulfate as a biomarker of senescence in male non-human primates. Experimental Gerontology 38:1077 – 85. [MVF] Nakao, T. & Kato, K. (2003) How do adult attachment scales relate to one another?: The correspondence between four category and three category scales. Kyushu University Psychological Research 4:57 – 66. [NL] Neberich, W., Lehnart, J., Penke, L. & Asendorpf, J. B. (in press) Family of origin, age at menarche, and reproductive strategies: A test of four evolutionary-developmental models. European Journal of Developmental Psychology. [LP] Nettle, D. (2006) The evolution of personality variation in humans and other animals. American Psychologist 61:622 – 31. [aMDG, JRH] (2007) Individual differences. In: Oxford handbook of evolutionary psychology, ed. R. Dunbar & L. Barrett, pp. 479 – 90. Oxford University Press. [aMDG] Nguyen, A. D. & Conley, A. J. (2008) Adrenal androgens in humans and non-human primates: Production, zonation, and regulation. In: Disorders of the human adrenal cortex, ed. C.E. Flu¨ck & W. L. Miller. Special issue of Endocrine Development 13:33– 54. [MVF] Noftle, E. E. & Shaver, P. R. (2006) Attachment dimensions and the big five personality traits: Associations and comparative ability to predict relationship quality. Journal of Research in Personality 40:179 – 208. [rMDG] O’Connor, T. G. & Byrne, J. G. (2007) Attachment measures for research and practice. Child and Adolescent Mental Health 12:187 – 92. [DKS] O’Connor, T. G. & Croft, C. M. (2001) A twin study of attachment in preschool children. Child Development 72:1501– 11. [aMDG] O’Connor, T. G., Croft, C. M. & Steele, H. (2000) The contributions of behavioural genetic studies to attachment theory. Attachment and Human Development 2:107– 22. [aMDG] Ong, K. K., Potau, N., Petry, C. J., Jones, R., Ness, A. R., ALSPAC Study Team, Honour, J. W., de Zegher, F., Ibanez, L. & Dunger, D. B. (2004) Opposing influences of prenatal and postnatal weight gain on adrenarche in normal boys and girls. Journal of Clinical Endocrinology and Metabolism 89:2647– 51. [MVF] Orcutt, H., Garcia, M. & Pickett, S. (2005) Female-perpetrated intimate partner violence and romantic attachment style in a college student sample. Violence and Victims 20:287 – 302. [AC, rMDG] Orentreich, N., Brind, J. L., Rizer, R. L. & Vogelman, J. H. (1984) Age changes and sex differences in serum dehydroepiandrosterone sulfate concentrations throughout adulthood. Journal of Clinical Endocrinology and Metabolism 59:551 – 55. [MVF] Overall, N. C. & Sibley, C. G. (2008) Attachment and attraction toward romantic partners versus relevant alternatives in daily interactions. Personality and Individual Differences 44:1126– 37. [rMDG] Palmert, M. R., Hayden, D. L., Mansfield, J., Crigler, J. F., Crowley, W. F., Chandler, D. W. & Boepple, P. A. (2001) The longitudinal study of adrenal maturation during gonadal suppression: Evidence that adrenarche is a gradual process. Journal of Clinical Endocrinology and Metabolism 86:4536– 42. [MVF, aMDG] Panksepp, J. (1998) Affective neuroscience: The foundations of human and animal emotions. Oxford University Press. [aMDG]

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

63

References/Del Giudice: Sex, attachment, and the development of reproductive strategies Pannebakker, F. (2007) Morality from infancy to middle childhood. Unpublished doctoral dissertation, Faculty of Social and Behavioral Sciences, Leiden University, The Netherlands. [rMDG, MJB-K] Parent, A. S., Teilmann, G., Juul, A., Skakkebaek, N. E., Toppari, J. & Bourgignon, J. P. (2003) The timing of normal puberty and the age limits of sexual precocity: Variations around the world, secular trends, and changes after migration. Endocrine Reviews 24:668 – 93. [rMDG] Parker, G. A., Royle, N. J. & Hartley, I. R. (2002) Intrafamilial conflict and parental investment: A synthesis. Philosophical Transactions of the Royal Society of London B 357:295– 307. [aMDG] Parker, J. G., Low, C. M., Walker, A. R. & Gamm, B. K. (2005) Friendship jealousy in young adolescents: Individual differences and links to sex, self-esteem, aggression, and social adjustment. Developmental Psychology 41:235 – 50. [AC] Pedersen, C. A., Ahnert, L., Anzenberger, G., Belsky, J., Draper, P., Fleming, A. S., Grossmann, K., Sachser, N., Sommer, S., Tietze, D. P. & Young, L. J. (2005) Beyond infant attachment: The origins of bonding in later life. In: Attachment and bonding: A new synthesis, ed. C.S. Carter, L. Ahnert, K. E. Grossmann, S. B. Hrdy, M. E. Lamb, S.W. Porges & N. Sachser, pp. 385 – 428. MIT Press. [aMDG] Pellegrini, A. D. (1995) A longitudinal study of boys’ rough-and-tumble play and dominance during early adolescence. Journal of Applied Developmental Psychology 16:77– 93. [aMDG] Pellegrini, A. D. & Bartini, M. (2001) Dominance in early adolescent boys: Affiliative and aggressive dimensions and possible functions. Merrill-Palmer Quarterly 47:142 – 63. [aMDG] Penke, L. & Asendorpf, J. B. (in press) Beyond global sociosexual orientations: A more differentiated look at sociosexuality and its effects on courtship and romantic relationships. Journal of Personality and Social Psychology. [rMDG, LP] Penke, L., Denissen, J. J. A. & Miller, G. F. (2007a) Evolution, genes, and inter-disciplinary personality research. European Journal of Personality 21:639 – 65. [LP] (2007b) The evolutionary genetics of personality. European Journal of Personality 21:549 – 87. [LP] Penke, L., Todd, P. M., Lenton, A. & Fasolo, B. (2007c) How self-assessments can guide human mating decisions. In: Mating intelligence: Sex, relationships, and the mind’s reproductive system, ed. G. Geher & G. F. Miller, pp. 37 – 75. Erlbaum. [LP] Pennington, R. & Harpending, H. (1988) Fitness and fertility among Kalahari !Kung. American Journal of Physical Anthropology 77:303 – 19. [aMDG] Penton-Voak, I. S., Jacobson, A. & Trivers, R. (2004) Populational differences in attractiveness judgements of male and female faces: Comparing British and Jamaican samples. Evolution and Human Behavior 25:355 – 70. [LEV] Penton-Voak, I. S., Perrett, D. I., Castles, D. L., Kobayashi, T., Burt, D. M., Murray, L. K. & Minamisawa, R. (1999) Menstrual cycle alters face preference. Nature 399:741 –42. [LEV] Petters, D. (2006a) Designing agents to understand infants. Unpublished PhD thesis in Cognitive Science, Department of Computer Science, University of Birmingham. Available at: http://www.cs.bham.ac.uk/ddp/petters-phd2006.pdf. [DP] (2006b) Implementing a theory of attachment: A simulation of the strange situation with autonomous agents. In: Proceedings of the Seventh International Conference on Cognitive Modelling, ed. D. Fum, F. D. Missier & A. Stocco, pp. 226 – 31. Edizioni Golardiche. Available at: http://www.cs.bham.ac.uk/ ~ddp/pettersiccm.pdf. [DP] Phillips, D. I. W. (2007) Programming of the stress response: A fundamental mechanism underlying the long-term effects of fetal environment? Journal of Internal Medicine 261:453– 60. [SGF] Pianka, E. R. (1970) On r- and K-selection. American Naturalist 104:592 – 96. [DJK] Picardi, A., Caroppo, E., Toni, A., Binetti, D. & Di Maria, G. (2005) Stability of attachment-related anxiety and avoidance and their relationships with the five-factor model and the psychobiological model of personality. Psychology and Psychotherapy: Theory, Research and Practice 78:327 – 45. [rMDG] Picardi, A., Vermigli, P., Toni, A., D’Amico, R., Bitetti, D. & Pasquini, P. (2002) Further evidence of the validity of the Italian version of the questionnaire “Experiences in Close Relationships” (ECR), a self-report instrument to assess adult attachment. Italian Journal of Psychopathology 8. (Online publication). [aMDG] Pigliucci, M. (2005) Evolution of phenotypic plasticity: Where are we going now? Trends in Ecology and Evolution 20:481 – 86. [LP] Pinker, S. (1997) How the mind works. Norton. [aMDG] Pressler, J. L. & Hepworth, J. T. (2002) A quantitative use of the NIDCAP tool: The effect of gender and race on very preterm neonate’s behavior. Clinical Nursing Research 11(1):89 –102. [SGF]

64

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

Promislow, D. & Harvey, P. (1990) Living fast and dying young: A comparative analysis of life-history variation among mammals. Journal of the Zoological Society of London 220:417– 37. [aMDG] (1991) Mortality rates and the evolution of mammal life histories. Acta Oecologica 12:119 – 37. [aMDG] Provost, M. P., Troje, N. F. & Quinsey, V. L. (2008) Short-term mating strategies and attraction to masculinity in point-light walkers. Evolution and Human Behavior 29:65 – 69. [LEV] Quinlan, R. J. (2003) Father absence, parental care, and female reproductive development. Evolution and Human Behavior 24:376 – 90. [CDG] (2006) Gender and risk in a matrifocal Caribbean community. American Anthropologist 108(3):464 – 79. [RJQ] (2007) Human parental effort and environmental risk. Proceedings of the Royal Society of London B 274:121 – 25. [arMDG] Quinlan, R. J. & Flinn, M. V. (2003) Intergenerational transmission of conjugal stability. Journal of Comparative Family Studies 34:569 – 84. [MVF] Quinlan, R. J. & Quinlan, M. B. (2007a) Evolutionary ecology of human pair-bonds. Cross-cultural tests of alternative hypotheses. Cross-Cultural Research 41:149 – 69. [aMDG] (2007b) Parenting and cultures of risk: A comparative analysis of infidelity, aggression, and witchcraft. American Anthropologist 109:164 – 79. [aMDG] Quinlan, R. J., Quinlan, M. B. & Flinn, M. V. (2003) Parental investment and age at weaning in a Caribbean village. Evolution and Human Behavior 24:1 – 16. [MVF] (2005) Local resource enhancement and sex-biased breastfeeding in a Caribbean community. Current Anthropology 46(3):411 –31. [RJQ] Quirin, M., Pruessner, J. C. & Kuhl, J. (2008) HPA system regulation and adult attachment anxiety: individual differences in reactive and awakening cortisol. Psychoneuroendocrinology. 33:581 – 90. [rMDG] Rainey, W. E., Carr, B. R., Sasano, H., Suzuki, T. & Mason, J. I. (2002) Dissecting human adrenal androgen production. Trends in Endocrinology and Metabolism 13:234 – 39. [MVF] Remer, T., Boye, K. R., Hartmann, M. F. & Wudy, S. A. (2005) Urinary markers of adrenarche: Reference values in healthy subjects, aged 3 – 18 years. Journal of Clinical Endocrinology and Metabolism 90(4):2015 – 21. [MVF] Renken, B., Egeland, B., Marvinney, D., Manglesdorf, S. & Sroufe, L. A. (1989) Early childhood antecedents of aggression and passive-withdrawal in early elementary school. Journal of Personality 57:257 – 81. [aMDG] Rodkin, P. C., Farmer, T. W., Ruth, P. & Acker, R. V. (2006) They’re cool: Social status and peer group support for aggressive boys and girls. Social Development 15:175 – 204. [aMDG] Roff, D. A. (2002) Life history evolution. Sinauer. [aMDG] Roisman, G. I., Holland, A., Fortuna, K., Fraley, C., Clausell, E. & Clarke, A. (2007) The Adult Attachment Interview and self-reports of attachment style: An empirical reapprochement. Journal of Personality and Social Psychology 92:678 – 97. [arMDG] Roney, J. R. & Maestripieri, D. (2002) The importance of comparative and phylogenetic analyses in the study of adaptation. Behavioral and Brain Sciences 25:525. [DM] Rose, A. J. & Rudolph, K. D. (2006) A review of sex differences in peer relationship processes: Potential trade-offs for the emotional and behavioral development of girls and boys. Psychological Bulletin 132:98 – 131. [AC] Roth, G. & Dicke, U. (2005) Evolution of the brain and intelligence. Trends in Cognitive Sciences 9(5):250 – 57. [MVF] Rowe, D. C. (2000a) Environmental and genetic influences on pubertal development: evolutionary life history traits? In: Genetic influences on human fertility and sexuality: recent empirical and theoretical findings, ed. J. L. Rodgers, D. C Rowe & W. B. Miller, pp. 147 –68. Kluwer. [aMDG] (2000b) Evolutionary ecology embraces early experience. Evolution and Human Behavior 21:352 –64. [JRH] Rowe, D. C., Vazsonyi, A. T. & Figueredo, A. J. (1997) Mating-effort in adolescence: A conditional or alternative strategy. Personality and Individual Differences 23:105 – 15. [aMDG] Rutter, M., Moffitt, T. E. & Caspi, A. (2006) Gene-environment interplay and psychopathology: Multiple varieties but real effects. Journal of Child Psychology and Psychiatry 47:226 – 61. [rMDG] Rutter, M., Pickles, A., Murray, R. & Eaves, L. (2001) Testing hypotheses on specific environmental causal effects on behavior. Psychological Bulletin 127:291 –324. [rMDG] Sadalla, E. I., Kenrick, D. T. & Vershure, B. (1987) Dominance and heterosexual attraction. Journal of Personality and Social Psychology 52:730 – 38. [aMDG] Sagi, A., Lamb, M. E., Lewkowicz, K. S., Shoham, R., Dvir, R. & Estes, D. (1985) Security of infant-mother, -father, and -metapelet attachment among kibbutz-reared Israeli children. Monographs of the Society for Research in Child Development 50:257 – 75. [aMDG] Sagi, A., van IJzendoorn, M. H., Aviezer, O., Donnell, F. & Mayseless, O. (1994) Sleeping out of home in a kibbutz communal arrangement: It makes a

References/Del Giudice: Sex, attachment, and the development of reproductive strategies difference for infant-mother attachment. Child Development 65:992 – 1004. [aMDG] Sanders, K. A. & Bruce, N. W. (1997) A prospective study of psychosocial stress and fertility in women. Human Reproduction 12:2324 –29. [rMDG] Santner, S. J., Albertson, B., Zhang, G., Zhang, G., Santulli, M., Wang, C., Demers, L. M., Shackleton, C. & Santen, R. J. (1998) Comparative rates of androgen production and metabolism in Caucasian and Chinese subjects. Journal of Clinical Endocrinology and Metabolism 83:2104– 109. [rMDG] Saudino, K. J. (1997) Moving beyond the heritability question: New directions in behavioral genetic studies of personality. Current Directions in Psychological Science 6:86 – 90. [JRH] Scarr, S. (1992) Developmental theories for the 1990s: Development and individual differences. Child Development 63:1 – 19. [rMDG] Schachner, D. A. & Shaver, P. R. (2002) Attachment style and human mate poaching. New Review of Social Psychology 1:122 – 29. [aMDG, VZ] (2004) Attachment dimensions and motives for sex. Personal Relationships 11:179 – 95. [VZ] Schaller, M. & Murray, D. R. (2008) Pathogens, personality and culture: Disease prevalence predicts worldwide variability in sociosexuality, extraversion, and openness to experience. Journal of Personality and Social Psychology 95:212 – 21. [LP] Scharfe, E. & Bartholomew, K. (1994) Reliability and stability of adult attachment patterns. Personal Relationships 1:23 – 43. [aMDG] Schmitt, D. P. (2005a) Is short-term mating the maladaptive result of insecure attachment? A test of competing evolutionary perspectives. Personality and Social Psychology Bulletin 31:747 – 68. [aMDG, LP] (2005b) Sociosexuality from Argentina to Zimbabwe: A 48-nation study of sex, culture, and strategies of human mating. Behavioral and Brain Sciences 28:247 – 75; discussion 275 – 311. [LB, rMDG, LP] (2008) Evolutionary perspectives on romantic attachment and culture: How ecological stressors influence dismissing orientations across genders and geographies. Cross-Cultural Research 42:220 – 47. [rMDG] Schmitt, D. P., Alcalay, L., Allensworth, M., Allik, J., Ault, L., Austers, I., Bennett, K. L., Bianchi, G., Boholst, F., Borg Cunen, M. A., Braeckman, J., Brainerd Jr., E. G., Caral, L. G. A., Caron, G., Casullo, M. M., Cunningham, M., Daibo, I., De Backer, C., De Souza, E., Diaz-Loving, R., Diniz, G., Durkin, K., Echegaray, M., Eremsoy, E., Euler, H. A., Falzon, R., Fisher, M. L., Fowler, R., Fry, D. P., Fry, S., Ghayur, M. A., Giri, V. N., Golden, D. L., Grammer, K., Grimaldi, L., Halberstadt, J., Haque, S., Hefer, E., Herrera, D., Hertel, J., Hitchell, A., Hoffman, H., Hradilekova, Z., Hudek-Kene-evi, J., Huffcutt, A., Jaafar, J., Jankauskaite, M., Kabangu-Stahel, H., Kardum, I., Khoury, B., Kwon, H., Laidra, K., Laireiter, A., Lakerveld, D., Lampart, A., Lauri, M., Lavalle´e, M., Lee, S., Leung, L.C., Locke, K. D., Locke, V., Luksik, I., Magaisa, I., Marcinkeviciene, J., Mata, A., Mata, R., McCarthy, B., Mills, M. E., Mkhize, N. J., Moreira, J., Moreira, S., Moya, M., Munyea, M., Noller, P., Olimat, H., Opre, A., Panayiotou, A., Petrovic, N., Poels, K., Popper, M., Poulimenou, M., P’yatokha, V., Raymond, M., Reips, U., Reneau, S. E., Rivera-Aragon, S., Rowatt, W. C., Ruch, W., Rus, V. S., Safir, M. P., Salas, S., Sambataro, F., Sandnabba, K. N., Schleeter, R., Schulmeyer, M. K., Schu¨tz, A., Scrimali, T., Shackelford, T. K., Sharan, M. B., Shaver, P. R., Sichona, F., Simonetti, F., Sineshaw, T., Sookdew, R., Speelman, T., Su¨mer, H. C., Su¨mer, N., Supekova, M., Szlendak, T., Taylor, R., Timmermans, B., Tooke, W., Tsaousis, I., Tungaraza, F. S. K., Turner, A., Vandermassen, G., Vanhoomissen, T., Van Overwalle, F., Van Wesenbeek, I., Vasey, P. L., Verissimo, J., Voracek, M., Wan, W. W. N., Wang, T., Weiss, P., Wijaya, A., Woertment, L., Youn, G. & Zupane`ie`, A. (2003a) Are men universally more dismissing than women? Gender differences in romantic attachment across 62 cultural regions. Personal Relationships 10:307 – 31. [arMDG, LP] Schmitt, D. P., Alcalay, L., Allik, J., Ault, L., Austers, I., Bennett, K. L., Bianchi, G., Boholst, F., Borg Cunen, M. A., Braeckman, J., Brainerd Jr., E. G., Caral, L. G. A., Caron, G., Casullo, M. M., Cunningham, M., Daibo, I., De Backer, C., De Souza, E., Diaz-Loving, R., Diniz, G., Durkin, K., Echegaray, M., Eremsoy, E., Euler, H. A., Falzon, R., Fisher, M. L., Foley, D., Fry, D. P., Fry, S., Ghayur, M. A., Golden, D. L., Grammer, K., Grimaldi, L., Halberstadt, J., Herrera, D., Hertel, J., Hoffman, H., Hradilekova, Z., Hudek-Kene-evi, J., Jaafar, J., Jankauskaite, M., Kabangu-Stahel, H., Kardum, I., Khoury, B., Kwon, H., Laidra, K., Laireiter, A., Lakerveld, D., Lampart, A., Lauri, M., Lavalle´e, M., Lee, S., Leung, L.C., Locke, K. D., Locke, V., Luksik, I., Magaisa, I., Marcinkeviciene, J., Mata, A., Mata, R., McCarthy, B., Mills, M. E., Moreira, J., Moreira, S., Moya, M., Munyea, M., Noller, P., Opre, A., Panayiotou, A., Petrovic, N., Poels, K., Popper, M., Poulimenou, M., P’yatokha, V., Raymond, M., Reips, U., Reneau, S. E., Rivera-Aragon, S., Rowatt, W. C., Ruch, W., Rus, V. S., Safir, M. P., Salas, S., Sambataro, F., Sandnabba, K. N., Schulmeyer, M. K., Schu¨tz, A., Scrimali, T., Shackelford, T. K., Shaver, P. R., Sichona, F., Simonetti, F., Sineshaw, T., Sookdew, R., Speelman, T., Spyrou, S., Su¨mer, H.C., Su¨mer, N., Supekova, M., Szlendak, T., Taylor, R., Timmermans, B., Tooke, W., Tsaousis, I., Tungaraza, F. S. K.,

Vandermassen, G., Vanhoomissen, T., Van Overwalle, F., Vanwesenbeek, I., Vasey, P. L., Verissimo, J., Voracek, M., Wan, W. W. N., Wang, T., Weiss, P., Wijaya, A., Woertment, L., Youn, G., & Zupane`ie`, A. (2003b) Universal sex differences in the desire for sexual variety: Tests from 52 nations, 6 continents, and 13 islands. Journal of Personality and Social Psychology 85:85– 104. [aMDG, LP] Schmitt, D. P., Alcalay, L., Allensworth, M., Allik, J., Ault, L., Austers, I., et al. Bennett, K. L., Bianchi, G., Boholst, F., Borg Cunen, M. A., Braeckman, J., Brainerd E. G. Jr., Caral, L. Caron, G., G. A., Casullo, M. M., Cunningham, M., Daibo, I., De Backer, C., De Souza, E., Diaz-Loving, R., Diniz, G., Durkin, K., Echegaray, M., Eremsoy, E., Euler, H. A., Falzon, R., Fisher, M. L., Foley, D., Fowler, R., Fry, D. P., Fry, S., Ghayur, M. A., Giri, V. N., Golden, D. L., Grammer, K., Grimaldi, L. Halbertandt, J., Haque, S., Herrera, D., Hertel, J., Hitchell, A., Hoffmann, H., Hooper, D., Hradilekova, Z., Hudek-Kene-Evi, J., Huffcutt, A., Jaafar, J., Jankauskaite, M., Kabangu-Stahel, H., Kardum, I., Khoury, B., Kwon, H., Laidra, K., Laireiter, A., Lakerveld, D., Lampert, A., Lauri, M., Lavalle´e, M., Lee, S., Leung, L. C., Locke, K. D., Locke, V., Luksik, I., Magaisa, I., Marcinkeviciene, D., Mata, A., Mata, R., McCarthy, B., Mills, M. E., Mkhize, N. J:, Moreira, J., Moreira, S., Moya, M., Munyae, M., Noller, P., Olimat, H., Opre, A., Panayiotou, A., Petrovic, N., Poels, K., Popper, M., Poulimenou, ., P’Yatokha, V., Raymond, M., Reips, U., Reneau, S., Rivera-Aragon, S., Rowatt, W. C., Ruch, W., Rus, V. S., Safir, M. P., Salas, S., Sambataro, F., Sandnabba, K. N., Schleeter, R., Schulmeyer, M. K., Schu¨tz, A., Scrimali, T., Shackelford, T. K., Sharan, M. B., Shaver, P. R., Sichona, F., Simonetti, F., Sineshaw, T., Sookdew, R., Speelman, T., Spyrou, S., Su¨mer, H. C., Su¨mer, N., Supekova, M., Szlendak, T., Taylor, R., Timmermans, B., Tooke, W., Tsaousis, I., Tungaraza, F. S. K., Turner, A., Vandermassen, G., Vanhoomissen, T., van Overwalle, F., Vanwesenbeeck, I., vasey, P. L., Verissimo, J., Voracek., M., Wan, W. W. N., Wang, T., Weiss, P., Wijaya, A., Woertman, L., Youn, G., Zupane`ie`, A. (2004) Patterns and universals of adult romantic attachment across 62 cultural regions. Are models of self and others pancultural constructs? Journal of Cross-Cultural Psychology 35:367 – 402. [arMDG] Schwartz, J. P., Lindley, L.D. & Buboltz Jr., W. C. (2007) Adult attachment orientations: Relation to affiliation motivation. Counselling Psychology Quarterly 20:253 – 65. [rMDG] Serbin, L. A., Powlishta, K. K. & Gulko, J. (1993) The development of sex typing in middle childhood. Monographs of the Society for Research in Child Development 64: 1 – 74. [aMDG] Shaver, P. R., Belsky, J. & Brennan, K. A. (2000) Comparing measures of adult attachment: An examination of interview and self-report methods. Personal Relationships 7:25 – 43. [aMDG] Shaver, P. R. & Mikulincer, M. (2002) Attachment-related psychodynamics. Attachment and Human Development 4:133 – 61. [aMDG] Shemmings, D. (2006) Using adult attachment theory to differentiate adult children’s internal working models of later life filial relationships. Journal of Aging Studies 20:177 – 91. [NL] Shrewsbury, V. & Wardle, J. (2008) Socioeconomic status and adiposity in childhood: A systematic review of cross-sectional studies 1990 – 2005. Obesity 16:275 – 84. [JRH] Sibley, C. G. & Overall, N. C. (2008) Modeling the hierarchical structure of attachment representations: A test of domain differentiation. Personality and Individual Differences 44:238 –49. [aMDG] Siever, L. J. (2008) Neurobiology of aggression and violence. American Journal of Psychiatry 165:429– 42. [AC] Simon, N. G. & Lu, S. (2006) Androgens and aggression. In: Biology of aggression, ed. R. J. Nelson, pp. 211 – 30. Oxford University Press. [aMDG] Simpson, J. A. (1994) Pair bonding as an organizational model for integrating research on close relationships. Psychological Inquiry 5:58 – 61. [aMDG] (1999) Attachment theory in modern evolutionary perspective. In: Handbook of attachment: Theory, research and clinical applications, ed. J. Cassidy & P. R. Shaver. Guilford. [aMDG] Simpson, J. A. & Belsky, J. (2008) Attachment theory within a modern evolutionary framework. In Handbook of attachment: Theory, research, and clinical applications, ed. J. Cassidy & P. R. Shaver. Guilford. [LB] Simpson, J. A. & Gangestad, S. W. (1991) Individual differences in sociosexuality: Evidence for convergent and discriminant validity. Journal of Personality and Social Psychology 60:870 – 83. [rMDG] Simpson, J. A. & Rholes, W. S. (2002) Fearful-avoidance, disorganization, and multiple working models: Some directions for future theory and research. Attachment and Human Development 4:223 – 29. [arMDG] Simpson, J. A., Wilson, C. L. & Winterheld, H. A. (2004) Sociosexuality and romantic relationships. In: The handbook of sexuality in close relationships, ed. J. H. Harvey, A. Wenzel & S. Sprecher. Erlbaum. [rMDG] Solomon, J. & George, C. (1999) The measurement of attachment security in infancy and childhood. In: Handbook of attachment: Theory, research and clinical applications, ed. J. Cassidy & P. R. Shaver. Guilford. [aMDG]

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

65

References/Del Giudice: Sex, attachment, and the development of reproductive strategies Spear, L. P. (2000) The adolescent brain and age-related behavioral manifestations. Neuroscience and Biobehavioral Review 24:417 – 63. [aMDG] Sroufe, L. A. (2003) Attachment categories as reflections of multiple dimensions. Comment on Fraley and Spieker (2003). Developmental Psychology 39:413 – 16. [aMDG] Sroufe, L. A., Bennett, C., Englund, M., Urban, J. & Shulman, S. (1993) The significance of gender boundaries in preadolescence: Contemporary correlates and antecedents of boundary violation and maintenance. Child Development 64:455 – 66. [aMDG, MJK] Staller, K. M. (2004) Runaway youth system dynamics: A theoretical framework for analyzing runaway and homeless youth policy. Families in Society 85(3):379 – 90. [MJK] Strous, R. D., Spivak, B., Yoran-Hegesh, R., Maayan, R., Averbuch, E., Kotler, M., Mester, R. & Weizman, A. (2001) Analysis of neurosteroid levels in attention deficit hyperactivity disorder. The International Journal of Psychopharmacology 4:259 – 64. [aMDG] Suess, G. J., Grossmann, K. E. & Sroufe, L. A. (1992) Effects of infant attachment to mother and father on quality of adaptation in preschool: From dyadic to individual organisation of self. International Journal of Behavioral Development 15:43– 65. [aMDG] Sulcova, J., Hill, M., Hampl, R. & Starka, L. (1997) Age and sex related differences in serum levels of unconjugated dehydroepiandrosterone and its sulfate in normal subjects. Journal of Endocrinology 154:57 – 62. [MVF] Suomi, S. J. (1999) Attachment in rhesus monkeys. In: Handbook of attachment: Theory, research and clinical applications, ed. J. Cassidy & P. R. Shaver. Guilford. [aMDG] Svensson, E. & Sheldon, B. C. (1998) The social context of life history evolution. Oikos 83:466 – 77. [aMDG] Szielasko, A. L., Symons, D. K. & Price, L. (2007) The role of attachment in adolescent sexual risk-taking. Paper presented at the Annual Convention of the Canadian Psychological Association, Ottawa, Ontario, 2007. [DKS] Szielasko, A. L., Symons, D. K. & Price, E. L. (under review) Attachment styles in sexual and romantic relationships predict invasive sexual behaviors and approaches. [DKS] Tanner, J. M. (1978) Fetus into man: Physical growth from conception to maturity. Harvard University Press. [MVF] Target, M., Fonagy, P. & Shmueli-Goetz, Y. (2003) Attachment representations in school-age children: The development of the Child Attachment Interview (CAI). Journal of Child Psychotherapy 29:171 – 86. [KAK] Taylor, S. E. (2006) Tend and befriend: Biobehavioral bases of affiliation under stress. Current Directions in Psychological Science 15:273 – 77. [aMDG] Taylor, S. E., Klein, L. C., Lewis, B. P., Gruenenwald, T. L., Gurung, R. A. & Updegraff, J. A. (2000) Biobehavioral responses to stress in females: Tend-and-befriend, not fight-or-flight. Psychological Review 107:411 – 29. [aMDG] Thiessen, D. (1994) Environmental tracking by females: Sexual lability. Human Nature 5:167 – 202. [JJJ] Thompson, R. A. (1999) Early attachment and later development. In: Handbook of attachment: Theory, research and clinical applications, ed. J. Cassidy & P. R. Shaver. Guilford. [aMDG] Tinbergen, N. (1963) On aims and methods in ethology. Zeitschrift fu¨r Tierpsychologie 20:410 – 33. [DJK] Tither, J. M. & Ellis, B. J. (2008) Impact of fathers on daughters’ age at menarche: A genetically- and environmentally-controlled sibling study. Developmental Psychology 44:1409 –20. [rMDG] Todd, P. M., Penke, L., Fasolo, B. & Lenton, A. P. (2007) Different cognitive processes underlie human mate choices and mate preferences. Proceedings of the National Academy of Sciences USA 104:15011 –16. [LP] Tomaszycki, M. L., Davis, J. E., Gouzoules, H. & Wallen, K. (2001) Sex differences in infant rhesus macaque separation – rejection vocalizations and effects of prenatal androgens. Hormones and Behavior 39:267 – 76. [aMDG] Tooby, J. & Cosmides, L. (2005) Conceptual foundations of evolutionary psychology. In: The handbook of evolutionary psychology, ed. D. M. Buss, pp. 5 – 67. Wiley. [LP] Torgersen, A. M., Grova, B. K. & Sommersted, R. (2007) A pilot study of attachment patterns in adult twins. Attachment and Human Development 9:127 –38. [aMDG] Toth, I., Szollosi, A., Danis, I., Green, J. & Gervai, J. (2006) Changes of attachment to the mother from infancy to childhood: Girls are less at risk? Infant Mental Health Journal 27:343. [arMDG] Townsend, J. M. (1995) Sex without emotional involvement: An evolutionary interpretation of sex differences. Archives of Sexual Behavior 24:173 – 76. [rMDG] (2005) Sex, sex differences, and the new polygyny. Behavioral and Brain Sciences 28:295 – 96. [rMDG]

66

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

Tracy, J. L., Shaver, P. R., Albino, A. W. & Cooper, M. L. (2003) Attachment styles and adolescent sexuality. In: Adolescent romance and sexual behavior: Theory, research, and practical implications, ed. P. Florsheim. Erlbaum. [rMDG, VZ] Treboux, D., Crowell, J. A. & Waters, E. (2004) When “new” meets “old”: Configurations of adult attachment representations and their implications for marital functioning. Developmental Psychology 40:295 – 314. [aMDG] Treloar, S. A. & Martin, N. G. (1990) Age at menarche as a fitness trait: Nonadditive genetic variance detected in a large twin sample. American Journal of Human Genetics 47:137– 48. [aMDG] Trivers, R. L. (1971) The evolution of reciprocal altruism. Quarterly Review of Biology 46:35– 57. [CDG] (1972) Parental investment and sexual selection. In: Sexual selection and the descent of man 1871 – 1971, ed. B. Campbell, pp. 136 – 79. Aldine. [arMDG] (1974) Parent-offspring conflict. American Zoologist 14:249 – 64. [aMDG] (1985) Social evolution. Benjamin/Cummings. [aMDG] Trivers, R. L. & Willard, D. E. (1973) Natural selection of parental ability to vary the sex ratio of offspring. Science 179:90 – 92. [AJL, RJQ] Troy, M. & Sroufe, L. A. (1987) Victimization among preschoolers: Role of attachment relationship history. Journal of the American Academy of Child and Adolescent Psychiatry 26:166– 72. [aMDG] True, M. M. (1994) Mother-infant attachment and communication among the Dogon of Mali. Unpublished doctoral thesis. University of California at Berkeley. [rMDG] Turkheimer, E. (2004) Spinach and ice cream: Why social science is so difficult. In: Behavior genetics principles: Perspectives in development, personality, and psychopathology, ed. L. F. DiLalla. American Psychological Association. [rMDG] Turner, P. J. (1991) Relations between attachment, gender, and behavior with peers in preschool. Child Development 62:1475 –88. [aMDG] Uchino, B. N., Cacioppo, J. T. & Kiecolt-Glaser, J. K. (1996) The relationship between social support and physiological processes: A review with emphasis on underlying mechanisms and implications for health. Psychological Bulletin 119:488 –531. [VZ] van Goozen, S. H. M., Matthys, W., Cohen-Kettenis, P. T., Thijssen, J. H. H. & van Engeland, H. (1998) Adrenal androgens and aggression in conduct disorder prepubertal boys and normal controls. Biological Psychiatry 43:156– 58. [aMDG] van Goozen, S. H. M., van den Ban, E., Matthys, W., Cohen-Kettenis, P. T., Thijssen, J. H.H. & van Engeland, H. (2000) Increased adrenal androgen functioning in children with oppositional defiant disorder: A comparison with psychiatric and normal controls. Journal of the American Academy of Children and Adolescent Psychiatry 39:1446– 51. [aMDG] Van IJzendoorn, M. H. (2000) The similarity of siblings’ attachments to their mother. Child Development 71:1086 – 98. [aMDG] Van IJzendoorn, M. H. & Bakermans-Kranenburg, M. J. (1996) Attachment representations in mothers, fathers, adolescents, and clinical groups: A meta-analytic search for normative data. Journal of Consulting and Clinical Psychology 64:8 – 21. [aMDG, MJB-K] (2006) DRD4 7-repeat polymorphism moderates the association between maternal unresolved loss or trauma and infant disorganization. Attachment and Human Development 8:291 –307. [rMDG] (2008) The distribution of adult attachment representations in clinical groups: Meta-analytic search for patterns of attachment. In: The Adult Attachment Interview in clinical context, ed. H. Steele & M. Steele, pp. 69– 98. Guilford. [MJB-K] (in preparation) Distributions of adult attachment representations in clinical and non-clinical groups: A meta-analysis. [MJB-K] Van IJzendoorn, M. H. & Sagi, A. (1999) Cross-cultural patterns of attachment: Universal and contextual dimensions. In: Handbook of attachment: Theory, research and clinical applications, ed. J. Cassidy & P. R. Shaver, pp. 713 –34. Guilford Press. [aMDG, JH] Van IJzendoorn, M. H., Schuengel, C. & Bakermans-Kranenburg, M. J. (1999) Disorganized attachment in early childhood: Meta-analysis of precursors, concomitants, and sequelae. Development and Psychopathology 11:225 – 49. [AJL] Van IJzendoorn, M. H., Vereijken, C. M. J. L., Bakermans-Kranenburg, M. J. & Riksen-Walraven, J. M. A. (2004) Is the Attachment Q-Sort a valid measure of attachment security in young children? Child Development 75:1188 – 1213. [MJB-K] Veith, J. L., Buck, M., Getzlaf, S., Van Dalfsen, P. & Slade, S. (1983) Exposure to men influences the occurrence of ovulation in women. Physiology and Behavior 31:313 –15. [VZ] Verschueren, K. & Marcoen, A. (1999) Representation of self and socioemotional competence in kindergartners: Differential and combined effects of attachment to mother and to father. Child Development 70:183 – 201. [aMDG, MJB-K]

References/Del Giudice: Sex, attachment, and the development of reproductive strategies (2002) Perceptions of self and relationship with parents in aggressive and nonaggressive rejected children. Journal of School Psychology 40:501 – 22. [aMDG] Viau, V. (2002) Functional cross-talk between the hypothalamic-pituitary-gonadal and -adrenal axes. Journal of Neuroendocrinology 14:506 – 13. [aMDG] Vitzthum, V. J. (2008) Evolutionary models of women’s reproductive functioning. Annual Review of Anthropology. 37:53 – 73. [rMDG] Voland, E. (1998) Evolutionary ecology of human reproduction. Annual Review of Anthropology 27:347 – 74. [aMDG] Vondra, J. I., Hommerding, K. D. & Shaw, D. S. (1999) Stability and change in infant attachment in a low-income sample. Monographs of the Society for Research in Child Development 64:119 – 44. [aMDG] Wakefield, J. C. (2005) Biological function and dysfunction. In: The handbook of evolutionary psychology, ed. D. M. Buss. Wiley. [AJL] Wallen, K. (2005) Hormonal influences on sexually differentiated behavior in nonhuman primates. Frontiers of Neuroendocrinology 26:7 – 26. [aMDG] Wan, L. & Li, T. (in press) Adult attachment and handling interpersonal conflict of employees. Chinese Journal of Clinical Rehabilitation. [NL] Wang, C., Christenson, P. & Swerdloff, R. (2007) Clinical relevance of racial and ethnic differences in sex steroids. Journal of Clinical Endocrinology and Metabolism 92:2433 – 35. [rMDG] Wang, E., Ding, Y. C., Flodman, P., Kidd, J. R., Kidd, K. K., Grady, D. L., Ryder, O. A., Spence, M. A., Swanson, J. M. & Moyzis, R. K. (2004) The genetic architecture of selection at the human dopamine receptor D4 (DRD4) gene locus. American Journal of Human Genetics 74:931 – 44. [rMDG] Wang, Y. & Ollendick, T. H. (2001) A cross-cultural and developmental analysis of self-esteem in Chinese and Western children. Clinical Child and Family Psychology Review 4:253– 71. [NL] Wasser, S. K. & Barash, D. P. (1983) Reproductive suppression among female mammals: Implications for biomedicine and sexual selection theory. The Quarterly Review of Biology 58:513 – 38. [arMDG] Wasser, S. K. & Place, N. J. (2001) Reproductive filtering and the social environment. In: Reproductive ecology and human evolution, ed. P. Ellison, pp. 137 – 57. Aldine De Gruyter. [rMDG] Waters, E. (2002) Live long and prosper. Available at: http://www.psychology. sunysb.edu/attachment/gallery/live_long/live_long.html. [DP] Waters, E., Crowell, J., Elliot, M., Corcoran, D. & Treboux, D. (2002) Bowlby’s secure base theory and the social/personality psychology of attachment styles: Work(s) in progress. Attachment and Human Development 4:230– 42. Available at: www.psychology.sunysb.edu/attachment/online/waters_ahd_ comment.pdf. [DP] Waters, E. & Beauchaine, T. P. (2003) Are there really patterns of attachment? Comment on Fraley and Spieker (2003). Developmental Psychology 39:417 – 22. [aMDG] Waters, E., Merrick, S., Treboux, D., Crowell, J. & Albersheim, L. (2000) Attachment security in infancy and early adulthood: a twenty-year longitudinal study. Child Development 71:684 – 89. [aMDG] Waters, E., Weinfield, N. S. & Hamilton, C. E. (2000) The stability of attachment security from infancy to adolescence to early adulthood: General discussion. Child Development 71:703 – 06. [aMDG] Watson, D., Klohnen, E. C., Casillas, A., Simms, E. N., Haig, J. & Berry, D. S. (2004) Match makers and deal breakers: Analyses of assortative mating in newlywed couples. Journal of Personality 72:1029– 68. [rMDG]

Weeden, J., Cohen, A. B. & Kenrick, D. T. (2008) Religious attendance as reproductive support. Evolution and Human Behavior 29:327 – 34. [JMA] Weinfield, N. S., Sroufe, L. A., Egeland, B. & Carlson, E. A. (1999) The nature of individual differences in infant-caregiver attachment. In: Handbook of attachment: Theory, research and clinical applications, ed. J. Cassidy & P. R. Shaver. Guilford. [aMDG] (2000) Attachment from infancy and adulthood in a high-risk sample: Continuity, discontinuity, and their correlates. Child Development 71:695 – 702. [aMDG] Weinstock, M. (2005) The potential influence of maternal stress hormones on development and mental health of the offspring. Brain, Behavior and Immunity 19:296 – 308. [MVF] Weisfeld, G. E. (1999) Evolutionary principles of human adolescence. Basic Books. [aMDG] Weisfeld, G. E., Bloch, S. A. & Ivers, J. W. (1983) A factor analytic study of peer-perceived dominance in adolescent boys. Adolescence 18:229 – 43. [aMDG] Weisfeld, G. E., Muczenski, Weisfeld & Omark (1987) Stability of boys’ social success among peers over an eleven-year period. In: Interpersonal relations: Family, peers, friends, ed. J. A. Meacham. Karger. [aMDG] West-Eberhard, M. J. (2003) Developmental plasticity and evolution. Oxford University Press. [MVF, arMDG, LP] Wilkins, A. S. (2002) The evolution of developmental pathways. Sinauer. [aMDG] Williams, G. C. (1966) Adaptation and natural selection. Princeton University Press. [JMA] Wilson, E. O. (1975) Sociobiology: The new synthesis. Harvard University Press. [DJK] (1998) Consilience: The unity of knowledge. Knopf. [rMDG] Wilson, M. & Daly, M. (1985) Competitiveness, risk-taking, and violence: The young male syndrome. Ethology and Sociobiology 6:59– 73. [aMDG] Winking, J., Kaplan, H., Gurven, M. & Rucas, S. (2007) Why do men marry and why do they stray? Proceedings of the Royal Society of London. B. 274:1643– 49. [aMDG] Wright, J. D. (1990) Poor people, poor health: The health status of the homeless. Journal of Social Issues 46:49– 64. [MJK] Xu, A., Xie, X., Liu, W., Xia, Y. & Liu, D. (2007) Chinese family strengths and resiliency. Marriage and Family Review 41:143 –64. [NL] Zak, P. J., Kurzban, R. & Matzner, W. T. (2005) Oxytocin is associated with human trustworthiness. Hormones and Behavior 48:522 – 27. [rMDG] Zemel, B. S. & Katz, S. H. (1986) The contribution of adrenal and gonadal androgens to the growth in height of adolescent males. American Journal of Physical Anthropology 71:459 – 66. [MVF] Ziegler, T. E., Savage A., Scheffler G. & Snowdon, C. T. (1987) The endocrinology of puberty and reproductive functioning in female cotton-top tamarins (Saguinus oedipus) under varying social conditions. Biology of Reproduction 37(3):618 – 27. [LJS] Ziv, Y., Aviezer, O., Gini, M., Sagi, A., Karie, N. K. & Korean-Karie, N. (2000) Emotional availability in the mother-infant dyad as related to the quality of infant-mother attachment relationship. Attachment and Human Development 2(2):149 – 69. [MJK]

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

67

BEHAVIORAL AND BRAIN SCIENCES (2009) 32, 69 –120 Printed in the United States of America

doi:10.1017/S0140525X09000284

Pre´cis of Bayesian Rationality: The Probabilistic Approach to Human Reasoning Mike Oaksford School of Psychology, Birkbeck College London, London, WC1E 7HX, United Kingdom [email protected] www.bbk.ac.uk/psyc/staff/academic/moaksford

Nick Chater Division of Psychology and Language Sciences, and ESRC Centre for Economic Learning and Social Evolution, University College London, London, WC1E 6BT, United Kingdom [email protected] www.psychol.ucl.ac.uk/people/profiles/chater_nick.htm

Abstract: According to Aristotle, humans are the rational animal. The borderline between rationality and irrationality is fundamental to many aspects of human life including the law, mental health, and language interpretation. But what is it to be rational? One answer, deeply embedded in the Western intellectual tradition since ancient Greece, is that rationality concerns reasoning according to the rules of logic – the formal theory that specifies the inferential connections that hold with certainty between propositions. Piaget viewed logical reasoning as defining the end-point of cognitive development; and contemporary psychology of reasoning has focussed on comparing human reasoning against logical standards. Bayesian Rationality argues that rationality is defined instead by the ability to reason about uncertainty. Although people are typically poor at numerical reasoning about probability, human thought is sensitive to subtle patterns of qualitative Bayesian, probabilistic reasoning. In Chapters 1 – 4 of Bayesian Rationality (Oaksford & Chater 2007), the case is made that cognition in general, and human everyday reasoning in particular, is best viewed as solving probabilistic, rather than logical, inference problems. In Chapters 5 – 7 the psychology of “deductive” reasoning is tackled head-on: It is argued that purportedly “logical” reasoning problems, revealing apparently irrational behaviour, are better understood from a probabilistic point of view. Data from conditional reasoning, Wason’s selection task, and syllogistic inference are captured by recasting these problems probabilistically. The probabilistic approach makes a variety of novel predictions which have been experimentally confirmed. The book considers the implications of this work, and the wider “probabilistic turn” in cognitive science and artificial intelligence, for understanding human rationality. Keywords: Bayes’ theorem, conditional inference, logic, non-monotonic reasoning, probability, rational analysis, rationality, reasoning, selection task, syllogisms

Bayesian Rationality (Oaksford & Chater 2007, hereafter BR) aims to re-evaluate forty years of empirical research in the psychology of human reasoning, and cast human rationality in a new and more positive light. Rather than viewing people as flawed logicians, we focus instead on the spectacular success of human reasoning under uncertainty. From this perspective, everyday thought involves astonishingly rich and subtle probabilistic reasoning – but probabilistic reasoning which is primarily qualitative, rather than numerical. This viewpoint leads to a radical re-evaluation of the empirical data in the psychology of reasoning. Previously baffling logical “errors” in reasoning about even the simplest statements can be understood as arising naturally from patterns of qualitative probabilistic reasoning. Why “Bayesian” rationality, rather than mere “probabilistic” rationality? The answer is that our approach draws crucially on a particular interpretation of probability, not merely on the mathematics of probability itself. # 2009 Cambridge University Press

0140-525X/09 $40.00

Probability is often taught as capturing “objective” facts about something, for example, gambling devices such as dice or cards. It is sometimes presumed to be a fact, for example, that the probability of a fair coin producing three consecutive heads is 1/8. However, in the context of cognitive science, probability refers not to objective facts about gambling devices or anything else, but rather, it describes a reasoner’s degrees of belief. Probability theory is then a calculus not for solving mathematical problems about objects in the world, but a calculus for rationally updating beliefs. This perspective is the subjective, or Bayesian view of probability. We thus argue that human rationality, and the coherence of human thought, is defined not by logic, but by probability. The Bayesian perspective on human reasoning has radical implications. It suggests that the meaning of even the most elementary natural language sentences may have been fundamentally mischaracterized: many such statements may make probabilistic rather than logical 69

Oaksford & Chater: Pre´cis of Bayesian Rationality claims. And the most elementary aspects of human reasoning may have been misunderstood – what appeared to be logically certain inferences may often instead be better understood as plausible, probabilistic reasoning. Shifting from a logical to a Bayesian perspective entirely changes our predictions concerning the patterns of reasoning that we should expect people to exhibit. And experimental work in the psychology of reasoning provides the data against which these predictions can be compared. This Pre´cis outlines the argument of BR chapter by chapter; the section numbering corresponds to the chapter numbering of the book, with occasional modifications to assist the flow of what is now a somewhat compressed argument. The first section of the book, Chapters 1 –4, outlines the theoretical background of our shift from logical to Bayesian rationality as an account of everyday human reasoning, drawing on relevant areas of psychology, philosophy, and artificial intelligence. The second section of the book, Chapters 5 –7, relates this approach to the key empirical data in the psychology of reasoning: conditional reasoning, Wason’s selection task, and syllogistic reasoning. We argue that the patterns of results observed in the empirical data consistently favour a Bayesian analysis, even for purportedly paradigmatically “logical” reasoning problems. Chapter 8 reflects on the implications of this approach. 1. Logic and the Western conception of mind Since the Greeks, the analysis of mind has been deeply entwined with logic. Indeed, the study of logical argument and the study of mind have often been viewed as overlapping substantially. One swift route to such a deep connection is to argue that minds are distinctively rational; and that rationality is partly, or perhaps even wholly, characterized by logic. That is, logical relations are viewed primarily

MIKE OAKSFORD is Professor of Psychology and Head of the School of Psychology at Birkbeck College, London. He is the author of over one hundred scientific publications in psychology, philosophy, and cognitive science, and has written or edited six books. He currently serves on the Editorial Boards of the journals Psychological Review, Memory and Cognition, and Thinking and Reasoning. He has also served as Associate Editor for the Quarterly Journal of Experimental Psychology and on the Editorial Board of the Journal of Experimental Psychology: Learning, Memory, & Cognition. His research explores the nature of reasoning, argumentation, and the influence of emotion on cognition. NICK CHATER is Professor of Cognitive and Decision Sciences at University College London. He is the author of over one hundred and fifty scientific publications in psychology, philosophy, linguistics, and cognitive science, and has written or edited eight books. He currently serves as Associate Editor for Psychological Review. He has also served as Associate Editor for Cognitive Science and on the Editorial Boards of Psychological Review and Trends in Cognitive Sciences. His research explores formal models of inference, choice, and language.

70

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

as unbreakable, inferential relations between thoughts; and a coherent, intelligible agent must respect such relations. In particular, logic aims to specify inferential relations that hold with absolute certainty: logical inference is truth preserving, that is, if the premises are true, the conclusion must also be true. But which inferences are absolutely certain? Which can be relied upon to preserve truth reliably? We may feel confident, from our knowledge of science, that, for example, all women are mortal. We might generalize from the mortality of all other living things; or note that even the most long-lived creature will succumb in the heat-death of the universe of the far future. But such considerations, however convincing, do not give the certainty of logic – they depend on contingent facts, and such facts are not themselves certain. Aristotle answered these questions by providing the first logical system: the theory of the syllogism. Syllogisms involve two premises, such as, All women are people; All people are mortal. Aristotle argued that these premises imply with absolute certainty that All women are mortal. Logical certainty is more than mere overwhelming confidence or conviction. A logical argument depends purely on its structure: thus, Aristotle noted, our logical argument put forth here is of the form All A are B; All B are C; therefore, All A are C. And this argument is valid whatever A, B, or C stand for; hence there is no appeal to contingent facts of any kind. Aristotle’s spectacular discovery was, therefore, that patterns of reliable reasoning could be obtained merely by identifying the structure of that reasoning. Logic, then, aims to provide a theory that determines which argument structures are truth-preserving, and which are not. In a very real sense, in a logical inference, if you believe the premises, you already believe the conclusion – the meaning of the conclusion is, somehow, contained in the meaning of the premises. To deny the constraints of logic would thus be incoherent, rather than merely mistaken. Thus, logic can be viewed as providing crucial constraints on the thoughts that any rational agent can entertain (Davidson 1984; Quine 1953). Aristotle’s theory of the logical structure of the syllogism proceeded by enumeration: Aristotle identified 64 forms of the syllogism, along with a systematic, though intuitive, approach to deciding which of these syllogisms had a valid conclusion, and if so, what the nature of this conclusion is. For more than two thousand years, Aristotle’s theory of the syllogism almost exhausted logical theory—and indeed, Kant considered all logical questions to have been decisively resolved by Aristotle’s account, stating: “It is remarkable also, that to the present day, it has not been able to make one step in advance, so that, to all appearance, it [i.e., logic] may be considered as completed and perfect” (Kant 1787/1961, p. 501). As we have suggested, although Aristotle’s logic is defined over patterns of verbally stated arguments (converted from everyday language into the appropriate formal structure), it is nonetheless tempting to view the primary subject matter of logic as thought itself. If the mind is viewed as constituted by rational thought, and logic captures patterns of rational thought, it seems natural to view logic as a central part of psychology. Such was Boole’s perspective, in going beyond Aristotle’s enumeration of patterns of logical argument. Boole aimed to describe the “The Laws of Thought” (Boole

Oaksford & Chater: Pre´cis of Bayesian Rationality 1854/1958); and, in doing so, provided, for the first time, explicit mathematical rules for logical reasoning. This allowed him to develop a calculus for logical reasoning, albeit limited in scope. Boole also opened up the possibility that logical reasoning could be carried out mechanistically, purely by the manipulation of logical symbols. This insight provided a partial foundation for modern computation and, by extension, cognitive science. The view that rational thought is governed by logic, which we term the logicist conception of the mind (Oaksford & Chater 1991), was adopted wholeheartedly by early cognitive theorists such as Piaget (e.g., Inhelder & Piaget 1955). Piaget viewed the pinnacle of cognitive development as attaining the “formal operational” stage, at which point the mind is capable of reasoning according to a particular formal system of logic: propositional logic. He viewed the process of cognitive development as a series of stages of enrichment of the logical apparatus of the child, enabling increasingly abstract reasoning, which is less tied to the specific sensory-motor environment. Similarly, the early foundations of cognitive science and artificial intelligence involved attempting to realize logical systems practically, by building computer programs that can explicitly derive logical proofs. Tasks such as mathematical reasoning and problem solving were then viewed as exercises in logic, as in Newell and Simon’s Logic Theorist and General Problem Solver (see Newell & Simon 1972; Newell et al. 1958). Moreover, Chomsky’s (1957; 1965) revolutionary work in linguistics showed how the syntactic structure of language could be organized in a deductive logical system, from which all and only the grammatical sentences of the language could be generated. And in the psychology of adult reasoning, this logical conception of mind was again used as the foundation for explaining human thought. Simultaneous with the construction of the logicist program in cognition, there were some discordant and puzzling observations. Specifically, researchers such as Wason, who attempted to verify the Piagetian view of the adult mind as a perfect logic engine, found that people appeared surprisingly and systematically illogical in some experiments. Given the dissonance between these results and the emerging logicist paradigm in cognitive science, these results were largely set aside by mainstream cognitive theorists, perhaps to be returned to once the logicist approach had reached a more developed state. But the general form that an account of apparent irrationality might take was that all illogical performance resulted from misunderstandings and from the faulty way in which the mind might sometimes apply logical rules. For example, Henle stated: “I have never found errors which could unambiguously be attributed to faulty reasoning” (Henle 1978, p. xviii). But the central notion that thought is based on logic was to be retained. This fundamental commitment to logic as a foundation for thought is embodied in contemporary reasoning theory in two of the main theoretical accounts of human reasoning. The mental logic view (Braine 1978; Rips 1983; 1994) assumes that human reasoning involves logical calculation over symbolic representations, using systems of proof which are very similar to those developed by Hilbert in mathematics, and used in computer programs for theorem-proving in artificial intelligence and computer

science. By contrast, the mental models view (JohnsonLaird 1983; Johnson-Laird & Byrne 1991) takes its starting point as the denial of the assumption that reasoning involves formal operations over logical formulae, and instead assumes that people reason over concrete representations of situations or “models” in which the formulae are true. This provides a different method of proof (see Oaksford & Chater 1991; 1998a, for discussion), but one that can achieve logical performance by an indirect route. Although mental logic and mental models both give logic a central role in human reasoning, they explain apparent irrationalities in different ways. For example, mental logics may explain errors in terms of the accessibility of different rules, whereas mental models explain errors in terms of limitations in how mental models are constructed and checked, and how many models must be considered. These logicist reactions to data appearing to show human irrationality seem entirely reasonable. Every new theory in science could be immediately refuted if the mere existence of data apparently inconsistent with the theory were assumed to falsify it decisively (Kuhn 1962; Lakatos 1970). The crucial question is: Can a more plausible explanation of these puzzling aspects of human reasoning be provided? We argue that the Bayesian approach provides precisely such an alternative. 2. Rationality and rational analysis BR aims to promote a Bayesian, rather than a logical, perspective on human reasoning. But to make sense of any debate between the logical and Bayesian standpoints, we need first to clarify how we interpret the relationship between a normative mathematical theory of reasoning (whether logic or probability), and empirical findings about human reasoning. In particular, how do we deal with any systematic clashes between the theory’s dictates concerning how people ought to reason, and empirical observations of how they actually do reason? Various viewpoints have been explored. One option is to take observed human intuitions as basic, and hence as the arbiter of what counts as a good formal theory of reasoning (e.g., Cohen 1981). Another is to take the mathematical theory as basic, and view it as providing a standpoint from which to evaluate the quality of observed reasoning performance (e.g., Rips 1994). Still a further possibility is that clashes between the formal theory and actual reasoning may arise because human thought itself is divided between two systems of reasoning (e.g., Evans & Over 1996a). Here, we take a different line: We view normative theory as a component of the project of providing a “rational analysis” which aims to capture empirical data concerning thought and behavior. Rational analysis (e.g., Anderson 1990; 1991a; Oaksford & Chater 1998b) has six steps: 1. Specify precisely the goals of the cognitive system. 2. Develop a formal model of the environment to which the system is adapted. 3. Make minimal assumptions about computational limitations. 4. Derive the optimal behaviour function given steps 1– 3. BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

71

Oaksford & Chater: Pre´cis of Bayesian Rationality (This requires formal analysis using rational norms, such as probability theory, logic, or decision theory.) 5. Examine the empirical evidence to see whether the predictions of the behaviour function are confirmed. 6. Repeat, iteratively refining the theory. So the idea of rational analysis is to understand the problem that the cognitive system faces, and the environmental and processing constraints under which it operates. Behavioral predictions are derived from the assumption that the cognitive system is solving this problem, optimally (or, more plausibly, approximately), under these constraints. The core objective of rational analysis, then, is to understand the structure of the problem from the point of view of the cognitive system, that is, to understand what problem the brain is attempting to solve. In the psychology of reasoning, this point is particularly crucial. We shall see that even when the experimenter intends to confront a participant with a logical reasoning puzzle, the participant may interpret the problem in probabilistic terms. If so, the patterns of reasoning observed may be well described in a Bayesian framework, but will appear to be capriciously errorful from a logical point of view. In Chapters 5– 7 of BR, and summarized further on here, we argue that the core data in the psychology of reasoning, which has focussed on putatively “logical” reasoning tasks, can be dramatically clarified by adopting a Bayesian rational analysis. It might appear that Step 2, concerning the environment, could not be relevant to rational analysis of the reasoning, as opposed to, say, perception. Mathematical theories of reasoning are supposed to apply across topics, and hence should surely be independent of environmental structure. We shall see further on that the reverse is the case. Very general features of the environment, such as the fact that almost all natural language categories occur with a low probability and that arbitrarily chosen probabilistic constraints are often independent or nearly independent, turn out to have substantial implications for reasoning. Indeed, the project of providing a rational analysis of human reasoning gains its empirical purchase precisely by explaining how a “topic neutral” mathematical theory applies to a specific goal, given a particular set of environmental and computational constraints. Two caveats are worth entering concerning Bayesian rational analysis. The first is that rational analysis is not intended to be a theory of psychological processes. That is, it does not specify the representations or algorithms that are used to carry out this solution. Indeed, as Anderson (1990; 1991a) points out, these representations and algorithms might take many different forms – but certain general aspects of their behavior will follow irrespective of such specifics; they will arise purely because the cognitive system is well-adapted to solving this particular problem. Hence, the correct analysis of the rational structure of the cognitive problem at hand can have considerable explanatory power. The second caveat is that the aim of understanding the structure of human reasoning, whether from a logical or a Bayesian perspective, should be carefully distinguished from the goal of measuring people’s performance on logical or probabilistic problems (Evans et al. 1993; Kahneman et al. 1982). Indeed, both logic and probability provide a fresh and bracing challenge to each generation 72

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

of students; performance on logical and probability problems results from explicit instruction and study, rather than emerging from capacities that are immanent within the human mind. But this observation need not impact our evaluation of logic or probability as explanations for patterns of everyday thought. Even if the mind is a probabilistic or logical calculating engine, it may not be possible to engage that engine with verbally, symbolically, or numerically stated probabilistic or logical puzzles, which it is presumably not adapted to handle. This point is no deeper than the observation that, although the early visual processes in the retina may compute elaborate convolutions and decorrelations of the image, this does not mean that people can thereby readily apply this machinery to solve mathematics problems concerning convolution or decorrelation. Thus, empirical evidence from the psychology of reasoning is not used, in BR, to evaluate people’s logical or probabilistic reasoning competence. Rather, this evidence is used to explore the patterns of reasoning that people find natural; and to relate such patterns to how people reason outside the experimental laboratory. From the standpoint of rational analysis, the question of whether logic or probability is the appropriate framework for understanding reasoning is an empirical question: Which rational analysis of human reasoning best captures the data? In Chapters 5 – 7 of BR, we argue, case-by-case, that a Bayesian rational analysis provides a better account of core reasoning data than its logicist rivals. First, though, we consider why. In Chapter 3, we argue that real-world, informal, everyday, reasoning is almost never deductive, that is, such reasoning is almost always logically invalid. In Chapter 4, we consider what has driven the broader “probabilistic turn” in cognitive science and related fields, of which the Bayesian analysis of human reasoning is a part. 3. Reasoning in the real world: How much deduction is there? Logic provides a calculus for certain reasoning – for finding conclusions which follow, of necessity, from the premises given. But in everyday life, people are routinely forced to work with scraps of knowledge, each of which may be only partially believed. Everyday reasoning seems to be more a matter of tentative conjecture, rather than of water-tight argument. Notice, in particular, that a successful logical argument cannot be overturned by any additional information that might be added to the premises. Thus, if we know that All people are mortal, and All women are people, then we can infer, with complete certainty, that All women are mortal. Of course, on learning new information we may come to doubt the premises – but we cannot come to doubt that the conclusion follows from the premises. This property of classical logic is known as monotonicity, meaning that adding premises can never overturn existing conclusions. In reasoning about the everyday world, by contrast, nonmonotonicity is the norm: almost any conclusion can be overturned, if additional information is acquired. Thus, consider the everyday inference from It’s raining and I am about to go outside to I will get wet. This inference

Oaksford & Chater: Pre´cis of Bayesian Rationality is uncertain – indefinitely many additional premises (the rain is about to stop; I will take an umbrella; there is a covered walkway) can overturn the conclusion, even if the premises are correct. The nonmonotonicity of everyday inference is problematic for the application of logical methods to modelling thought. Nonmonotonic inferences are not logically valid and hence fall outside the scope of standard logical methods. The nonmonotonicity of everyday reasoning often strikes in subtle and unexpected ways. Most notorious is the “frame problem” (McCarthy & Hayes 1969), which arose in early applications of logical methods in artificial intelligence. Suppose an agent, with knowledge base K, makes an action A (e.g., it turns a cup upside down). Which other information in K needs to be updated to take account of this action? Intuitively, almost all other knowledge should be unchanged (e.g., that the street is empty, or that the burglar alarm is off). But, from a logical point of view, the “interia” of such everyday knowledge does not follow, because it is logically possible that A may have all manner of consequences. For example, given the additional information that the cup is valuable and placed in an alarmed glass case, then turning it over may trigger the burglar alarm and may fill the street with curious bystanders. The difficulties generated by the frame problem have had a paralyzing effect on logical approaches to planning, action, and knowledge representation in artificial intelligence. Analogous problems arise more generally (Fodor 1983; Pylyshyn 1987). Given a database with knowledge K, adding a new fact F (not necessarily concerning an action) can typically overthrow many of the previous consequences of K, in highly idiosyncratic ways. It proves to be impossible to delimit the inferential consequences of a new fact in advance. Learning a new fact about football can, for example, readily modify my beliefs about philosophy. For example, suppose one has been told footballing facts and philosophical facts by the same person, of uncertain trustworthiness. Then learning that a footballing fact is incorrect may cause one to doubt a putative philosophical fact. Thus, nonmonotonicty may apply to arbitrarily remote pieces of knowledge. And note, of course, that an inference that can be overturned by additional premises cannot be logically valid – because standard logic is monotonic by definition. Inferences which are nonmonotonic, and hence cannot be captured by conventional logic, are described in different literatures using a variety of terms: non-demonstrative inference, informal argument, and common-sense reasoning. For the purposes of our arguments, these terms are interchangeable. But their import, across psychology, artificial intelligence, and philosophy, is the same: nonmonotonic arguments are outside the scope of deductive logic. This conclusion has alarming implications for the hypothesis that thought is primarily based on logical inference. This is because the scope of monotonic inference is vanishingly small – indeed, it scarcely applies anywhere outside mathematics. As we shall see in Chapters 5 –7, this point applies even to verbally stated inferences that are typically viewed as instances of deduction. For example, consider the argument from if you put 50p in the coke machine, you will get a coke and I’ve put 50p in the coke machine, to I’ll get a coke. This argument

appears to be an instance of a canonical monotonic logical inference: modus ponens. Yet in the context of commonsense reasoning, this argument does not appear to be monotonic at all. There are innumerable possible additional factors that may block this inference (power failure, the machine is empty, the coin or the can become stuck, etc.). Thus, you can put the money in, and no can of coke may emerge. Attempting to maintain a logical analysis of this argument, these cases could be interpreted as indicating that, from a logical point of view, the conditional rule is simply false – precisely because it succumbs to counterexamples (Politzer & Braine 1991). But this is an excessively rigorous standpoint, from which almost all everyday conditionals will be discarded as false. But how could a plethora of false conditional statements provide a useful basis for thought and action. From a logical point of view, after all, we can only make inferences from true premises; a logical argument tells us nothing, if one or more of its premises are false. In sum, there appears to be a fundamental mismatch between the nonmonotonic, uncertain character of everyday reasoning, and the monotonicity of logic; and this mismatch diagnoses the fundamental problem with logic-based theories of reasoning and logicist cognitive science more broadly. In BR, we draw a parallel with a similar situation in the philosophy of science, where there has been a gradual retreat from early positive claims that theoretical claims somehow logically follow from observable premises, to Popper’s (1935/1959) limitation of logical deduction, to the process of drawing predictions from theories, to the abandonment of even this position, in the light of the nonmonotonicity of predictive inference (there are always additional forces, or factors, that can undo any prediction; Putnam 1974). Indeed, modern philosophy of science has taken a resolutely Bayesian turn (e.g., Bovens & Hartmann 2003; Earman 1992; Horwich 1982; Howson & Urbach 1993). BR also considers attempts to deal with the apparent mismatch by attempting to deal with uncertainty by developing nonmonotonic logics (e.g., Reiter 1980), a project that rapidly became mired in difficulties (see, e.g., Oaksford & Chater 1991). Perhaps it is time to shift our attention to a calculus that deals directly with uncertainty: probability theory. 4. The probabilistic turn We have seen how uncertainty, or nonmonotonicity, is a ubiquitous feature of everyday reasoning. Our beliefs, whether arising from perception, commonsense thought, or scientific analysis, are tentative and provisional. Our expectation that the car will start, that the test tube will turn blue, or that one arrow is longer than another, are continually being confounded by faulty batteries, impure chemicals, or visual illusions. Interestingly, Aristotle, the founder of logic, was keenly aware of the limits of the logical enterprise. After all, he was interested not only in mathematical and philosophical reasoning, but also with the scientific description and analysis of the everyday world, and with practical affairs and human action. An often quoted passage from the Nicomachean Ethics (1094b, Aristotle 1980, p. 3) notes that BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

73

Oaksford & Chater: Pre´cis of Bayesian Rationality “it is the mark of an educated man to look for precision in each class of things just so far as the nature of the subject admits: it is evidently equally foolish to accept probable reasoning from a mathematician and to demand from a rhetorician demonstrative reasoning.” Indeed, one key motivation for developing a theory of probability was closely connected with Aristotle’s rhetorician. The goal in rhetoric, in its traditional sense, is to provide reasoned arguments for why people should hold certain opinions concerning matters about which certainty is impossible. Thus, in deciding court cases by jury, a different piece of evidence (e.g., eye-witness testimony, forensic evidence, evidence of previous good character) must somehow be combined to yield a degree of belief concerning the likely guilt of the defendant. Here, probability is interpreted subjectively, in terms of a person’s strength of opinion, rather than concerning an assumption about the external world. Indeed, the very word “probability” initially referred to the degree to which a statement was supported by the evidence at hand (Gigerenzer et al. 1989). Jakob Bernoulli explicitly endorsed this interpretation when he entitled his definitive book Ars Conjectandi, or the Art of Conjecture (Bernoulli 1713). This subjectivist, or Bayesian, conception of probability ran through the eighteenth and into the nineteenth centuries (Daston 1988), frequently without clear distinctions being drawn between probability theory as a model of actual thought (or more usually, the thought of “rational”, rather than common, people [Hacking 1975; 1990]) or as a set of normative canons prescribing how uncertain reasoning should be conducted. As with logic, early probability theory itself was viewed as a model of mind. Over the latter part of the twentieth century, the Bayesian perspective has been increasingly influential across the cognitive sciences and related disciplines. Chapter 4 of BR surveys some of these developments. For example, if everyday inference is inherently probabilistic, this raises the possibility that natural language statements should be interpreted as making probabilistic, rather than logical, claims. So, for example, Adams (e.g., 1975; 1998) directly imports probability into logical theory, arguing that the conditional If A then B should, roughly, be interpreted as saying that B is probable, if A is true. Later we shall see how this, and other probabilistic analyses of familiar “logical” structures (e.g., concerning the quantifiers All, Some, etc.), cast new light on the empirical reasoning data. It is, we suggest, significant that three key domains in which uncertain inference is ubiquitous, philosophy of science, artificial intelligence, and cognitive psychology, have all embraced the Bayesian approach. BR reviews some of the key developments: the application of Bayes’ theorem to hypothesis confirmation (e.g., Earman 1992); the development of graphical models for knowledge representation and causal reasoning (Pearl 1988; 2000); and the application of Bayesian methods in rational models of cognitive phenomena (Chater & Oaksford 2008b; Oaksford & Chater 1998b) in areas as diverse as categorization (Anderson 1991b; Anderson & Matessa 1998), memory (Anderson & Milson 1989; Anderson & Schooler 1991), conditioning (Courville et al. 2006; Kakade & Dayan 2002), causal learning (Griffiths & Tenenbaum 2005; Novick & Cheng 2004), natural language processing 74

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

(Chater et al. 1998; Chater & Manning 2006), and vision (Knill & Richards 1996; Yuille & Kersten 2006). There has, in short, been a “probabilistic turn” across a broad range of domains – a move away from the attempt to apply logical methods to uncertain reasoning, and towards dealing with uncertainty by the application of probability theory. In Chapters 5– 7, we illustrate how the switch from logical to Bayesian rationality leads to a radical re-evaluation of the psychology of human reasoning – so radical, in fact, that even apparently paradigmatic “logical” reasoning tasks turn out to be better understood from a probabilistic point of view. 5. Does the exception prove the rule? How people reason with conditionals In Chapters 5 –7 of BR, we describe Bayesian probabilistic models for the three core areas of human reasoning research: conditional inference (Ch. 5), data selection (Ch. 6), and quantified syllogistic inference (Ch. 7). The key idea behind all these models is to use conditional probability, P(qjp), to account for the meaning of conditional statements, if p then q (e.g., if you turn the key then the car starts). The aim is to show that what appear to be “errors and biases” from a logicist standpoint are often entirely rational from a Bayesian point of view. In this Pre´cis, for each area of reasoning, we introduce the task, the standard findings, and existing logicist accounts. We then introduce a Bayesian rational analysis for each problem, show how it accounts for the core data, and provide a snapshot of some of the further data that we discuss in BR. Finally, for each area of reasoning, we summarise and describe one or two outstanding problems confronting the Bayesian approach. Chapter 5 of BR begins with conditional inference, that is, inferences directly involving the conditional if p then q. In the conditional, p is called the “antecedent” and q is called the “consequent.” Four inference patterns have been extensively studied experimentally (see Fig. 1). Each inference consists of the conditional premise and one of four possible categorical premises, which relate either to the antecedent or consequent of the conditional, or their negations ( p, : p, q, : q where “ : ” ¼ not). For example, the inference Modus Ponens (MP) combines the conditional premise if p then q with the categorical premise p; and yields the conclusion q. According to standard logic, two of these inferences are logically valid (MP and Modus Tollens [MT], see Fig. 1), and two are fallacies (Denying the Antecedent [DA] and

Figure 1. The four inference patterns investigated in the psychology of conditional inference: Modus Ponens (MP) and Modus Tollens (MT) are logically valid. Denying the Antecedent (DA) and Affirming the Consequent (AC) are logically fallacious. These inference schemata read that if the premises above the line are true then so must be the conclusion below the line. “p ) q” signifies the “material conditional” of standard logic, which is true unless p is true and q is false.

Oaksford & Chater: Pre´cis of Bayesian Rationality

Figure 2. The fits to the experimental data (Schroyens & Schaeken 2003) of standard logic (Panel A), standard logic plus the biconditional interpretation and error (Panel B), the original probabilistic model (Panel C), and the probabilistic model adjusted for rigidity violations (Panel D).

Affirming the Consequent [AC], see Fig. 1). Figure 2 (Panel A) shows the results of a meta-analysis of experiments where people are asked whether they endorse each of these four inferences (Schroyens & Schaeken 2003). Panel A also shows the predictions of the standard logical model, revealing a large divergence. From a logicist standpoint, this divergence may be reduced by assuming that some people interpret the conditional as a biconditional, that is, that if p then q also means that if q then p. This move from conditional to biconditional is, of course, logically invalid. For example, if a bird is a swan, then it is white clearly does not entail that if a bird is white, then it is a swan. Nonetheless, the biconditional interpretation may be pragmatically reasonable, in some cases. For example, promises such as if you mow the lawn, I will pay you £5 do seem to allow this pragmatic inference; it seems reasonable to assume that I will only pay you £5 if you mow the lawn (or, at least, that I will not pay you £5 if you refuse). By assuming that people make this pragmatic inference for the stimuli used in experimental tasks and by making some allowance for random error, the best fit that standard logic can provide is shown in Figure 2 (Panel B) (see Oaksford & Chater 2003a). To further close the gap with the data in Figure 2, logicist theories of conditional inference typically assume not only that people adopt the pragmatic inference to the biconditional interpretation, but also that they fail to represent logic completely in their cognitive system. For example, mental logic (e.g., Rips 1994) is typically assumed to involve an MP inference rule, but no MT rule. This means that MT inferences must be drawn in a more complex way, often leading to error. Similarly, according to mental models theory, people do not initially represent the full meaning of the conditional (JohnsonLaird & Byrne 2002). To draw an MT inference, they must “flesh out” their representations to fully capture the meaning of the conditional. In both cases, logically unwarranted pragmatic inferences and assumptions about cognitive limitations are invoked to explain the data. In contrast, the Bayesian approach only invokes probability theory. There are four key ideas behind the probabilistic account of conditional inference. First, the probability of a conditional is the conditional probability, that is, P(if p then q) ¼ P(qjp). In the normative literature, this identification is simply called “The Equation” (Adams 1998; Bennett 2003; Edgington 1995). In the psychological

literature, the Equation has been confirmed experimentally by Evans et al. (2003) and by Oberauer and Wilhelm (2003). Second, as discussed earlier, probabilities are interpreted “subjectively,” that is, as degrees of belief. It is this interpretation of probability that allows us to provide a probabilistic theory of inference as belief updating. Third, conditional probabilities are determined by a psychological process called the “Ramsey Test” (Bennett 2003; Ramsey 1931/1990b). For example, suppose you want to evaluate your conditional degree of belief that if it is sunny in Wimbledon, then John plays tennis. By the Ramsey test, you make the hypothetical supposition that it is sunny in Wimbledon and revise your other beliefs so that they fit with this supposition. You then “read off” your hypothetical degree of belief that John plays tennis from these revised beliefs. The final idea concerns standard conditional inference: how we reason when the categorical premise is not merely supposed, but is actually believed or known to be true. This process is known as conditionalization. Consider an MP inference, for example, If it is sunny in Wimbledon, then John plays tennis, and It is sunny in Wimbledon, therefore, John plays tennis. Conditionalization applies when we know (instead of merely supposing) that it is sunny in Wimbledon; or when a high degree of belief can be assigned to this event (e.g., because we know that it is sunny in nearby Bloomsbury). By conditionalization, our new degree of belief that John plays tennis should be equal to our prior degree of belief that if it is sunny in Wimbledon, then John plays tennis (here “prior” means before learning that it is sunny in Wimbledon). More formally, by the Equation, we know that P0(if it is sunny in Wimbledon, then John plays tennis) equals P0(John plays tennisjit is sunny in Wimbledon), where “P0(x)” ¼ prior probability of x. When we learn it is sunny in Wimbledon, then P1(it is sunny in Wimbledon) ¼ 1, where “P1(x)” ¼ posterior probability of x. Conditionalizing on this knowledge tells us that our new degree of belief in John plays tennis¸ P1(John plays tennis), should be equal to P0(John plays tennisjit is sunny in Wimbledon). That is, P1(q) ¼ P0(qjp), where p ¼ it is sunny in Wimbledon, and q ¼ John plays tennis.1 So from a probabilistic perspective, MP provides a way of updating our degrees of belief in the consequent, q, on learning that the antecedent, p, is true. So, quantitatively, if you believe that P0(John plays tennisjit is sunny in Wimbledon) ¼ 0.9, then given you discover that it is sunny in Wimbledon (P1(it is sunny in Wimbledon) ¼ 1) your new degree belief that John plays tennis should be 0.9, that is, P1(John plays tennis) ¼ 0.9. This contrasts with the logical approach in which believing the conditional premise entails with certainty that the conclusion is true, so that P0(John plays tennisjit is sunny in Wimbledon) ¼ 1. This is surely too strong a claim. The extension to the other conditional inferences is not direct, however. Take an example of AC, if it is sunny in Wimbledon, John plays tennis and John plays tennis, therefore, it is sunny in Wimbledon. In this case, one knows or strongly believes that John play tennis (perhaps we were told by a very reliable source), so P1(q) ¼ 1. But to use Bayesian conditionalization to infer one’s new degree of belief that it is sunny in Wimbledon, P1(p), one needs to know one’s conditional degree of belief BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

75

Oaksford & Chater: Pre´cis of Bayesian Rationality that it is sunny in Wimbledon given John plays tennis, that is, P0(pjq). However, the conditional premise of AC, like that of MP, is about P0(qjp) not about P0(pjq) (Sober 2002). The solution proposed by Oaksford et al. (2000; see also Wagner 2004) is that that people also know the prior marginal probabilities (at least approximately). That is, they know something about the probability of a sunny day in Wimbledon, P0( p), and the probability that John plays tennis, P0(q), before learning that it is in fact a sunny day in Wimbledon. With this additional information, P0(pjq) can be calculated from the converse conditional probability, P0(qjp), using Bayes’ Theorem.2 The same approach also works for DA and MT where the relevant probabilities are P0( : qj : p) and P0( : pj : q), respectively. The fact that the conditional premises of AC, DA, and MT do not determine the appropriate conditional probability marks an important asymmetry with MP. For these inferences, further knowledge is required to infer the relevant conditional degrees of belief. The rest of Chapter 5 in BR shows how the errors and biases observed in conditional inference are a consequence of this rational probabilistic model. The first set of “biases” relates directly to the data in Figure 2. These are what, in BR, we call “the inferential asymmetries.” That is, MP is drawn more than MT and AC is drawn more than DA (MT is also drawn more than AC). Figure 2, Panel C shows how well a probabilisitic account can explain these asymmetries. Here we have calculated the values of P0(qjp), P0(p), and P0(q) that best fit the data, that is, they minimize the sum of squared error between the data and the models predictions (“model” in Fig. 2). As Panel C shows, a probabilistic account can capture the asymmetries without pragmatic inference or appeal to process limitations. Panel C also shows, however, that this probabilistic model (Oaksford et al. 2000) does not capture the magnitudes of the inferential asymmetries (Evans & Over 2004; Schroyens & Schaeken 2003). It underestimates the MP – MT asymmetry and overestimates the DA – AC asymmetry. In BR, we argue that this is because learning that the categorical premise is true can have two inferential roles. The first inferential role is in conditionalization, as we have described. The second inferential role is based on the pragmatic inference that being told that the categorical premise is true often suggests that there is a counterexample to the conditional premise. For example, consider the MT inference on the rule: If I turn the key, the car starts. If you were told that the car did not start, it seems unlikely that you would immediately infer that the key was not turned. Telling someone that the car did not start seems to presuppose that an attempt has been made to start it, presumably by turning the key. Consequently, the categorical premise here seems to suggest a counterexample to the conditional itself, that is, a case where the key was turned but the car did not start. Hence, one’s degree of belief in the conditional should be reduced on being told that the car did not start. Notice, here, the contrast between being told that the car did not start (and drawing appropriate pragmatic inferences), and merely observing a car that has not started (e.g., a car parked in the driveway). In this latter situation, it is entirely natural to use the conditional rule to infer that the key has not been turned. 76

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

Where the second, pragmatic, inferential role of the categorical premise is operative, this violates what is called the rigidity condition on conditionalization, P0(qjp) ¼ P1(qjp) (Jeffrey 1983). That is, learning the categorical premise alters one’s degree of belief in the conditional premise. In BR, we argue that taking account of such rigidity violations helps capture the probability of the conditional; and that, for MT this modified probability is then used in conditionalization. Furthermore, we argue that DA and AC also suggest violations of the rigidity condition, concerning the case where the car starts without turning the key. These violations lead to reductions in our degree of belief that the cars starts, given that the key is turned (P0(qjp)). Using this lower estimate to calculate the relevant probabilities for DA, AC, and MT can rationally explain the relative magnitudes of the MP – MT and DA – AC asymmetries (see Fig. 2, Panel D). We now turn to one of the other biases of conditional inference that we explain in Chapter 5 of BR: negative conclusion bias. This bias arises when negations are used in conditional statements, for example, If a bird is a swan, then it is not red. In Evans’ (1972) Negations Paradigm, four such rules are used: If p then q; if p then not-q; if not-p then q; if not-p then not-q. The most robust finding is that people endorse DA, AC, and MT more when the conclusion contains a negation (see Fig. 3). So, for example, DA on if p then q (see Panel A in Fig. 3) yields a negated conclusion, not-q. Whereas, DA on if p then not-q (see Panel B in Fig. 3) yields an affirmative conclusion, q (because not-not-q ¼ q). In Figure 3, it is clear that the frequency with which DA is endorsed for if p then q is much higher than for if p then not-q. To explain negative conclusion bias, we appeal to the idea that most categories apply only to a minority of objects (Oaksford & Stenning 1992). Hence, the probability of an object being, say, red is lower than the probability of it not being red, that is, P0(Red) , P0( : Red). Consequently, the marginal probabilities (P0(p) and P0(q)) will take on higher values when p or q are negated. Higher values of the prior probabilities of the conclusion imply higher values of the relevant conditional probabilities for DA, AC, and MT, that is, to higher values of the posterior probability of the conclusion. So, for example, for our rule if a bird is a swan, then it is white, the prior probability of the conclusion of the DA inference (P0(:White)) is high. This means that the conditional probability (P0(:Whitej :

Figure 3. The results of Oaksford et al.’s (2000) meta-analysis of the negations paradigm conditional inference task for if p then q (Panel A), if p then : q (Panel B), if : p then q (Panel C), and if : p then : q (Panel D), showing the fit of the original probabilistic model.

Oaksford & Chater: Pre´cis of Bayesian Rationality

Figure 4. The results of Oaksford et al.’s (2000) Experiment 1 for the low P(p), low P(q) rule (Panel A), the low P( p), high P(q) rule (Panel B), the high P(p), low P(q) rule (Panel C), and the high P(p), high P(q) rule (Panel D), showing the fit of the original probabilistic model.

Swan)) is also high and, consequently, so is the probability of the conclusion (P1(:White)). Therefore, an apparently irrational negative conclusion bias can be seen as a rational “high probability conclusion” effect. Oaksford et al. (2000) tested this explanation by manipulating directly P0(p) and P0(q) rather than using negations and showed results closely analogous to negative conclusion bias (see Fig. 4). To conclude this section on conditional inference, we briefly review one of the most cited problems for a probabilistic account. Like any rational analysis, this account avoids theorising about the specific mental representations or algorithms involved in conditional reasoning. This may seem unsatisfactory. We suggest, by contrast, that it is premature to attempt an algorithmic analysis. The core of our approach interprets conditionals in terms of conditional probability, that is, using the Equation; and our current best understanding of conditional probability is given by the Ramsey test (Bennett 2003). But there is currently no possibility of building a full algorithmic model to carry through the Ramsey test, because this involves solving the notorious frame problem, discussed in Chapter 3. That is, it involves knowing how to update one’s knowledge-base, in the light of a new piece of information – and this problem has defied 40 years of artificial intelligence research. Nonetheless, an illustrative small-scale implementation of the Ramsey test is provided by the operation of a constraint satisfaction neural network (Oaksford 2004a). In such a model, performing a Ramsey test means clamping on or off the nodes or neurons corresponding to the categorical premise of a conditional inference. Network connectivity determines relevance relations and the weight matrix encodes prior knowledge. Under appropriate constraints, such a network can be interpreted as computing true posterior probabilities (McClelland 1998). A challenge for the future is to see whether such small-scale implementations can capture the full range of empirically observed effects in conditional inference.

other. They are asked which cards they should turn over, in order to test the hypothesis that if there is an A (p) on one side of a card, then there is a 2 (q) on the other. The upturned faces of the four cards show an A ( p), a K ( : p), a 2 (q), and a 7 ( : q) (see Fig. 5). The typical pattern of results is shown in Figure 6 (Panel A, Data). As Popper (1935/1959) argued, logically one can never be certain that a scientific hypothesis is true in the light of observed evidence, as the very next piece of evidence one discovers could be a counterexample. So, just because all the swans you have observed up until now have been white, is no guarantee that the next one will not be black. Instead, Popper argues that the only logically sanctioned strategy for hypothesis testing is to seek falsifying cases. In testing a conditional rule if p then q, this means seeking out p, : q cases. This means that, in the standard selection task, one should select the A (p) and the 7 ( : q) cards, because these are the only cards that could potentially falsify the hypothesis. Figure 6 (Panel A, Model) shows the logical prediction, and, as for conditional inference, the divergence from the data is large. Indeed, rather than seek falsifying evidence, participants seem to select the cases that confirm the conditional ( p and q). This is called “confirmation bias.” The range of theories of the selection task parallels the range of accounts of the conditional inference task described earlier. Mental logic theories (e.g., Rips 1994) assume that people attempt to perform conditional inferences, using the upturned face as the categorical premise to infer what is on the hidden face. Again, a biconditional interpretation is invoked: that if A then 2 may pragmatically imply if 2 then A. If people perform an MP inference on both conditionals, this will yield a confirmatory response pattern. To infer that the 7 card should be turned, involves considering the hidden face. If people consider the possibility that the hidden face is not an A, then the complex inference pattern required for MT can be applied. A problem for mental logic is that, on this explanation, selection task performance

Figure 5. The four cards in Wason’s Selection Task.

6. Being economical with the evidence: Collecting data and testing hypotheses Chapter 6 of BR presents a probabilistic model of Wason’s selection task. In this task, people see four double-sided cards, with a number on one side and a letter on the

Figure 6. The fits to the experimental data on the Wason Selection Task of standard logic (Panel A) and of the optimal data selection model (Oaksford & Chater 1994) (Panel B). BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

77

Oaksford & Chater: Pre´cis of Bayesian Rationality should look like conditional inference task performance where selecting the 2 (q) card corresponds to AC and selecting the 7 ( : q) card corresponds to MT. However, in conditional inference, MT is endorsed more than AC, but in the selection task this is reversed, that is, q (AC) is selected more than : q (MT).3 For mental models, similar predictions are made if people initially represent the conditional as a biconditional and do not “flesh out” this representation. The optimal data selection (ODS) model of this task (Oaksford & Chater 1994; 1996; 2003b) is a rational analysis derived from the normative literature on optimal experimental design in Bayesian statistics (Lindley 1956). The idea again relies on interpreting a conditional in terms of conditional probability. For example, the hypothesis, if swan (p) then white (p), is interpreted as making the claim that the probability of a bird being white given that it is a swan, P(qjp), is high, certainly higher than the base rate of being a white bird, P(q). This hypothesis is called the dependence hypothesis (HD). Bayesian hypothesis testing is comparative rather than exclusively concentrating on falsification. Specifically, in the ODS model, it is assumed that people compare HD with an independence hypothesis (HI) in which the probability of a bird being white, given it is a swan, is the same as the base rate of a bird being white, that is, P(qjp) ¼ P(q). We assume that, initially, people are maximally uncertain about which hypothesis is true (P(HD) ¼ P(HI) ¼ 0.5) and that their goal in selecting cards is to reduce this uncertainty as much as possible while turning the fewest cards. Take, for example, the card showing swan (p). This card could show white on the other side (p, q) or another color (p, : q). The probabilities of each outcome will be quite different according to the two hypotheses. For example, suppose that the probability of a bird being white, given that it is a swan is 0.9 (P(qjp, HD) ¼ 0.9) in the dependence hypothesis; the marginal probability that a bird is swan is 0.2 (P(p) ¼ 0.2); and the marginal probability that a bird is white is 0.3 (P(q) ¼ 0.3). Then, according to the dependence hypothesis, the probability of finding white (q) on the other side of the card is 0.9, whereas according to the independence hypothesis it is 0.3 (as the antecedent and consequent are, in this model, independent, we need merely consult the relevant marginal probability). And, according to the dependence hypothesis, the probability of finding a color other than white (: q) on the other side of the card is 0.1, whereas according to the independence hypothesis it is 0.7. With this information it is now possible to calculate one’s new degree of uncertainty about the dependence hypothesis after turning the swan card to find white on the other side (P(HDjp, q)). According to Bayes’ theorem (see Note 2), this probability is 0.75. Hence, one’s new degree of belief in the dependence model should be 0.75 and one’s degree of belief in the independence model should be 0.25. Hence, the degree of uncertainty about which hypothesis is true has been reduced. More specifically, the ODS model is based on information gain, where information is measured in bits as in standard communication theory. Here, the initial uncertainty is 1 bit (because P(HD) ¼ P(HI) ¼ 0.5, equivalent to the uncertainty of a single fair coin flip) and in this example this is reduced to 0.81 bits (because now P(HD) ¼ 0.75 and P(HI) ¼ 0.25). This is an information gain of 0.19 bits. 78

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

In Wason’s task, though, participants do not actually turn the cards, and hence they cannot know how much information they will gain by turning a card before doing so. Consequently, they must base their decision on expected information gain, taking both possible outcomes (p, q and p, : q) into account. The ODS model assumes that people select each card in direct proportion to its expected information gain. The ODS model also makes a key assumption about the task environment – that is, Step 2, in rational analysis: The properties that occur in the antecedents and consequents of hypotheses are almost always rare and so have a low base rate of occurrence. For example, most birds are not swans and most birds are not white. This assumption has received extensive independent verification (McKenzie et al. 2001; McKenzie & Mikkelsen 2000; 2007). The ODS model predicts that the two cards that lead to the greatest expected information gain are the p and the q cards. Figure 6 (Panel B) shows the fit of the model to the standard data (Oaksford & Chater 2003b). The value of P(qjp, HD) was set to 0.9 and the best fitting values of P(p) and P(q) were 0.22 and 0.27 respectively, that is, very close to the values used in the earlier example. The ODS model suggests that performance on the selection task displays rational hypothesis testing behavior, rather than irrational confirmation bias. Taking rarity to an extreme provides a simple intuition here. Suppose we consider the (rather implausible) conditional: If a person is bitten by a vampire bat ( p), they will develop pointed teeth (q). Clearly, we should check people who we know to have been bitten, to see if their teeth are pointed (i.e., turn the p card); and, uncontroversially, we can learn little from people we know have not been bitten (i.e., do not turn the : p card). If we see someone with pointed teeth, it is surely worth finding out whether they have been bitten – if they have, this raises our belief in the conditional, according to a Bayesian analysis (this is equivalent to turning the q card). But it seems scarcely productive to investigate someone without pointed teeth (i.e., do not turn the : q card) to see if they have been bitten. To be sure, it is possible that such a person might have been bitten, which would disconfirm our hypothesis, and lead to maximum information gain; but this has an almost infinitesimal probability. Almost certainly, we shall find that they have not been bitten, and learn nothing. Hence, with rarity, the expected informativeness of the q card is higher than that of the : q card, diverging sharply from the falsificationist perspective, but agreeing with the empirical data. It has been suggested, however, that behaviour on this task might be governed by what appears to be a wholly non-rational strategy: matching bias. This bias arises in the same context as negative conclusion bias that we discussed earlier, that is, in Evans’ (1972) negations paradigm. Take, for example, the rule if there is an A on one side, then there is not a 2 on the other side (if p then : q). The cards in this task are described using their logical status, so for this rule, 2 is the false consequent (FC) card and 7 is the true consequent (TC) card. For this negated consequent rule, participants tend to select the A card (TA: true antecedent) and the 2 card (FC). That is, participants now seem to make the falsifying response. However, as Evans and Lynch (1973) pointed out, participants may simply ignore the negations entirely and match

Oaksford & Chater: Pre´cis of Bayesian Rationality the values named in the conditional, that is, A and 2. Prima facie, this is completely irrational. However, the “contrast set” account of negation shows that because of the rarity assumption – that most categories apply to a minority of items – negated categories are high probability categories (discussed earlier). Having a high probability antecedent or consequent alters the expected information gains associated with the cards. If the probability of the consequent is high, then the ODS model predicts that people should make the falsifying TA and FC responses, because these are associated with the highest information gain. Consequently, matching bias is a rational hypothesis testing strategy after all. Probabilistic effects were first experimentally demonstrated using the reduced array version of Wason’s selection task (Oaksford et al. 1997), in which participants can successively select up to 15 q and 15 : q cards (there are no upturned p and : p cards that can be chosen). As predicted by the ODS model, where the probability of q is high (i.e., where rarity is violated), participants select more : q cards and fewer q cards. Other experiments have also revealed similar probabilistic effects (Green & Over 1997; 2000; Kirby 1994; Oaksford et al. 1999; Over & Jessop 1998). There have also been some failures to produce probabilistic effects, however (e.g., Oberauer et al. 1999; 2004). We have argued that these arise because of weak probability manipulations or other procedural problems (Oaksford & Chater 2003b; Oaksford & Moussakowski 2004; Oaksford & Wakefield 2003). We therefore introduced a natural sampling (Gigerenzer & Hoffrage 1995) procedure in which participants sample the frequencies of the card categories while performing a selection task (Oaksford & Wakefield 2003). Using this procedure, we found probabilistic effects using the same materials as Oberauer et al. (1999), where these effects were not evident. In further work on matching bias, Yama (2001) devised a crucial experiment to contrast the matching bias and the information gain accounts. He used rules that introduced a high and a low probability category, relating to the blood types Rhesus Negative (Rh –) and Positive (Rhþ). People were told that one of these categories, Rh –, was rare. Therefore, according to the ODS model and the rule if p then : Rhþ should lead participants to select the rare Rh – card. In contrast, according to matching bias they should select the Rhþ card. Yama’s (2001) data were largely consistent with the information gain model. Moreover, this finding was strongly confirmed by using the natural sampling procedure with these materials (Oaksford & Moussakowski 2004). Alternative probabilistic accounts of the selection task have also been proposed (Evans & Over 1996a; 1996b; Klauer 1999; Nickerson 1996; Over & Evans 1994; Over & Jessop 1998). Recently, Nelson (2005) directly tested the measures of information underpinning these models, including Bayesian diagnosticity (Evans & Over 1996b; McKenzie & Mikkelsen 2007; Over & Evans 1994), information gain (Hattori 2002; Oaksford & Chater 1994; 1996; 2003b), Kullback-Liebler distance (Klauer 1999; Oaksford & Chater 1996), probability gain (error minimization) (Baron 1981; 1985), and impact (absolute change) (Nickerson 1996). Using a related data selection task, he looked at a range of cases in which these norms predicted

different orderings of informativeness, for various data types. Nelson found the strongest correlations between his data and information gain (.78). Correlations with diagnosticity (– .22) and log diagnosticity (–.41) were actually negative. These results mirrored Oaksford et al.’s (1999) results in the Wason selection task. Nelson’s work provides strong convergent evidence for information gain as the index that most successfully captures people’s intuitions about the relative importance of evidence. There has been much discussion in the literature of the fact that selection task results change dramatically for conditionals that express rules of conduct, rather than putative facts, about the world (Cheng & Holyoak 1985; Manktelow & Over 1991). In such tasks, people typically do select the p and : q cards – the apparently “logical” response. One line of explanation is that reasoning is domain-specific, rather than applying across-the-board; a further claim, much discussed in evolutionary psychology, is that such tasks may tap basic mechanisms of social reasoning, such as “cheater-detection” (Cosmides 1989), which enables “correct” performance. A Bayesian rational analysis points, we suggest, in a different direction – that such deontic selection tasks (i.e., concerning norms, not facts) require a different rational analysis. In the deontic selection task, participants are given conditionals describing rules concerning how people should behave, for example, if you enter the country, you must have an inoculation against cholera. The rule is not a hypothesis under test, but a regulation that should be obeyed (Manktelow & Over 1987). Notice, crucially, that it makes no sense to confirm or disconfirm a rule concerning how people should behave: People entering the country should be inoculated, whether or not they actually are. The natural interpretation of a deontic task is for the participant to check whether the rule is being disobeyed – that is, to look for p, : q cases (people who enter the country, but are not inoculated); and indeed, in experiments, very high selections of the p and : q cards are observed. This is not because people have suddenly become Popperian falsifiers. This is because the task is no longer about attempting to gain information about whether the conditional is true or false. The conditional now concerns how people should behave, and hence can neither be confirmed nor disconfirmed by any observations of actual behavior. We adopted a decision theoretic approach to these tasks (Oaksford & Chater 1994; Perham & Oaksford 2005). Violators are people who enter the country (p) without a vaccination (: q). Thus, we assume that participants whose role it is to detect violators attach a high utility to detecting these cases, that is, U( p, : q) is high. However, every other case represents a cost, as it means wasted effort. We argue that people calculate the expected utility associated with each card. So, for example, take the case where someone does not have an inoculation ( : q). She could be either entering the country (p, : q) or not entering the country ( : p, : q). Just as in calculating expected information gain, both possible outcomess have to be taken into account in calculating expected utility (EU(x)): EU(:q) ¼ P(pj:q)U(p, :q) þ P(:pj:q)U(:p, :q)

We argue that people select cards in the deontic selection task to maximise expected utility. As only the utility of detecting a violator – someone trying to enter without an BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

79

Oaksford & Chater: Pre´cis of Bayesian Rationality inoculation – is positive, this means that only the p and the : q cards will have positive utilities (because only these cards could show a violator). This model can account for a broad range of effects on deontic reasoning, including the effects of switched rules (Cosmides 1989), perspective changes (Manktelow & Over 1991), utility manipulations (Kirby 1994), and probability manipulations (Manktelow et al. 1995). Recently, we have also applied this model to rules that contain emotional content, for example, if you clean up blood, then you must wear gloves (Perham & Oaksford 2005). With the goal of detecting cheaters (Cosmides 1989), you will look at people who are not cleaning up blood but who are wearing gloves ( : p, q). With the goal of detecting people who may come to harm, you will want to check people who are cleaning up blood but who are not wearing gloves ( p, : q). Perham and Oaksford (2005) set up contexts in which cheater detection should dominate, but in which the goal of detecting people who may come to harm may still be in play. That is, U( : p, q) . U( p, : q) . 0. The threatening word “blood” can appear for either the p, q case or the p, : q case. In calculating generalized expected utility (Zeelenberg et al. 2000), a regret term (Re) is subtracted from the expected utility of an act of detection, if the resulting state of the world is anticipated to be threatening. For example, by checking someone who is not wearing gloves ( : q), to see if they are at risk of harm, one must anticipate encountering blood (p). Because “blood” is a threatening word, the utility for the participant of turning a : q card is reduced; that is, the utility of encountering a p, : q card is now U( p, : q) – Re, for regret term Re. Consequently, selections of the “not wearing gloves” card ( : q) should be lower for our blood rule than for a rule that does not contain a threatening antecedent, such as, if you clean up packaging, then you must wear gloves. In two experiments, Perham and Oaksford (2005) observed just this effect. When participants’ primary goal was to detect cheaters, their levels of : p and q card selection were the same for the threat (blood rule) as for the nothreat rule. However, their levels of p and : q card selection were significantly lower for the threatening than for the non-threatening rules. This finding is important because it runs counter to alternative theories, in particular the evolutionary approach (Cosmides 1989; Cosmides & Tooby 2000), which makes the opposite prediction, that p and : q card selections should, if anything, increase for threat rules. Of the models considered in BR, the optimal data selection and expected utility models have been in the literature the longest, and have been subject to most comment. In the rest of Chapter 6, we respond in detail to these comments, pointing out that many can be incorporated into the evolving framework, and that some concerns miss their mark.

7. An uncertain quantity: How people reason with syllogisms Chapter 7 of BR presents a probabilistic model of quantified syllogistic reasoning. This type of reasoning relates two quantified premises. Logic defines four types of 80

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

quantified premise: All, Some, Some. . .not, and None. An example of a logically valid syllogistic argument is:

Therefore

Some Londoners (P) are soldiers (Q) All soldiers (Q) are well fed (R) Some Londoners (P) are well fed (R)

In this example, P and R are the end terms and Q is the middle term, which is common to both premises. In the premises, these terms can only appear in four possible configurations, which are called figures. When one of these terms appears before the copula verb (“are”) it is called the subject term (in the example, P and Q) and when one appears after this verb it is called the predicate term (Q and R). As the premises can appear in either order, there are 16 combinations, and as each can be in one of four figures, there are 64 different syllogisms. There are 22 logically valid syllogisms. If people are reasoning logically, they should endorse these syllogisms and reject the rest. However, observed behavior is graded, across both valid and invalid syllogisms; and some invalid syllogisms are endorsed more than some valid syllogisms. Table 1 shows the graded behaviour over the 22 logically valid syllogisms. There are natural breaks dividing the valid syllogisms into three main groups. Those above the single line are endorsed most, those below the double line are endorsed least, and those in between are endorsed at an intermediate level. Table 1. Meta-analysis of the logically valid syllogisms showing the form of the conclusion, the number of mental models (MMs) needed to reach that conclusion, and the percentage of times the valid conclusion was drawn, in each of the five experiments analyzed by Chater and Oaksford (1999b) Syllogism

Conclusion

MMs

Mean

All(Q,P), All(R,Q) All(P,Q), All(Q,R)

All All

1 1

89.87 75.32

All(Q,P), Some(R,Q) Some(Q,P), All(Q,R) All(Q,P), Some(Q,R) Some(P,Q), All(Q,R)

Some Some Some Some

1 1 1 1

86.71 87.97 88.61 86.71

No(Q,P), All(R,Q) All(P,Q), No(R,Q) No(P,Q), All(R,Q) All(P,Q), No(Q,R)

No No No No

1 1 1 1

92.41 84.81 88.61 91.14

All(P,Q), Some. . .not(R,Q) Some. . .not(P,Q), All(R,Q) All(Q,P), Some. . .not(Q,R) Some. . .not(Q,P), All(Q,R)

Some. . .not Some. . .not Some. . .not Some. . .not

2 2 2 2

67.09 56.33 66.46 68.99

Some(Q,P), No(R,Q) No(Q,P), Some(R,Q) Some(P,Q), No(R,Q) No(P,Q), Some(R,Q) Some(Q,P), No(Q,R) No(Q,P), Some(Q,R) Some(P,Q), No(Q,R) No(P,Q), Some(Q,R)

Some. . .not Some. . .not Some. . .not Some. . .not Some. . .not Some. . .not Some. . .not Some. . .not

3 3 3 3 3 3 3 3

16.46 66.46 30.38 51.90 32.91 48.10 44.30 26.56

Note. The means in the final column are weighted by sample size.

Oaksford & Chater: Pre´cis of Bayesian Rationality Alternative theories of syllogistic reasoning invoke similar processes to explain these data as for conditional inference and the selection task. However, here both mental logic and mental models have to introduce new machinery to deal with quantifiers. For mental logic (Rips 1994), this requires new logical rules for All and Some, and a guessing mechanism to account for the systematic pattern of responses for the invalid syllogisms. For mental models, dealing with quantifiers requires reinterpreting the lines of a mental model as objects described by their properties (P, Q, and R) rather than as conjunctions of propositions. For the different syllogisms different numbers of mental models are consistent with the truth of the premises. Only conclusions that are true in all of these possible models are logically valid. As Table 1 shows, for the most endorsed valid syllogisms, there is only one model consistent with the truth of the premises, and so the conclusion can be immediately read off. For the remaining valid syllogisms, more than one model needs to be constructed. If people only construct an initial model, then errors will occur. As Table 1 shows, mental models theory provides a good qualitative fit for the valid syllogisms, that is, the distinction between 1, 2, and 3 model syllogisms maps on to the key qualitative divisions in the data. The probabilistic approach to syllogisms was developed at both the computational and the algorithmic levels in the Probabilistic Heuristics Model (PHM, Chater & Oaksford 1999b). One of the primary motivations for this model was the hypothesis that, from a probabilistic point of view, reasoning about all and some might be continuous with reasoning about more transparently probabilistic quantifiers, such as most and few. By contrast, from a logical standpoint, such generalised quantifiers require a different, and far more complex, treatment (Barwise & Cooper 1981), far beyond the resources of existing logicbased accounts in psychology. Perhaps for this reason, although generalised quantifiers were discussed in early mental models theory (Johnson-Laird 1983), no empirical work on these quantifiers was carried out in the psychology of reasoning. In deriving PHM, the central first step is to assign probabilistic meanings to the central terms of quantified reasoning using conditional probability. Take the universally quantified statement, All P are Q (we use capitals to denote predicates; these should be applied to variables x, which are bound by the quantifier, e.g., P(x), but we usually leave this implicit). Intuitively, the claim that All soldiers are well fed can naturally be cast in probabilistic terms: as asserting that the probability that a person is well fed given that they are a soldier is 1. More generally, the probabilistic interpretation of All is straightforward: because its underlying logical form can be viewed as a conditional, that is, All(x)(if P(x) then Q(x)). Thus, the meaning is given as P(QjP) ¼ 1, as specifying the conditional probability of the predicate term (Q), given the subject term (P). Similar constraints can be imposed on this conditional probability to capture the meanings of the other logical quantifiers. So, Some P are Q means that P(QjP) . 0; Some P are not Q means that P(QjP) , 1; and No P are Q means that P(QjP) ¼ 0. Thus, for example, “Some Londoners are soldiers” is presumed to mean that the probability that a person is a soldier given that he or she is a Londoner is greater than zero, and similarly for the

other quantifiers. Such an account generalises smoothly to the generalised quantifiers most and few. Most P are Q means that 1 2 D , P(QjP),1 and Few P are Q means that 0 , P(QjP) , D, where D is small. So, for example, Most soldiers are well fed may be viewed as stating that the probability that a person is well fed, given that they are a soldier, is greater than, say, 0.8, but less than 1. At the level of rational analysis, these interpretations are used to build very simple graphical models (e.g., Pearl 1988) of quantified premises, to see if they impose constraints on the conclusion probability. For example, take the syllogism:

Therefore

Some P are Q All Q are R Some P are R

P!Q!R

The syllogistic premises on the left define the dependencies on the right because of their figure, that is, the arrangement of the middle term (Q) and the end terms (P and R) in the premises. There are four different arrangements or figures. The different figures lead to different dependencies, with different graphical structures. Note that these dependency models all imply that the end terms (P and R) are conditionally independent, given the middle term, because there is no arrow linking P and R, except via the middle term Q. Assuming conditional independence as a default is a further assumption about the environment (Step 2 in rational analysis). This is an assumption not made in, for example, Adams’ (1998) probability logic. These dependency models can be parameterised. Two of the parameters will always be the conditional probabilities associated with the premises. One can then deduce whether the constraints on these probabilities, implied by the earlier interpretations, impose constraints on the possible conclusion probabilities, that is, P(RjP) or P(PjR). In this example, the constraints that P(QjP) . 0, and P(RjQ) ¼ 1, and the conditional independence assumption, entail that P(RjP) . 0. Consequently, the inference to the conclusion Some P are R is probabilistically valid (p-valid). If each of the two possible conclusion probabilities, P(RjP) or P(PjR), can fall anywhere in the [0, 1] interval given the constraints on the premises, then no p-valid conclusion follows. It is then a matter of routine probability to determine which inferences are pvalid, of the 144 two premise syllogisms that arise from combining most and few and the four logical quantifiers (Chater & Oaksford 1999b). In the PHM, however, this rational analysis is also supplemented by an algorithmic account. We assume that people approximate the dictates of this rational analysis by using simple heuristics. Before introducing these heuristics, though, we introduce two key notions: the notions of the informativeness of a quantified claim, and the notion of probabilistic entailment between quantified statements. According to communication theory, a claim is informative in proportion to how surprising it is: informativeness varies inversely with probability. But what is the probability of an arbitrary quantified claim? To make sense of this idea, we begin by making a rarity assumption, as in our models of the conditional reasoning and the selection task, that is, the subject and predicate terms apply to only small subsets of objects. On this assumption, if we selected BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

81

Oaksford & Chater: Pre´cis of Bayesian Rationality subject term P, and predicate term, Q, at random, then it is very likely that they will not cross-classify any object (this is especially true, given the hierarchical character of classification; Rosch 1975). Consequently, P(QjP) ¼ 0 and so No P are Q is very likely to be true (e.g., No toupees are tables). Indeed, for any two randomly chosen subject and predicate terms it is probable that No P are Q. Such a statement is therefore quite uninformative. Some P are not Q is even more likely to be true, and hence still less informative, because the probability interval it covers includes that for No P are Q. The quantified claim least likely to be true is All P are Q, which is therefore the most informative. Overall, the quantifiers have the following order in informativeness: I(All) . I(Most) . I(Few) . I(Some) . I(None) . I(Somenot) (see Oaksford et al. 2002, for further analysis and discussion). Informativeness applies to individual quantified propositions. The second background idea, probabilistic entailment, concerns inferential relations between quantified propositions. Specifically, the use of one quantifier frequently provides evidence that another quantifier could also have been used. Thus, the claims that All swans are white is strong evidence that Some swans are white – because P(whitejswan) ¼ 1 is included in the interval P(whitejswan) . 0 (according to standard logic, this does not follow logically, as there may be no swans). Thus, we say that All probabilistically entails (or p-entails) Some. Similarly, Some and Some. . .not are mutually p-entailing because the probability intervals P(QjP) . 0 and P(QjP) , 1 overlap almost completely. With this background in place, we can now state the probabilistic heuristics model (PHM) for syllogistic reasoning. There are two types of heuristic: generate heuristics, which produce candidate conclusions, and test heuristics, which evaluate the plausibility of the candidate conclusions. The PHM account also admits the possibility that putative conclusions may also be tested by more analytic test procedures such as mental logics or mental models. The generate heuristics are: (G1) Min-heuristic: The conclusion quantifier is the same as that of the least informative premise (minpremise) (G2) P-entailments: The next most preferred conclusion quantifier will be the p-entailment of the min-conclusion (G3) Attachment-heuristic: If just one possible subject noun phrase (e.g., Some R) matches the subject noun phrase of just one premise, then the conclusion has that subject noun phrase. The two test heuristics are: (T1) Max-heuristic: Be confident in the conclusion generated by G1 – G3 in proportion to the informativeness of the most informative premise (max-premise) (T2) Some_not-heuristic: Avoid producing or accepting Some_not conclusions, because they are so uninformative. We show how the heuristics combine in the following example: All P are Q (max-premise) Some R are not Q (min-premise) Therefore Some_not (by min-heuristic) Some R are not P (by attachment-heuristic) and a further conclusion can be drawn: Some R are P [by p-entailment]

82

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

In BR, we compare the results of these heuristics with probabilistic validity, and show that where there is a p-valid conclusion, the heuristics generally identify it. For example, the idea behind the min-heuristic is to identify the most informative conclusion that validly follows from the premises. Out of the 69 p-valid syllogisms, the min-heuristic identifies that conclusion for 54; for 14 syllogisms the p-valid conclusion is less informative than the min-conclusion. There is only one violation; that is, where the p-valid conclusion is more informative than the min-conclusion. In turning to the experimental results, in BR we first show how all the major distinctions between standard syllogisms captured by other theories are also captured by PHM. So, returning to Table 1, all the syllogisms above the double line have the most informative max-premise, All (see heuristic T1). Moreover, all the syllogisms below the single line have uninformative conclusions, Somenot (see heuristic T2), and those below the double line violate the min-heuristic (heuristic G1) and require a pentailment (heuristic G2), that is, Some. . .not $ Some. Consequently, this simple set of probabilistic heuristics makes the same distinctions among the valid syllogisms as the mental models account. In this Pre´cis, we concentrate on novel predictions that allow us to put clear water between PHM and other theories. As we discussed earlier, the most important feature of PHM is the extension to generalised quantifiers, like most and few. No other theory of reasoning has been applied to syllogistic reasoning with generalised quantifiers. Table 2 shows the p-valid syllogisms involving generalised quantifiers showing the conclusion type and the percentage of participants selecting that conclusion type in Chater and Oaksford’s (1999b) Experiments 1 and 2. The single lines divide syllogisms with different maxpremises, showing a clear ordering in levels of endorsements dependent on heuristic T1. All those above the double line conform to the min-heuristic (heuristic G1), whereas those below it do not and require a p-entailment (heuristic G2). As Chater and Oaksford (1999b) pointed out, one difference with experiments using standard logical quantifiers was that the Some. . .not conclusion was not judged to be as uninformative, that is, heuristic T2 was not as frequently in evidence. However, in general, in experiments using generalised quantifiers in syllogistic arguments the heuristics of PHM predict the findings just as well as for the logical quantifiers (Chater & Oaksford 1999b). Many further results have emerged that confirm PHM. The min-heuristic captures an important novel distinction between strong and weak possible conclusions introduced by Evans et al. (1999). They distinguished conclusions that are necessarily true, possibly true, or impossible. For example, taking the syllogism discussed earlier (with premises, Some P are Q, All Q are R), the conclusion Some P are R follows necessarily, No P are R is impossible, and Some P are not R is possible. Some possible conclusions are endorsed by as many participants as the necessary conclusions (Evans et al. 1999). Moreover, some of the possible conclusions were endorsed by as few participants as the impossible conclusions. Evans et al. (1999) observed that possible conclusions that are commonly endorsed all conform to the min-heuristic, whereas those which are rarely endorsed violate the

Oaksford & Chater: Pre´cis of Bayesian Rationality Table 2. The p-valid syllogisms less the syllogisms that are also logically valid (shown in Table 1), showing the form of the conclusion and the proportion of participants picking the p-valid conclusion in Chater and Oaksford’s (1999b) Experiments 1 and 2 Syllogism

Conclusion

Mean

All(Q,P), Most(R,Q) Most(Q,P), All(R,Q) All(P,Q), Most(Q,R) Most(P,Q), All(Q,R) Few(P,Q), All(R,Q) All(P,Q), Few(R,Q) Few(P,Q), All(R,Q) All(P,Q), Few(Q,R)

Most Most Most Most Few Few Few Few

85 65 70 55 80 85 85 75

Most(Q,P), Most(R,Q) Most(P,Q), Most(Q,R) Few(Q,R), Most(R,Q) Most(Q,R), Few(R,Q) Most(P,Q), Few(Q,R)

Most Most Few Few Few

65 50 60 75 70

Most(Q,P), Some. . .not(R,Q) Some. . .not(Q,P), Most(R,Q) Some. . .not(Q,P), Most(Q,R) Most(Q,P), Some. . .not(Q,R) Most(P,Q), Some. . .not(Q,R) Some. . .not(P,Q), Most(Q,R)

Some. . .not Some. . .not Some. . .not Some. . .not Some. . .not Some. . .not

80 60 75 65 75 75

Few(Q,P), Some. . .not(R,Q) Some. . .not(Q,P), Few(R,Q) Some. . .not(Q,P), Few(Q,R) Few(Q,P), Some. . .not(Q,R) Few(P,Q), Some. . .not(Q,R) Some. . .not(P,Q), Few(Q,R)

Some. . .not Some. . .not Some. . .not Some. . .not Some. . .not Some. . .not

60 40 30 60 60 40

All(P,Q), Most(R,Q) Most(P,Q), All(R,Q) Few(Q,P), Few(R,Q) Few(P,Q), Few(Q,R) Few(P,Q), Most(Q,R)

Some. . .not Some. . .not Some. . .not Some. . .not Some. . .not

35 35 35 30 30

Note. This table excludes the eight MI, IM, FI, and IF syllogisms, which have two p-valid conclusions only one of which was available in Chater and Oaksford’s (1999b) Experiment 2.

min-heuristic (with one exception). Hence, PHM captures this important new set of data. Some experiments designed to test the claim that syllogism difficulty is determined by the number of alternative mental models can also be interpreted as confirming PHM (Newstead et al. 1999). Participants wrote down or drew diagrams consistent with the alternative conclusions they entertained, during syllogistic reasoning. No relationship was found between the number of models a syllogism requires (according to mental models theory) for its solution and the number of conclusions or diagrams participants produced. This suggests that sophisticated analytical procedures, such as those described in mental models, play, at most, a limited role in the outcome of syllogistic reasoning. By contrast, participants’ responses agreed with those predicted by the min- and attachment-heuristics. Furthermore, no

differences in task difficulty dependent on syllogistic figure were observed, a finding consistent with PHM, but not mental models. Recent work relating memory span measures to syllogistic reasoning has also confirmed PHM (Copeland & Radvansky 2004). PHM makes similar predictions to mental models theory because the number of heuristics that need to be applied mirrors the one, two, and three model syllogism distinction (see Table 1). For one model syllogisms, just the min-heuristic and attachment is required (two heuristics). For two model syllogisms, the some__not-heuristic is also required (three heuristics). In addition, for three model syllogisms a p-entailment is required (four heuristics). The more mental operations that need to be performed, the more complex the inference will be, and the more working memory it will require. Copeland and Radvansky (2004) found significant correlations between working memory span and strategy use, for both mental models and PHM. While not discriminating between theories, this work confirmed the independent predictions of each theory for the complexity of syllogistic reasoning and its relation to working memory span. As with Chapters 5 and 6, Chapter 7 of BR closes by addressing the critiques of PHM that have arisen since the theory first appeared. One criticism is that PHM does not generalise to cardinal quantifiers (Geurts 2003) such as Exactly three P are Q, which have no probabilistic interpretation. Yet, such quantifiers can, nonetheless, naturally mesh with the generalized quantifiers, to yield interesting inferences. For example, suppose you are told that exactly three birds in the aviary are black. If there are twenty birds in the aviary, then few of the birds are black; if there are four, then most of the birds are black; and, in either case, further inferences from these generalized quantifiers can be drawn, as appropriate. 8. Conclusion As we have seen, Chapters 5 to 7 of BR provide the empirical support for taking a probabilistic approach to human reasoning and rationality. The final chapter provides further arguments for pursuing this research strategy in the form of a dialogue between an adherent of the probabilistic approach and a sceptic. In this Pre´cis, we concentrate on two key issues that emerge from that debate. The first topic we consider is whether the brain is a probabilistic inference machine. BR focuses primarily, as we have seen, on providing rational analyses of human reasoning – and we have noted that rational analysis does not make direct claims about underlying computational operations. But, to what extent can the mind or brain be viewed as a probabilistic (or for that matter, a logical) calculating engine? Although not the primary focus in this book, this is nonetheless a fundamental question for the behavioural and brain sciences. We suspect that, in general, the probabilistic problems faced by the cognitive system are too complex to be solved by direct probabilistic calculation. Instead, we suspect that the cognitive system has developed relatively computationally “cheap” methods for reaching solutions that are “good enough” probabilistic BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

83

Oaksford & Chater: Pre´cis of Bayesian Rationality solutions to be acceptable. In particular, where we propose a specific processing theory (in our account of syllogistic reasoning) this account consists of simple, but surprisingly effective, heuristics – heuristics that, however, lead to errors, which we argue are revealed in the empirical data. Moreover, in related work on the topic of rational choice and decision making, which we do not consider here, we and others have proposed models that solve probabilistic/decision making problems, but do so using relatively cheap, and hence approximate, methods (Gigerenzer & Goldstein 1996; Gigerenzer et al. 1999; Stewart et al. 2006). To the degree that algorithmic models can be formulated, is rational analysis simply redundant? We argue that it is not. Rational analysis is essential because it explains why the particular algorithms used by the cognitive system are appropriate. That is, without a characterization of what problem the cognitive system solves, we cannot ask, let alone answer, the questions of why the algorithm has its particular form, or how effectively it works. Moreover, it may be that a good deal of empirical data about human reasoning (and indeed, human cognition more generally) can be understood as arising from the structure of the problem itself – that is, the nature of the problem drives any reasonable algorithmic solution to have particular properties, which may be evident in the data. This idea is a core motivation for the rational analysis approach (Anderson 1990; 1991a); and we have seen that a broad spectrum of data on human reasoning can be understood purely at the rational level – that is, without formulating an algorithmic theory of any kind. The second topic we consider is the importance of qualitative patterns of probabilistic reasoning, rather than precise numerical calculations. Suppose, for concreteness, we consider a person reasoning about a game of dice. If the dice are unbiased, then it is easy, of course, for the theorist to formulate a probabilistic model specifying that each throw is independent, and that each face has a probability of 1/6. But this model is both too strong and too weak. It is too strong because it generates all manner of subtle mathematical predictions, concerning, say, the relative probabilities of rolling at least one six out of six dice rolls versus rolling at least two sixes out of twelve dice rolls, predictions that are not available to everyday intuition. And it is too weak because it ignores many factors of crucial importance in everyday reasoning. For example, watching a dice being thrown, we have not only a model of the probability that each face will be uppermost, but a rough model of where it will land, how likely it is to fall off the table, how loud that impact is likely to be, how another player is likely to react to a particular outcome, given their temperament, the gamble they have placed, and so on. This observation implies that, if the cognitive system is indeed building probabilistic models of the world, then it is building models of considerable complexity – models that can take into account any aspect of knowledge, from naı¨ve physics to folk psychology. This implies that the probabilistic turn does not resolve the difficulty of representing knowledge – rather it provides a framework into which this knowledge must be integrated. The advantage of the probabilistic viewpoint, though, is that it provides a powerful framework for 84

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

dealing with an uncertain world; and, indeed, for assessing competing explanations of observed phenomena (rival interpretations of perceptual input; competing grammars; alternative interpretations of sentences, stories, or court-cases). Moreover, probabilistic models of complex domains do not need to be fully specified, at a numerical level – most critical is that the functional relationships between pieces of information are represented. What tends to cause what? What is evidence for what? The direction and existence of functional dependencies between pieces of information may be mentally represented, even though precise numerical probabilities may be unknown. Thus, probability theory can provide a framework for qualitative reasoning, without using numerical values (e.g., Pearl 2000). We tentatively suggest that much of the power, and limitations, of human reasoning about the everyday world flows from this qualitative style of reasoning. From this point of view, it is perhaps not surprising that people are not good at explicit reasoning with probabilities – indeed, they fall into probabilistic fallacies just as readily as they fall into logical contradictions (e.g., Kahneman et al. 1982). The probabilistic mind is not, of course, a machine for solving verbally or mathematically specified problems of probability theory. Instead, we suggest, the mind is a qualitative probabilistic reasoner, in the sense that the rational analysis of human reasoning requires understanding how the mind deals qualitatively with uncertainty. As we have stressed, this does not imply that the mind is a probabilistic calculating machine (although it may be); still less does it imply that the mind can process probabilistic problems posed in a verbal or mathematical format. Nonetheless, the concepts of probability are, we suggest, as crucial to understanding the human mind as the concepts of aerodynamics are in understanding the operation of a bird’s wing.

ACKNOWLEDGMENTS The book Bayesian Rationality owes huge debts to more people and institutions than we can summarize here. We would especially like to express our gratitude to the University of Edinburgh and the Centre for Cognitive Science (formerly the School of Epistemics, and now part of the Division of Informatics) for starting us on the adventure of understanding human reasoning from a formal and empirical standpoint, which has engaged so much of our lives for the last twenty years.

NOTES 1. The case where the categorical premise is uncertain can be accommodated using a generalization of this idea, Jeffrey conditionalization (Jeffrey 1983). The new degree of belief that John plays tennis (q), on learning that it is sunny in Bloomsbury (which confers only a high probability that it is sunny in Wimbledon [ p]), is: P1 (q) ¼ P0 (qjp)P1 (p) þ P0 (qj:p)P1 (:p) 2. Bayes’ theorem is an elementary identity of probability theory that allows a conditional probability to be calculated from its converse conditional probability and the priors: P( pjq) ¼ (P(qjp)P( p))/P(q). 3. However, this may be because of the different way that negations are used in each task (see Evans & Handley 1999; Oaksford 2004b).

Commentary/Oaksford & Chater: Pre´cis of Bayesian Rationality

Open Peer Commentary Classical logic, conditionals and “nonmonotonic” reasoning doi:10.1017/S0140525X09000296 Nicholas Allott and Hiroyuki Uchida Department of Linguistics, University College London, London WC1N 1PF, United Kingdom. [email protected] [email protected] http://www.phon.ucl.ac.uk/home/nick/

Abstract: Reasoning with conditionals is often thought to be nonmonotonic, but there is no incompatibility with classical logic, and no need to formalise inference itself as probabilistic. When the addition of a new premise leads to abandonment of a previously compelling conclusion reached by modus ponens, for example, this is generally because it is hard to think of a model in which the conditional and the new premise are true.

We doubt two claims made by Oaksford & Chater (O&C), one linguistic and one about inference, as they relate to conditionals (Bayesian Rationality, Oaksford & Chater 2007, henceforth BR as in target article). We do not think that the case has been made that sentences of the form “If A then B” have a semantics diverging from the material implication of propositional logics. We focus here on the related claim that human inference is probabilistic. Classical Propositional Logic (CPL) is often claimed to be inadequate for explaining our spontaneous propositional inferences. The claim is based on the observation that, whereas CPL is monotonic, human inferences seem to be non-monotonic. For example, it is argued that the validity of some inferences, such as the one in (1a), may be cancelled by addition of another premise, such as the premise R in (1b). The claim is that now that John has broken his left leg, John will no longer run, even if the weather is fine. This is claimed to show the inadequacy of CPL as a tool to explain our propositional inferences, because in CPL, addition of another premise does not influence the provability of the sequent, as is shown in the inference from top to bottom in (1c).

(1a) (Premise 1) P ! Q: If the weather is fine, John runs for a mile. (Premise 2) P: The weather is fine. (Conclusion) 3. Q: John runs for a mile. The valid inference: Whenever Premise 1 and Premise 2 are true, 3 is also true. (1b) R: John breaks his left leg. (1c) P, P ! Q r Q P, P ! Q, R r Q Weakening However, it is not clear that the addition of the premise R in (1b) to the set of premises in (1a) actually cancels out the validity of the initial sequent, P, P ! Q r Q, in our propositional inferences. The point becomes clearer if we add the new proposition R to the set of premises, as in P, P ! Q, R r Q. With the specification of the premise propositions as in (1a) to (1b), we do not find this revised sequent to be valid, but this is only because we no longer find the proposition “If the weather is fine, John runs for a mile” (¼ P ! Q) to be true when we reevaluate the inference. In contrast, if we force ourselves to assume that all the three premise propositions P, P ! Q, R are true in a model of interpretation, then in that model, we have to conclude Q. Some might find it difficult to think of such an interpretation model, because common-sense knowledge tells us that a

person normally does not run with a broken leg, but we may sometimes make a claim such as, “If the weather is fine, John (always) runs for a mile. It does not matter if he gets injured. He always runs for a mile.” Thus, it is not impossible to force ourselves to think of models in which the proposition, “If the weather is fine, John runs for a mile” is true despite the fact that John has broken his leg. In such models, whenever P is true, Q is also true: that is, the inference goes through. As is clear from the foregoing discussion, typical interpretation data that allegedly show that monotonic logic cannot capture our propositional inferences include a change of models in which we evaluate the propositional sequents. The valid inference in (1) can be re-stated as “In each model in which P and P ! Q are both true, Q is also true.” The cancellation of the truth of Q arises because in the new model in which we “re-evaluate” the sequent, the premise P ! Q is no longer true (or we find it more difficult to think of a model in which both P ! Q and R are true). Because the initial valid inference in (1a) concludes R as a true proposition only on condition that P ! Q and P are both true, this revision does not really involve the cancellation of the validity of the initial inference. Given that specification of models in which logical formulas/ sequents are evaluated is not part of either the syntax or the semantics of propositional logic languages, it is not clear that alleged non-monotonicity of reasoning, which arises because of the revision of models, requires formalising logical inference as non-monotonic or probabilistic. No doubt people hold beliefs with varying degrees of strength, and a result of reasoning is that these degrees of strength are changed. But one can agree with O&C on this point without thinking that the formal inference system itself is probabilistic. Space here does not permit discussion of how some propositions are accepted and others rejected, but we are sympathetic to O&C’s claim that heuristics that are sensitive to information gain must be involved, with the caveat that it cannot simply be information that is sought, but information that is important to the reasoner at a reasonable processing cost. This recalls discussion of relevance in Gricean pragmatics and Sperber and Wilson’s relevance theory. As Oakford & Chater note, for one’s belief in the conditional in (2) it matters whether one discovers, for example, an unstarted car or is told that a car did not start. (2) If the key is turned the car will start.

A pragmatic explanation in terms of the tendency of speakers to produce utterances relevant to their audience is natural. Effects of the order in which information is presented (see BR, pp. 157ff) also require such an explanation, we believe. This raises a methodological point. To understand human reasoning, both classical logic and O&C’s probabilistic account of conditionals and of inference must be supplemented by accounts of processing, and of the pragmatics of utterance interpretation. Thus, it is not obvious that the probabilistic account is more parsimonious.

Identifying the optimal response is not a necessary step toward explaining function doi:10.1017/S0140525X09000302 Henry Brighton and Henrik Olsson Center for Adaptive Behavior and Cognition, Max Planck Institute for Human Development, 14195 Berlin, Germany. [email protected] [email protected]

Abstract: Oaksford & Chater (O&C) argue that a rational analysis is required to explain why a functional process model is successful, and that, when a rational analysis is intractable, the prospects for BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

85

Commentary/Oaksford & Chater: Pre´cis of Bayesian Rationality understanding cognition from a functional perspective are gloomy. We discuss how functional explanations can be arrived at without seeking the optimal response function demanded by a rational analysis, and argue that explaining function does not require optimality.

Oaksford & Chater (O&C) argue in Bayesian Rationality (Oaksford & Chater 2007, henceforth BR) that a rational analysis is an essential step toward understanding process models from a functional perspective and that “[d]oing this requires developing an account of the optimal behaviour” (p. 268). We argue that relative uses of rational principles of inductive inference can be used to explain function without knowing the optimal response function, and propose that multiple forms of functional analysis are required to understand the cognitive system from a functional perspective. Rational principles of inductive inference such as Bayesian and simplicity principles are perhaps most frequently used as criteria in relative statements of the form “P is a better response than Q for problem X” For instance, the use of rational principles as model selection criteria (Kearns et al. 1997; Pitt et al. 2002). In contrast, step four of Anderson’s (1991a) rational analysis excludes anything but statements of the form “P is the optimal response to task X.” Because rational principles can be used to compare the behavior of process models without knowing the optimal response, rational analysis is an instance of a broader class of methodologies adopting rational principles to understand function (e.g., Gigerenzer et al. 1999). Given this, does knowledge of the optimal response offer any intrinsic advantage when explaining why a process is successful; and what price do we pay by demanding knowledge of the optimal response function? Knowledge of the optimal solution is one way of establishing that a particular process or organism is successful; but explaining why the process is successful is not implied by this finding. For example, knowing the optimality conditions of the naı¨ve Bayes classifier does not by itself tell us why it is successful. One explanation for why naı¨ve Bayes is successful might be that the independence assumption results in fewer parameters. In certain contexts which violate this independence assumption, the assumption nevertheless causes a reduction in the variance component of error relative to a learning algorithm that assumes that the features are dependent (Domingos & Pazzani 1997). This causal explanation for why naı¨ve Bayes is successful does not require knowledge of the optimal response. It can be established using relative uses of rational principles when the optimal response is incalculable. Furthermore, for realistic contexts of sparse exposure the optimality conditions for naı¨ve Bayes, despite its simplicity, are not fully known (Kuncheva 2006). Although typically unavailable, knowing the optimality conditions for an algorithm would undoubtedly provide a good starting point to understand its function; but optimality conditions are neither a required starting point, nor do they by themselves offer a causal explanation for why the algorithm is functional. Functional explanations arising from Bayesian rational analyses also aim to tell us why a pattern of behavior is rationally justified given that the environment has a certain probabilistic structure. The optimal response function provides a reference point against which to measure behavior without committing to how that behavior is achieved. Because relative uses of rational principles are order relations over theories, or models, of problem solving, they cannot be used in this way. However, abstracting from the process level limits what causal explanation one can offer for why the organism does what it does. One can say that it is successful, but whether or not such a finding provides a satisfactory causal explanation for why it is successful is not so clear (Danks 2008). Problems arising from the intractability of reliably identifying the optimal response can also make it less desirable. O&C contemplate the possibility that when a rational analysis is intractable, cognitive science may also be intractable (BR, p. 283). Alternatively, this makes complementary forms of functional analysis all the more necessary. The distinction between relative

86

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

and absolute uses of rational principles mirrors the distinction between, and relative difficulty of, verification and search problems in complexity theory. For example, for an instance of the traveling salesperson problem, the comparative statement “Tour P is shorter than tour Q” is trivial to verify, but the absolute statement “Tour P is the shortest tour” will often be intractable to establish. Many problems take this form and are NP complete (Nondeterministic Polynomial time), including the computation of optimal Bayesian responses and approximations in many settings (e.g., Cooper 1990). Functional analyses based on rational principles of induction rest on several idealizations: (a) Although different rational principles of induction point to a coherent theoretical picture of what makes a good inference, their practical implementations are often inconsistent, and point to different conclusions (Kearns 1997); (b) functional models will not capture all forms of uncertainty impacting on the problem, some of which may change the character of a functional response (Bookstaber & Langsam 1985); (c) functional models always consider local goals, which only partially inherit the properties of the global goal being examined; (d) rational principles of inductive inference are approximate models of function, which do not consider functional pressures arising from, for example, processing (Brighton & Gigerenzer 2008; Todd & Gigerenzer 2003). These are unavoidable realities of modeling, and apply to both relative and absolute uses of rational principles of induction. An explanation requiring an optimal response function must also consider that: (e) for problems of inductive inference, the optimal response is often analytically intractable to determine with exact methods, and will not be unique; (f) behavioral responses are typically approximately optimal, revealing a tendency rather than a correspondence; (g) successfully optimizing a local goal does not necessarily take us toward the global optimal when other dependencies are known to be only approximately fulfilled (Lipsey & Lancaster 1956). These additional factors lead to increased flexibility in what behaviors we choose to label optimal. Our point is that these further assumptions are a choice rather than a necessity, and are only required to support certain forms of explanation. We agree with (O&C) on the importance of understanding the ecological function of cognitive processes. We also agree that rational analysis represents a powerful move in this direction. But functional analysis can also proceed without seeking to establish optimality with respect to the organism.

Explaining norms and norms explained doi:10.1017/S0140525X09000314 David Danksa and Frederick Eberhardtb a Department of Philosophy, Carnegie Mellon University and Institute for Human and Machine Cognition, Pittsburgh, PA 15213; bInstitute of Cognitive and Brain Sciences, University of California, Berkeley, Berkeley, CA 94720, and Department of Philosophy, Washington University in St. Louis, St. Louis, MO 63130. [email protected] http://www.hss.cmu.edu/philosophy/faculty-danks.php [email protected] http://www.phil.cmu.edu/  fde/

Abstract: Oaksford & Chater (O&C) aim to provide teleological explanations of behavior by giving an appropriate normative standard: Bayesian inference. We argue that there is no uncontroversial independent justification for the normativity of Bayesian inference, and that O&C fail to satisfy a necessary condition for teleological explanations: demonstration that the normative prescription played a causal role in the behavior’s existence.

In Bayesian Rationality (Oaksford & Chater 2007, henceforth BR) we understand Oaksford & Chater (O&C) as providing the

Commentary/Oaksford & Chater: Pre´cis of Bayesian Rationality following argument for Bayesian models as teleological explanations in psychology: 1. Normative Justification: The normativity of Bayesian inference can be established independently of empirical observations. (p. 31 of their book) 2. Normative Force: Appropriate normative principles lead to behavior conforming to those principles. (p. 33) 3. Default Assumption: People are pretty good at achieving their goals (i.e., they are “everyday rational”). (p. 19) Therefore, people’s behavior is a guide to the appropriate normative principles. (p. 30) 4. Empirical Results: People’s behavior conforms to the prescriptions of Bayesian models. (Chs. 6 & 7) Conclusion: Therefore, Bayesian models explain why that behavior occurs. (Ch. 8) The first three premises are formal claims that are instances of necessary parts of proper teleological explanations. Without independent justification (Premise 1), there is no normative power behind Bayesian inference. Without influence on behavior (Premise 2), normative principles provide no explanation. And if the normative prescription cannot be reasonably interpreted in terms of desirable human goals (Premise 3), then the explanation is implausible. We suggest, however, that neither Premise 1 nor Premise 2 is supported by O&C’s arguments. Premise 1 is usually grounded in two standard arguments. Dynamic Dutch book arguments aim to establish Bayesian inference as the unique belief change procedure that avoids irrational, between-time inconsistencies, understood as acceptance of bets (potential, not actual) over time that result in guaranteed loss (Teller 1973). Standard forms of these arguments, however, also imply absurd prescriptions, including: reasoners should not protect themselves against predictable irrationalities (Maher 1992); reasoners should not retrospect on their past beliefs (Levi 1988); and reasoners should never change their conditional commitments (Levi 1988; 2002). If these arguments are weakened to avoid these unwanted implications, then Bayesian inference is only one of infinitely many ways to avoid dynamic Dutch book; it has no particularly special status. The second standard argument is long-run convergence: Roughly, any non-zero degree of belief in the truth will converge to 1 (using Bayesian inference) as one collects more evidence (Savage 1972), and no other reliable method always converges to the truth faster (Schulte 1999). However, the convergence arguments have the unreasonable requirement that the Bayesian reasoner be logically omniscient. There are also problems that can be solved by a naı¨ve falsificationist, but which the Bayesian can solve only if she can “compute” uncomputable functions (Juhl 1993; Osherson et al. 1988). Long-run convergence thus cannot provide a conceptual justification for the normativity of Bayesian inference. There might be other arguments for Premise 1, but we have not seen them, nor do O&C provide them. We agree with O&C that the normative principles underlying rationality may vary between situations. But independent justification of normativity must be provided for whatever principles are appropriate, else one cannot have an adequate teleological explanation. The dependence of the normative principle on the situation is inversely correlated with the explanatory power provided by the normative principle. A situation-independent justification of Premise 1 (or some similar premise) is necessary for teleological explanations. Our second worry is that, even if Premise 1 holds, O&C require only an empirical match between behavior and normative prescription. Their argument thus neglects the requirement for a teleological explanation that the normative principle must have played a causal role 2ontogenetic, phylogenetic, or both 2 in ˙ are the behavior’s existence or persistence. “Origin stories” required for teleological explanation, but are never provided by O&C. Behavior B could be optimal for task T even though behavior B results from considerations that are independent of task T; B’s optimality might be coincidental. In this case, the claim

“Behavior B because it is optimal for task T” is an incorrect explanation, even though B conforms (empirically) to the normative prescription. O&C seem to have two lines of response. First, if there is widespread alignment between the normative prescription and people’s behavior, then it is arguably improbable that the behavior is only coincidentally optimal. Consequently, research confirming correspondence in a wide range of conditions is evidence in favor of the normative principle. This response fails, however, because widespread alignment is actually to be expected for Bayesian models, given the many “free parameters” in such models: hypothesis space, prior probabilities, utilities, likelihood functions, and various plausible computational approximations. In particular, for any evidence and any behavior (represented as a probability distribution over possible choices), there exists a prior probability distribution such that the predicted posterior distribution after inference on the evidence matches the observed behavior.1 That is, for (almost) any psychological experiment, we know a priori that a Bayesian model will be capable of fitting the observed behavior, whatever it might be. Repeatedly developing sui generis Bayesian models for each task does not compensate for a lack of “origin stories,” even if the models successfully predict. O&C’s second line of response is more promising: successful model predictions for experiments that vary features of one particular task make it plausible that the normative constraints played a causal role in shaping the observed behavior. We support this line of response to the lack of “origin stories,” but are doubtful about how much support the evidence provides for the normativity of Bayesian inference. O&C here primarily cite manipulations of the base-rate (BR, pp. 146ff, 178ff), and as they recognize, there are several competing models with similar normative appeal. Moreover, there is substantial controversy about which model provides the best fit of the data. There is a potential response, but it is currently only potential. O&C admirably search for teleological explanations of human behavior. We are in complete agreement with them that such explanations are desirable, but we believe that their enterprise requires a stronger foundation. They have neither properly established Bayesian inference as a normative principle, nor shown – directly or indirectly – that the optimality of Bayesian inference (if it is optimal in some respect!) is actually a cause of people approximating those prescriptions. They offer only an existence proof – “Behavior could be caused by a potentially normative principle” – and such accounts are too weak to provide teleological explanations. ACKNOWLEDGMENT The second author, Frederick Eberhardt, is funded by a fellowship from the James S. McDonnell Foundation. NOTE 1. There are other technical assumptions (e.g., nonextremal likelihoods), but they are irrelevant to our core point.

Beyond response output: More logical than we think doi:10.1017/S0140525X09000326 Wim De Neys Lab for Experimental Psychology, University of Leuven, 3000 Leuven, Belgium. [email protected] http://ppw.kuleuven.be/reason/wim/

Abstract: Oaksford & Chater (O&C) rely on a data fitting approach to show that a Bayesian model captures the core reasoning data better than its logicist rivals. The problem is that O&C’s modeling has focused

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

87

Commentary/Oaksford & Chater: Pre´cis of Bayesian Rationality exclusively on response output data. I argue that this exclusive focus is biasing their conclusions. Recent studies that focused on the processes that resulted in the response selection are more positive for the role of logic.

It is well established that in a wide range of reasoning tasks people fail to select the response that standard logic dictates. Oaksford & Chater (O&C) in Bayesian Rationality (Oaksford & Chater 2007) nicely demonstrate that a Bayesian model fits the available response selection data better than its logicist rivals. However, it is crucial to stress that O&C’s modeling has not moved beyond this mere response output data. Other measures such as basic latencies or brain-imaging data that can help to examine the processes that led to the outputs are not being considered. This limitation is bound to bias their conclusion. It is clear that people’s final response selection does not tell us how they arrived at this response. The fact that people do not select the predicted logical response does not suffice to discard the logicist view. It is always possible that people attempt to reason in line with standard logic but fail to do so because of specific processing difficulties. Addressing this issue requires that one digs bellow the output surface. In O&C’s defense, it may be noted that the fixation on response output is characteristic for the whole psychology of reasoning (e.g., Hoffrage 2000). In the vast majority of classic reasoning studies, response accuracy has been the sole dependent variable of interest. From a pragmatic point of view, one might argue that O&C were simply focusing on the most prominent data and might move to more process-related measures in the future. However, the key point is that in the meantime caution is needed when drawing any strong conclusions about the role of logic in reasoning. I will illustrate the point with some examples from recent brain-imaging and latency studies. This processing data suggests that although people very often fail to select the predicted logical response, they are nevertheless trying to adhere to the logical norm. Houde´ et al. (2000), for example, scanned people’s brain activation while they were trying to solve the selection task. Participants were scanned before and after they received training aimed at increasing their inhibitory control capacities. Although the training did not instruct people about the standard logical response (i.e., P and not-Q card selection), it did boost the selection rate of this pattern and resulted in an increased activation of prefrontal brain areas believed to be involved in inhibitory processing. This suggests that the initial low selection rate of the predicted logical response does not result from a lack of logical knowledge, but from a failure to inhibit the intuitive appealing matching response. If people did not know what the logical response was or considered it to be irrelevant, merely training their inhibitory capacities should not affect its selection rate. Likewise, numerous studies on belief bias in syllogistic reasoning have shown that people typically fail to respond logically to reasoning problems in which intuitive beliefs conflict with the logical response. Nevertheless, latency and eye-tracking studies indicate that these problems take longer to respond to and are more thoroughly inspected compared to problems in which beliefs and logic do not conflict (e.g., Ball et al. 2006). If people were not trying to adhere to the logical norm, violating it should not affect their reasoning behavior. Furthermore, consistent with the latency findings, De Neys and Franssens (2007) observed that after solving syllogisms in which beliefs and logic conflicted, reasoners showed an impaired access to words that were associated with the cued beliefs. Such an impaired memory access is a key marker of inhibitory processing. Even people who were always biased by their beliefs showed a minimal impairment, indicating that they had attempted to inhibit the intuitive beliefs, but failed to complete the process. If people were not trying to adhere to the logical norm, there would be no reason for them to block the conflicting intuitive response. Bluntly put, although people did not manage to select the logical response, they were at least trying to do so.

88

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

Interestingly, brain-imaging of classic tasks from the judgment and decision-making literature is pointing towards a similar conclusion. As O&C note, people hardly ever reason in line with classic probabilistic norms in these tasks. People typically select responses that are cued by intuitive heuristics. Nevertheless, a recent fMRI study showed that a specific brain-area, the anterior cingulate cortex, involved in the detection of conflict between competing responses, is activated when people select the heuristic response and violate the norm (De Neys et al. 2008). This indicates that people detect that their heuristic response conflicts with the classic normative response. If people were not taking the norm into account, they would not experience such a conflict (see also De Neys & Glumicic 2008). The general point is that although people’s response selections in all these cases deviate from the predicted classic normative pattern, additional process data indicates that people are nevertheless trying to adhere to the norm. This should give pause for thought before discarding the logical perspective. O&C have modeled but a small, superficial subset of relevant data. If one moves beyond the static response output, it becomes clear that people can be more logical than their answers suggest. ACKNOWLEDGMENT The authors’ research is supported by the Fund for Scientific Research Flanders (FWO-Vlaanderen).

Does rational analysis stand up to rational analysis? doi:10.1017/S0140525X09000338 Jonathan St. B. T. Evans Centre for Thinking and Language, School of Psychology, University of Plymouth, Plymouth PL4 8AA, United Kingdom. [email protected]

Abstract: I agree with Oaksford & Chater (O&C) that human beings resemble Bayesian reasoners much more closely than ones engaging standard logic. However, I have many problems with their “rational analysis” framework, which appears to be rooted in normative rather than ecological rationality. The authors also overstate everyday rationality and neglect to account for much relevant psychological work on reasoning.

There is much that I can admire in Bayesian Rationality (Oaksford & Chater 2007, henceforth BR) and the lengthy research programme that it describes. There is also much that I agree with: for example, that human beings resemble Bayesian reasoners much more closely than ones engaging standard logic. I also share part of the authors’ research programme, in that I have argued against logicism and excessive use of the deduction paradigm in the psychology of reasoning (Evans 2002), as well as proposed that ordinary conditionals are much better modelled by probability logic than by standard propositional logic (Evans & Over 2004). So why is it that when I read the Socratic dialogue presented in Chapter 8 of BR, my sympathy is constantly with the sceptic? The problems for me lie with “rational analysis” and the manner in which Oaksford & Chater (O&C) go about doing their science. Rarely do we see anything approaching a descriptive or algorithmic account of reasoning (the model of syllogistic reasoning being the closest) or even – in spite of their philosophy – a genuinely computational one. What we typically get is an alternative norms account. Thus, choices on the Wason selection task look erroneous from the viewpoint of standard logic, but can be seen as rational attempts to maximise information gain (given a misreading of the task instructions that makes it about categories instead of four individual cards). Similarly, we can predict people’s conditional inference rates on the assumption

Commentary/Oaksford & Chater: Pre´cis of Bayesian Rationality that they are somehow reflecting the probability of the conclusion given the minor premise (despite the oddity that uncertainty resides exclusively in the major premise), and so on. The object of the exercise seems to be to show that there is some normative account of behaviour that can be offered. As a cognitive psychologist, I find this unsatisfying, because I want to know what people are actually doing and how. The basic philosophy of the authors is set out in Chapter 2, on which I will focus. They start out by describing essentially the same paradox as the one that inspired our earlier book on rationality (Evans & Over 1996a). Why is it that people’s reasoning and decision making seems essentially rational, intelligent, and adaptive in the real world, but results in numerous violations of logic and other standard normative systems when they are tested in the laboratory? We suggested that while people have some limited capacity for explicitly following normative rules (rationality2), they could often achieve everyday goals by implicit processes such as associative learning (rationality1). O&C (BR, p. 29) object, however, to any form of rationality that cannot be justified by a normative system. But elsewhere in the book, they seem to be describing something very similar to rationality1. For example, they describe some examples of “rational” behaviour with the comment: “Such behaviour may be built in by evolution or acquired via a long process of learning – but need not require on-line computation of the optimal solution” (BR, p. 36). Examples discussed here include Gigerenzer’s research programme on fast and frugal heuristics (Gigerenzer 2004). But Gigerenzer appeals to ecological and not normative rationality, so the link between the two programmes is unclear. The authors overstate everyday rationality when they ask (BR, p. 30), “why do the cognitive processes underlying everyday rationality consistently work?” Well, they don’t, and to say otherwise is to ignore a massive body of psychological literature. For example, in the study of judgement and decision making, many cognitive biases have been shown to occur reliably in everyday contexts and with expert judges: phenomena such as outcome bias, hindsight bias, overconfidence, and the planning fallacy, to name just a few (for evidence, see papers from two recent collections on these topics: Gilovich et al. 2002; Koehler & Harvey 2004). There is also a massive accumulation of evidence for dual-process accounts of reasoning and decision making (Evans 2007; 2008), something which plays no visible role at all in the account of human reasoning offered by O&C. Nor do O&C feel moved to explain why the achievement of standard normative solutions to the task for which they provide alternative normative accounts is associated with those of high general intelligence (Stanovich 1999). It is particularly disappointing that they make no reply to Stanovich’s carefully constructed challenge to their position. A final issue is whether their methodology provides computational or normative accounts of behaviour, because these are not the same thing (Elqayam 2008). The authors try to combine them as follows: “Formal rational principles spell out the optimal solution for the information processing problem that the agent faces. The assumption is that a well-adapted agent will approximate this solution to some degree” (BR, p. 38). But the adaptive mechanisms of nature – evolution and learning – do not optimise. If evolution had an optimal solution, we would not have billions of different species of every conceivable form and function. If learning was optimised, animals would not, for example, generally match the probability of their foraging to that of food sources. But O&C are quite explicit (BR, pp. 30 – 31) that they are not worried about the performance of a priori normative theories in providing such quasi-computational accounts. On the contrary, they are looking for empirical rationality, which means in essence: observe some behaviour, assume that it is rational, find a normative theory that deems it to be so, and then . . . nothing else, apparently. I can understand the benefits of rational analysis when it is rooted in ecological rationality, so that by assuming that

behaviour is adapted to the environment, we can look to the structure of that environment for cues as to how our cognitive mechanisms are designed. However, I really don’t understand why O&C feel the need to fit a normatively rational model to human reasoning, and still less why that should put an end to the inquiry, with no algorithmic account to follow. Nevertheless, I agree strongly with them that the Bayesian model is a far more appropriate reference for real world reasoning than one based on truth-functional logic, and that it is a standard much more likely to be approximated in the inferences that people actually make.

The strengths of – and some of the challenges for – Bayesian models of cognition doi:10.1017/S0140525X0900034X Thomas L. Griffiths Department of Psychology, University of California, Berkeley, Berkeley, CA 94720-1650. [email protected] http://cocosci.berkeley.edu

Abstract: Bayesian Rationality (Oaksford & Chater 2007) illustrates the strengths of Bayesian models of cognition: the systematicity of rational explanations, transparent assumptions about human learners, and combining structured symbolic representation with statistics. However, the book also highlights some of the challenges this approach faces: of providing psychological mechanisms, explaining the origins of the knowledge that guides human learning, and accounting for how people make genuinely new discoveries.

Oaksford & Chater’s (O&C’s) book Bayesian Rationality (Oaksford & Chater 2007, henceforth BR) is at heart a review of the authors’ research program exploring probabilistic models of human deductive reasoning. However, the review is nested within a set of chapters that make clear the ambitious goal of this research program: not just to explain how people reason, but to provide us with a new functional notion of rationality. This work, together with that of Anderson (1990) and Shepard (1987; 1995) was one of the early successes of rational analysis and Bayesian models of cognition (see also, Chater & Oaksford 1999a). As such, BR illustrates the strengths of rational analysis, and helps to highlight the challenges that Bayesian models face. One of the strengths of rational analysis is producing models that explain human behavior as an adaptive response to problems posed by the environment. Establishing how the notion of rationality appealed to by these explanations relates to traditional normative analyses in psychology and economics is one of the themes of the early chapters of the book, building on previous discussion by Anderson (1990). Loosely speaking, the thesis is that rationality can be a methodological assumption that guides us to explanations of behavior: Rather than deciding that people solve a problem poorly, we should consider the possibility that they are solving another problem well, and try to determine what that problem might be. This analysis can be more satisfying than traditional uses of normative models, in that it produces a systematic picture of the assumptions guiding human inferences rather than a list of deviations. This approach is perhaps best exemplified in BR by the Wason selection task, in which the traditional normative account is a combination of falsificationism and deductive logic, and the alternative construes the problem as one of confirmation and induction. Rational accounts of cognition have virtues that go beyond mere systematicity. One of the most important is the explicit treatment of the assumptions of learners that results from explaining behavior as an adaptive response to computational problems. Computational models of cognition typically focus on the processes that govern behavior. Their content and BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

89

Commentary/Oaksford & Chater: Pre´cis of Bayesian Rationality parameters concern the properties of hypothetical cognitive mechanisms – similarity, strength of association, and learning rates. The content and parameters of rational models are expressed in a different language, being about the problems learners solve and the assumptions they make. As a consequence, these models are more abstract, but more explicit about their commitments, providing a transparent account of what learners should find easy or hard to learn (their “inductive biases”) and what factors should influence behavior (such as the rarity of properties in the Wason task). Developing models formulated in these terms can be valuable, because many questions in cognitive science are posed at the same level of abstraction, concerning the constraints that guide human learning. The treatment of human reasoning presented in BR illustrates another strength of Bayesian models: the capacity to combine symbolic representations with statistical inference. As O&C discuss, logic and probability have often been viewed as distinct. Researchers in philosophy and artificial intelligence have touted the expressive power of deductive logic, whereas probability theory is typically applied to simpler representations such as sequences of coin flips. The models developed in the book begin to combine the expressive power of logic with the inferential capacity of probability, following a trend towards probabilistic logics that is now a hot topic in artificial intelligence (e.g., Friedman et al. 1999; Milch et al. 2004). More generally, the idea of combining structured representations with statistical inference brings together two traditions in cognitive science, where symbolic models employ structured representations and connectionism demonstrates power of statistical learning. As discussed in the book, the potential of the resulting synthesis is great, with structured statistical models being useful when both representation and learning are key, such as inferring the existence of causal relationships (e.g., Griffiths & Tenenbaum 2005) and learning and using language (e.g., Chater & Manning 2006). While the examples presented in BR illustrate the strengths of rational analysis and Bayesian models of cognition, they also highlight some of the challenges that this approach faces. One significant challenge is connecting rational models to more conventional psychological explanations – identifying mechanisms that can support the sophisticated probabilistic inferences required by structured statistical models. Another challenge is understanding where the knowledge that guides these inferences comes from – the source of the priors and hypothesis spaces for Bayesian inference. And finally, the parallels between induction and scientific reasoning drawn in the book raise another compelling question. Following Reichenbach (1938), philosophers of science distinguish between discovery – developing genuinely new hypotheses – and justification – explaining why one hypothesis should be preferred to another. Bayesian inference, at least in the form presented in most Bayesian models, is a logic of justification, with all possible hypotheses being evaluated. So, how might Bayesian models account for discovery, something that would seem to require going beyond the hypotheses available? These are significant challenges, but they might be overcome by drawing on work in other disciplines that use Bayesian models. Probabilistic inference is not just hard for people, it is difficult for computers too. The resulting work on approximate inference in computer science and statistics provides a potential source of psychological mechanisms (e.g., Sanborn et al. 2006; Shi et al. 2008). Hierarchical Bayesian models, originally developed in statistics to capture both individual and population effects, provide a way for learners to start out with vague priors and to refine their beliefs over time to develop more informative expectations (Tenenbaum et al. 2006). Finally, nonparametric Bayesian models use stochastic processes to define priors over infinite, structured hypothesis spaces, allowing new observations to be explained through new hypotheses (e.g., Sanborn et al. 2006; Griffiths & Ghahramani 2006), and perhaps paving the way towards a deeper understanding of discovery. Set in this broader context, BR is a significant step towards a genuinely interdisciplinary science of

90

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

induction, connecting human minds with algorithms from computer science and the abstract principles of statistics. ACKNOWLEDGMENTS I would like to thank J. Austerweil, F. Eberhardt, C. Lucas, F. Reali, J. Vuong, J. Williams, and other members of the Computational Cognitive Science Lab at Berkeley for helpful discussions of the book.

Explaining more by drawing on less doi:10.1017/S0140525X09000351 Ulrike Hahn School of Psychology, Cardiff University, Cardiff CF10 3AT, United Kingdom. [email protected]

Abstract: One of the most striking features of “Bayesian rationality” is the detail with which behavior on logical reasoning tasks can now be predicted and explained. This detail is surprising, given the state of the field 10 to 15 years ago, and it has been brought about by a theoretical program that largely ignores consideration of cognitive processes, that is, any kind of internal behavior that generates overt responding. It seems that an increase in explanatory power can be achieved by restricting a psychological theory.

It is undoubtable that over the last decade, Oaksford & Chater (O&C) have revolutionized the psychology of human reasoning. What were previously viewed as “logical reasoning” tasks, such as Wason’s card selection task, were given a probabilistic reinterpretation. At the same time, their work has formed a corner-stone in the wider probabilistic revolution that has taken place over the last 10 to 15 years: Finding evidence for human sensitivity to probabilities in as unlikely a corner as “logical” reasoning tasks, lends support to the wider claim that human cognition is fundamentally about probabilities. Given the current popularity of Bayesian approaches to cognition (see e.g., Chater et al. 2006; Chater & Oaksford 2008, for overviews), it is worth trying to remind oneself just how much has changed in cognitive science in that period. At the beginning of the 1990s, logic was taught extensively in philosophy and cognitive science courses, but little or no mention was given to probability theory; now core topics such as epistemology are incomprehensible without a basic understanding of probability. Cognitive science as a field was still in the throws of the symbolic versus connectionist debate (e.g., Fodor & Pylyshyn 1988; Smolensky 1990) and both Nick Chater and Mike Oaksford were involved in the connectionist project (e.g., Bullinaria & Chater 1995; Oaksford & Brown 1994). Arguably, the rise of connectionism was a major factor in cognitive science’s statistical “turn” (e.g., Chater 1995), but connectionism’s emphasis was firmly on “mechanism” and its (potentially) “brainstyle” processing. However, since then, the connectionist project has stagnated by comparison to what has been achieved in other areas, such as the psychology of reasoning. Crucial to this latter success, it seems, was the step back, or better, “upwards,” in terms of the degrees of abstraction, and levels of description at which explanation were sought. Retreating from considerations of implementation, or, by and large, even algorithms, O&C have stuck resolutely to computational level descriptions, and, in so doing, have become key exponents of Anderson’s project of rational analysis (Anderson 1990; Chater & Oaksford 2008; Oaksford & Chater 1998b). To this day, computational level descriptions provoke suspicion in mainstream experimental psychology. “But what is it that people are actually doing?”, “what is going on in the head?” are standard questions to researchers focused on the computational level, as is exemplified in the dialogue between Sceptic and Advocate in the final chapter of O&C’s new book, Bayesian Rationality (Oaksford & Chater 2007, henceforth BR).

Commentary/Oaksford & Chater: Pre´cis of Bayesian Rationality All are agreed that the project of understanding human behavior is not over until processes and their implementation have also been specified, but the thing that comes across most strikingly in O&C’s book is the level of explanatory specificity that has now been achieved in the psychology of reasoning. Before O&C’s seminal paper on the selection task (Oaksford & Chater 1994), data in the psychology of logical reasoning were a (in many ways somewhat haphazard) collection of qualitative phenomena (“context effects,” “supression effects,” etc.). Reading O&C’s summary of the state of affairs 14 years on, the psychology of reasoning has become an arena in which detailed quantitative predictions are evaluated. This is not only true of their own probabilistic approach; O&C have also forced the hand of rival approaches which have since taken on board quantitative model evaluation (Klauer et al. 2007; Oberauer 2006; Schroyens & Shaeken 2003). It is worth pausing to consider how remarkable this is. The reasoning tasks in question are firmly “high-level cognition” of the kind that is characteristic of what Fodor (1983) considered to be “central processing,” and hence an aspect of human cognition for which we would never have detailed theories and predictions due to the inherent open-endedness of high-level thought. That the field can capture subtle changes in behavior in response to changes in the content of high-level, verbal experimental materials in such detail is truly a success. Moreover, this specificity has been spreading through other aspects of human reasoning, as well (e.g., Tenenbaum et al. 2007). That greater predictive power with regard to human behavior might be achieved by taking a step backwards to a higher level of abstraction that disregards processes might seem counterintuitive, but it seems to be exactly what has happened in the psychology of reasoning. One further aspect of O&C’s work deserves mention in this context. While it has, in one way, considered “less” than has other research in the psychology of reasoning, there is also a way in which it has consistently considered “more”: Apparent throughout O&C’s book is the multi-disciplinarity of their project which has drawn support from both philosophy and Artificial Intelligence (AI). We disregard developments in adjacent disciplines at our peril: The problems of associating natural language if. . .then statements with the material conditional were long known in philosophy, but ignored by psychologists of reasoning (see Edgington 1995; Evans & Over 2004). Likewise, developments in AI spoke to the feasibility of the “logicist” program. Yet interdisciplinarity continues to be less of an asset and more of a stumbling block for publication in many of the core psychology journals that have housed the psychology of reasoning. At the same time, cognitive science as a discipline seems, if anything, to have lost in appeal over the last decade. Certainly in the United Kingdom, the number of degree courses in cognitive science has gone down – a trend that contrasts with the booming interest in cognitive neuroscience. Cognitive neuroscience, of course, seems first and foremost concerned with processes and their implementation. Although it undoubtedly will lead to results that are new and exciting in their own right, it seems worth pointing out that this trend is a move in the opposite direction to that taken so successfully by O&C; this should give pause for thought to those interested in high-level cognition.

Complexity provides a better explanation than probability for confidence in syllogistic inferences doi:10.1017/S0140525X09000363 Graeme S. Halford School of Psychology, Griffith University, Nathan, Queensland, 4111, Australia. [email protected]

Abstract: Bayesian rationality is an important contribution to syllogistic inference, but it has limitations. The claim that confidence in a conclusion is a function of informativeness of the max-premise is anomalous because this is the least probable premise. A more plausible account is that confidence is inversely related to complexity. Bayesian rationality should be supplemented with principles based on cognitive complexity.

The Bayesian account of reasoning proposed by Oaksford & Chater (O&C) in Bayesian Rationality (Oaksford & Chater 2007, henceforth BR) extends the historic trend away from regarding logic as the science of thought, and toward seeing reasoning as adaptation to the environment. Piaget, who is cited on page 4 of the book as wholeheartedly adopting a logicist conception of the mind, was in fact part of this trend, and he expressly repudiated logicism (Piaget 1957, p. 1). The Bayesian account arguably has potential to revolutionise our understanding of reasoning, but it has limitations. I want to draw attention to some issues that are additional to those that have been identified in the literature. The heuristics proposed by the Probability Heuristics Model (PHM), that is, min, max, and attachment, operate on an implicit ranking of the informativeness of the four premise moods (All, Some, Some not, None). According to the min-heuristic, individuals tend to draw conclusions that match the mood of the least informative premise, which by information theory is also the most probable proposition in a set of premises. The validity of the min-heuristic is that it reflects constraints between the mood of the min-premise and the mood of the valid conclusion (see Tables 7.2 and 7.3 in BR). Thus, PHM implies that reasoning depends on knowledge of constraints contained in world knowledge. PHM accounts for approximately the same proportion of variance in problem forms as Mental Models theory and Relational Complexity theory (Halford et al. 2007). PHM goes beyond these theories in dealing with premises “Most” and “Few,” but it only predicts mood of the conclusion, not actual conclusions, and it depends on estimated parameters. There is also a significant anomaly. The max-heuristic determines confidence in the conclusion in proportion to the informativeness of the max-premise, but the most informative premise is the least probable. There does not appear to be an explicit explanation for confidence being inversely related to probability, whereas selection of conclusions, by the min-heuristic, is directly related to probability. An alternative hypothesis is that high confidence is associated with lowest complexity, because the least probable forms according to PHM are also the least complex according to complexity metrics, including number of mental models, and the relational complexity metric. The simplest syllogisms according to the Relational Complexity metric (Halford et al. 2007), those based on binary or ternary relations, have at least one A (All) premise, or at least one I (Some) premise. A (All) premises are the most informative (least probable) and I (Some) premises are second most informative in PHM. Thus, the most informative, and least probable, syllogisms are of low complexity, which is a more plausible basis for confidence than low probability. Therefore, PHM might work better if it incorporated other principles that have been established in cognitive science, including those that define cognitive complexity of reasoning tasks.

Are stomachs rational? doi:10.1017/S0140525X09000375 Elias L. Khalil Department of Economics, Monash University, Clayton, Victoria 3800, Australia. [email protected] www.eliaskhalil.com BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

91

Commentary/Oaksford & Chater: Pre´cis of Bayesian Rationality Abstract: Oaksford & Chater (O&C) would need to define rationality if they want to argue that stomachs are not rational. The question of rationality, anyhow, is orthogonal to the debate concerning whether humans use classical deductive logic or probabilistic reasoning.

In Bayesian Rationality (Oaksford & Chater 2007, henceforth BR), the authors offer a definitive answer to what seems obvious to them: Although our stomachs can be “optimally efficient” in digesting the food in our environment, we still would not describe them as rational (BR, p. 38). For Oaksford & Chater (O&C), the question of rationality of stomachs does not even arise: Stomachs are not information-processing devices, have no beliefs/desires/knowledge, and make no decisions/ inferences. O&C use the stomach example in the context of pressing their distinction between rationality and optimization: Optimization is about a “local” end (such as digestion), whereas rationality “requires not just optimizing something but optimizing something reasonable” (p. 37). What is ironic about O&C’s “rationality analysis” approach, borrowed from Anderson (1990), is that their instrumental approach prohibits them from offering such an answer about stomachs. In fact, in the same context, they explicitly state that they would not offer an answer to the “ultimate question of what rationality is” (BR, p. 38). That is, if they have no definition of rationality, how could they answer at the abstract level that organs (such as stomachs) can be efficient, but not rational, while human organisms can be rational? To be clear, O&C recognize that humans exhibit rationality in everyday activity that allows us to discern between good and bad reasoning. O&C, therefore, want to avoid the empirical approach that elevates everyday behavior to the entry point of theorizing, as the literature expounding the Allais paradox attempts to do. O&C correctly note that the empirical approach leads to chaos: If everyday activity is taken as the entry point, there would be no ground to distinguish between good reasoning from bad reasoning. On the other hand, O&C want to avoid a definition of rationality. They do not want to invite what I call here the “specter of Plato”: O&C are afraid of invoking the pitfalls of the logical approach that postulates deductive logic without being informed by everyday, empirical activity. The logical approach, O&C argue, invites the questions of why humans should follow deductive logic and from where such logic arises. Of more importance to the authors’ project, experimental results have shown that humans perform poorly in classical logical tasks such as conditionals, Wason selection task, and syllogisms. O&C instead argue that humans reason in a world full of uncertainties and, hence, humans use natural language conditionals and probability assessment of premises when they make inferences. Humans do not use meaningless logical conditionals and deductive reasoning with premises that stand with absolute certainty. Let us agree with O&C that the probabilistic approach to human reasoning explains why humans, in laboratory settings, are bad at solving tasks formulated by classical logic. But, still, why does the advocacy of a probabilistic approach prevent one from providing a definition of rationality – a definition that is necessary if one wants to argue that stomachs are not rational? It seems that the issue of defining rationality is orthogonal to the debate between the logical approach and the probabilistic approach to human reasoning. Both approaches are about resolving the issue of how people can arrive at true beliefs – whether the truth is defined in absolute certainty or in degrees of belief. But the issue of rationality is not about how reasoning can arrive at truth. Rather, rationality is about the use of resources in the most efficient way (Becker 1976, Ch. 1). Organisms are not mainly concerned with becoming professional scientists, where costs of the truth are disregarded. Rather, organisms are mostly interested in harnessing food/energy from their environment in the most effective way in order to enhance their well-

92

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

being or, what is the same thing, to expand the quantity/ quality of offspring. So, the main goal of organisms, with the exception of professional scientists, is not reasoning in order to arrive at true beliefs – even when truth is recognized as degrees of belief. Rather, the main goal of organisms is reasoning that enhances well-being (or fitness). Such a goal entails that organisms may settle for inaccurate hypotheses or beliefs given that more accurate ones can be unjustifiably costly. That is, if the processing of new data – which may upset an older hypothesis or belief – is too costly, given the expected benefit of a new hypothesis of belief, then humans would be better off with the old, incorrect hypothesis. This elementary definition of rationality stands (or falls) orthogonally to the issue of whether agents use classical logical methods or probabilistic methods. And O&C can adopt a definition of rationality without adopting the classical logical method or invoking the specter of Plato. To wit, their insistence, on more than one occasion, that stomachs are not rational implies that they uphold a definition of rationality. O&C must implicitly view rationality as characteristic of the actor, and not merely, appealing to the work of Marr (1982), as an “explanatory tool” (BR, p. 35). The idea that rationality is merely a tool, called generally instrumentalism and advocated by van Fraassen (1980, Ch. 2; see also Friedman 1953), is critical of realism, which in this case means that rationality is characteristic of the decisionmaker. On the view of rationality as efficiency, we only need to assume, at least initially, that preferences are consistent (Kreps 1990), and that agents change their actions in response to a change in the environment (constraints). There is no need to suppose that the actor or entity has a cognitive ability informed by beliefs/desires/knowledge. As long as an entity changes behavior in response to the environment in order to economize on effort expenditure, it is rational. Behavioral ecologists, ethologists, and philosophers have recently started to realize that organisms do process information and respond to the environment in ways that qualify them as rational (e.g., Griffin 1992; Hurley & Nudds 2006; Raby et al. 2007; Vermeij 2004). Some of these organisms have no brains, such as plants and single-cell organisms. The same reasoning can be applied to organs (including stomachs) insofar as organs change pace of function, such as secretions or size, in response to changes in the environment (constraints). There is no need, therefore, to distinguish efficiency from rationality – when rationality is about optimizing “something reasonable.” Digestion seems also a reasonable goal. In fact, the efficiency/rationality distinction would beg the question of how to distinguish a reasonable from an unreasonable goal, that is, it would invite circularity of argument, as O&C clearly admit (BR, p. 37).

Is the second-step conditionalization unnecessary? doi:10.1017/S0140525X09000387 In-mao Liu Department of Psychology, National Chung-Cheng University, Chia-Yi 621, Taiwan. [email protected]

Abstract: Because the addition of the conditional premise tends to increase modus ponens (MP) inferences, Oaksford & Chater argue that the additional knowledge is assimilated to world knowledge before the Ramsey test is carried out to evaluate P(qjp), so that the process of applying the Ramsey test could become indistinguishable from the process of applying the second-step conditionalization.

In Bayesian Rationality (Oaksford & Chater 2007, henceforth BR), Oaksford & Chater (O&C) argue that the second-step

Commentary/Oaksford & Chater: Pre´cis of Bayesian Rationality conditionalization is unnecessary in the MP (modus ponens) argument. However, based on experimental findings, I show that the second-step conditionalization in the MP argument is sometimes necessary and sometimes unnecessary. I further argue that the second-step conditionalization is unnecessary because the conditional probability hypothesis holds. The conditional probability hypothesis, which is implied by the Ramsey test (see Over et al. 2007), states that the subjective probability of a natural language conditional is the subjective conditional probability. I will show that the conditional probability hypothesis holds only when reasoners explicitly evaluate the probability of conditionals, but that it may not hold in the context of making MP inferences. The second-step conditionalization is experimentally isolated from the first-step conditionalization by adding the reduced problem to the complete problem. The reduced MP problem is of the form: Given p, how probable is q? The complete MP problem is the original MP problem of the form: If p then q; given p, how probable is q? Thus, reduced MP problems measure the result of the first-step conditionalization, whereas complete MP problems measure the result of performing both steps in succession. To argue that the second-step conditionalization is unnecessary is tantamount to arguing that complete MP problems can be replaced by reduced MP problems. In other words, the major premise does not play its role in a complete MP problem. Thus, only studies that administered both reduced and complete problems in the same experiments can provide evidence of whether the major premise does play a role in complete MP problems. There are several studies – for example, those of Chou (2007), Liu (2003), Liu and Chou (2008), Liu et al. (1996), and Wang (1999) – that manipulated several variables to study their effect on reduced and complete MP problems in the same experiments. Three variables are known to affect reduced MP problems differently from complete MP problems. The first variable, perceived sufficiency, affects complete problems only by affecting the reduced problems. Because of a ceiling effect, reasoners’ responses in solving the reduced and complete MP problems are generally identical in the high sufficiency condition. The second variable, problem content, affects the complete MP problem without affecting the reduced MP problems. The third one, age differences, is a quasi-variable: Primary school children raised in a rural area of Taiwan, when tested several years ago, exhibit identical responses in solving reduced and complete MP problems, because they are still unable to perform the second-step conditionalization. O&C realize that adding the conditional premise tends to increase endorsement of the inferences. However, to account for the Ramsey test (1931/1990a), they argued that the additional knowledge the conditional premise provides is assimilated to world knowledge before the Ramsey test is carried out to evaluate P(qjp). The Ramsey test (1931/1990a) is compatible with recent experimental findings that tend to support the conditional probability hypothesis (Evans et al. 2003; Oberauer & Wilhelm 2003; Over et al. 2007). On the surface, however, the conditional probability hypothesis is incompatible with the vast literature on MP findings. In particular, the MP is nearly perfectly endorsed when the conditional statement involves abstract content (e.g., Evans 1977; Taplin 1971; Taplin & Staudenmayer 1973). More specifically, Evans (1977) used the following conditional to generate MP problems: If the letter is G, then the number is 9. The conditional probability of the number 9, given the letter G, would be extremely low. Since Evans found that his participants perfectly endorsed MP problems generated by this type of abstract conditional, the conditional probability interpretation of a conditional statement becomes questionable. Working with conditionals phrased in terms of thematic materials, such as “If Mary has an essay to write, she will stay late in the library,”

Byrne et al. (1999) also observed a mean of 95% endorsement of simple MP inferences in their experiment. Again, the conditional probability of Mary staying late in the library, given that she has an essay to write, would generally be small. Faced with this complex array of empirical findings, O&C argue that the conditional-statement premise provides additional evidence that q and p are related, thus increasing the assessment of P(qjp). They then argue that this additional knowledge is assimilated to world knowledge before the Ramsey test is carried out to evaluate P(qjp). Empirically, there is an increase in P(qjp) in conditionalizing the result of the first-step conditionalization on the conditional-statement premise when reasoners are able to detach it from reality (Liu 2003). In other words, although P(qjp) observed from the reduced MP problems of abstract and thematic contents could be the same, the response level observed from the complete MP problems is generally higher for abstract than for thematic content. On the basis of this empirical finding, it is impossible to know whether the process of applying the Ramsey test (1931/1990a) is the same as the process of applying the second-step conditionalization. The fact that reasoners might consider the antecedent as sufficient for the consequent in reading the conditional-statement premise (e.g., Byrne et al. 1999; Evans 1977) may not contradict the conditional probability hypothesis. This is because the former finding is observable in the context of MP inferences, whereas the conditional probability hypothesis receives its support through the reasoners’ direct evaluation of the conditional statement. ACKNOWLEDGMENTS This study was supported by NSC Grant 96-2413-H-194-010-MY2. I thank Peter Schonemann for reading an early draft.

Bayes plus environment doi:10.1017/S0140525X09000399 Craig R. M. McKenzie Rady School of Management and Department of Psychology, University of California, San Diego, La Jolla, CA 92093-0553. [email protected] http://psy.ucsd.edu/  mckenzie/

Abstract: Oaksford & Chater’s (O&C’s) account of deductive reasoning is parsimonious at a local level (because a rational model is used to explain a wide range of behavior) and at a global level (because their Bayesian approach connects to other areas of research). Their emphasis on environmental structure is especially important, and the power of their approach is seen at both the computational and algorithmic levels.

Oaksford & Chater (O&C) are to be commended for their comprehensive account of deductive reasoning in Bayesian Rationality (Oaksford & Chater 2007, henceforth BR). They don’t just explain Wason’s selection task, or conditional reasoning, or syllogistic reasoning, but all of these. Furthermore, the fact that their account is Bayesian is important for several reasons. First, an inductive account of deductive behavior is interesting and novel. Second, because the pattern of data which they explain consists largely of deductive errors, they are providing an alternative rational view of these “errors.” Third, because their account is both rational and explains behavior in these tasks, there is an important sense in which readers (such as myself) gain an understanding of why people respond as they do in these tasks. The explanation is, essentially, a teleological one. Finally, the approach connects deductive reasoning to other areas of psychological research that are otherwise disconnected, such as vision, categorization, and language, all of which are influenced by BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

93

Commentary/Oaksford & Chater: Pre´cis of Bayesian Rationality Bayesian approaches. In short, O&C offer an account that is parsimonious both locally, in the sense that it is a rational framework that can explain a wide range of deductive behavior, and globally, in that it ties in with a much broader collection of diverse research topics. Especially important is the environmental aspect of the “rational analysis” that they pursue (Anderson 1990). Researchers tend to point to the many examples of purportedly irrational behavior as evidence against a rational approach (e.g., Gilovich et al. 2002; Kahneman et al. 1982; Kahneman & Tversky 2000). Often overlooked by researchers, though, is that people make strong and reasonable assumptions about the structure of their environment. Unaware of these assumptions, researchers can draw misleading conclusions about behavior. For example, O&C point out that people appear to make the “rarity assumption,” that is, that named categories tend to be small (see also McKenzie & Amin 2002; McKenzie & Mikkelsen 2000; 2007; McKenzie et al. 2001). Without the rarity assumption, some behavior can appear irrational and even nonsensical, but with it, behavior is not only reasonable, it is consistent with a Bayesian approach. Similarly, McKenzie and Mikkelsen (2007) recently argued that the rarity assumption explains why people, when assessing the correlation between two variables that can be either present or absent, are especially influenced by observing the joint presence of the variables. Traditionally, the large influence of joint presence observations has been seen as an error, but it makes sense from a Bayesian perspective combined with the rarity assumption. A completely different example of the importance of environmental structure comes from framing effects, which refer to situations in which listeners respond differently to logically equivalent utterances (or frames). Framing effects have been considered irrational, but it has been pointed out that speakers do not choose randomly among frames. Instead, speakers choose frames systematically, and listeners know this. Thus, although many frames are logically equivalent, they are not information equivalent, and responding differently to them is not irrational (McKenzie 2004; McKenzie & Nelson 2003; Sher & McKenzie 2006; 2008). Understanding the structure of the environment, and what makes sense in light of this structure, goes a long way toward understanding people’s behavior. The power of rational analysis is apparent at the computational level (using Marr’s [1982] terminology), where the aim is to understand what problem the organism is trying to solve. O&C’s accounts of Wason’s selection task and conditional reasoning are essentially computational. But when O&C discuss syllogistic reasoning and their probability heuristics model, they also show what rational analysis can do at the algorithmic level, which is concerned with the processes used by the organism to solve the problems it faces. The rational model remains the same, but it becomes clear that behavior, at this level, diverges from the model. Relative to the rational model, people use simplified processes (i.e., heuristics) that lead to systematic errors (biases; e.g., Kahneman et al. 1982). One must be careful here because, as others have pointed out, the “heuristics and biases” paradigm in judgment and decision making has left that field with a “laundry list” of heuristics and biases rather than an overarching theory (Gigerenzer 1996; Krueger & Funder 2004). However, rational analysis severely constrains the set of potential heuristics that people might use, which, to a large extent, would guard against the explosion of heuristics that has plagued the judgment and decision making literature. Furthermore, because this constrained set of plausible heuristics would come from a rational analysis, the adaptiveness of heuristics would be less likely to be a contentious topic, as it often is in the field of judgment and decision making. It would be fascinating (and a little ironic) if rational analysis were to pave the way for the next step in the heuristics and biases paradigm.

94

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

Naı¨ve optimality: Subjects’ heuristics can be better motivated than experimenters’ optimal models doi:10.1017/S0140525X09000405 Jonathan D. Nelson Max Planck Institute for Human Development, Berlin 14195, Germany. [email protected] http://jonathandnelson.com/

Abstract: Is human cognition best described by optimal models, or by adaptive but suboptimal heuristic strategies? It is frequently hard to identify which theoretical model is normatively best justified. In the context of information search, naı¨ve subjects’ heuristic strategies are better motivated than some “optimal” models.

Bayesian Rationality (Oaksford & Chater 2007) nicely synthesizes the growing body of research. This approach offers promise to give principled explanation not only of human higher cognitive processes, but also much of perception and animal cognition. Is cognition best described by optimal models, or by adaptive but suboptimal heuristic strategies? Oaksford & Chater (O&C) explain that heuristic strategies approximate human syllogistic reasoning better than optimal Bayesian models. This commentary illustrates, in the context of information search, that it is frequently possible to be mistaken about what model is best motivated; and that naı¨ve heuristic strategies can perform better than “optimal” models! Consider the task of deciding which of two medical tests to order, assuming cost constraints only allow one test, to best diagnose a patient’s disease. We assume here that the patient either has Disease 1 or Disease 2, with equal (50 %) prior probability, and that Test 1 and Test 2 are each either positive or negative. How should a diagnostician decide which test to order? A great deal of cognitive psychological and statistical thinking (since I. J. Good’s [1950; 1975] work) claims that the optimal strategy is to conduct the test with highest expected Bayesian diagnosticity (expected weight of evidence) or expected log diagnosticity. [T1 ¼ p denotes that Test 1 is positive, T1 ¼ n that Test 1 is negative, D ¼ d1 that the patient has disease 1, etc.] As shown in Figure 1, the expected Bayesian diagnosticity for Test 1 (i.e., its expected utility, as measured with Bayesian diagnosticity) is:

Figure 1. and, as shown in Figure 2, its expected log diagnosticity is:

Figure 2. However, many subjects follow the feature difference strategy (Nelson 2005; Skov & Sherman 1986; Slowiaczek et al. 1992). This strategy involves calculating the absolute difference in feature likelihoods for each test; for example, as shown in Figure 3:

Commentary/Oaksford & Chater: Pre´cis of Bayesian Rationality

Figure 3.

and ordering the test with the highest fDiff. Thus, Skov and Sherman and Slowiaczek et al. concluded that many subjects use a suboptimal heuristic strategy that is highly correlated with the optimal strategy. Remarkably, however, both the claims (1) that Bayesian diagnosticity (and/or log diagnosticity) are theoretically optimal, and (2) that the feature difference strategy only imperfectly approximates optimal behavior, are in disrepute. Both expected Bayesian diagnosticity and expected log diagnosticity are poorly behaved as optimal models. To illustrate, suppose that Test 1 were positive in 99% of people with Disease 1, and in 100% of the people with Disease 2. Suppose further that Test 2 were positive in 1% of people with Disease 1, and 99% of people with Disease 2. Test 1 leads, on average, to 50.5% probability of identifying the correct disease; Test 2 leads, on average, to 99% probability of correctly identifying the true disease. Clearly, Test 2 would be more helpful than Test 1 to differentiate between the diseases. Yet diagnosticity and log diagnosticity maintain that Test 1 is infinitely more useful than Test 2! Both diagnosticity measures hold that any test that offers greater-than-zero probability of obtaining 100% certainty of the true disease is infinitely useful. This bizarre claim is not a desirable property of an “optimal” model. (In Nelson [2005; 2008] I discuss these and other theoretical flaws with the diagnosticity measures, and how redefining a single point cannot fix them.) Better-motivated theoretical models of the value of information, such as information gain-KL distance (Lindley 1956; Oaksford & Chater 1994), probability gain (error reduction; cf. Baron’s 1981 talk at the Psychonomic Society Meeting, as cited in Baron 1985), and impact (Klayman & Ha 1987, pp. 219–20; Nelson 2008; Nickerson 1996; Wells & Lindsay 1980) behave reasonably in this medical diagnosis scenario, and do not suffer from the diagnosticity measures’ aforementioned theoretical flaws. Does the feature difference strategy also approximate these better-motivated theoretical models? In fact, it exactly corresponds to impact! The highest fDiff feature also has the highest impact, irrespective of the prior probabilities of the diseases and the specific feature probabilities (Nelson 2005, footnote 2). Closer analysis of the supposedly optimal theoretical models used by some experimenters, and the supposedly suboptimal heuristics used by some subjects, showed that the subjects’ heuristic strategy corresponds to a normative model (impact) that is theoretically superior to the normative model that the experimenters had in mind! Put in the context of Marr’s (1982) levels of analysis, consideration of subjects’ behavior at the algorithmic level can inform thinking about the kinds of computational-level models (normative theories) that are most appropriate (also see Chater et al. 2003; Cohen 1981). Do all subjects use the feature difference strategy? No. As O&C discuss, the means with which information is presented is important. Different people use a variety of strategies, especially when environmental probabilities are presented in the standard probability format, with explicit prior probabilities and likelihoods. The standard probability format is not the most meaningful to subjects; frequency formats better facilitate Bayesian reasoning (Cosmides & Tooby 1996; Gigerenzer & Hoffrage 1995). Personal experience of environmental probabilities may be even more effective. When environmental probabilities are learned through personal experience, the vast majority of subjects maximize the probability of a correct guess (probability gain), rather than impact or information gain (Nelson et al., submitted). Note that impact (which the feature difference strategy implements) more reliably approximates probability gain than do Bayesian

diagnosticity or log diagnosticity (Nelson 2005), and impact is easily calculated when the standard probability format is used. Is cognition optimal? Adaptation can be impressive. Insects’ flight length distributions appear well-calibrated to natural environments (Viswanathan et al. 1999). But the modern world is evolutionarily novel. For instance, sugar, fat, and salt are available in unprecedented abundance. Similarly, modern media may exaggerate the incidence of plane crashes versus car crashes, or terrorism versus heart disease. The increasing rate of human genetic evolution (Hawks et al. 2007) may facilitate adaptation to some modern environments, over phylogenetic time. Among topics of interest in Bayesian and rational analysis, such as perception (e.g., Hoffman, in press), memory, information search, and category formation, the correct function to optimize is seldom clear. Baron (2004) noted that utilities are formed on the basis of reflection, and are constantly being modified. As a pragmatic matter, cognitive science would be wise to treat candidate normative models in similar fashion (also see McKenzie 2003). When there are clear and robust discrepancies between human behavior and a particular theoretical model, the normative status of the theoretical model should be reconsidered, as well as the rationality or adaptiveness of the human behavior. ACKNOWLEDGMENTS The author thanks Mike Oaksford and Nick Chater for their insightful suggestions. The writing of this commentary was supported by NIH MH57075 (Garrison W. Cottrell, P.I), when the author was at the Computer Science and Engineering Department, UCSD, and at the Computational Neurobiology Lab, Salk Institute.

Oaksford & Chater’s theory of reasoning: High prior, lower posterior plausibility doi:10.1017/S0140525X09000417 Klaus Oberauer Department of Experimental Psychology, University of Bristol, Bristol BS8 ITU, United Kingdom. [email protected] http://psychology.psy.bris.ac.uk/people/klausoberauer.htm

Abstract: Oaksford & Chater (O&C) subscribe to the view that a conditional expresses a high conditional probability of the consequent, given the antecedent, but they model conditionals as expressing a dependency between antecedent and consequent. Therefore, their model is inconsistent with their theoretical commitment. The model is also inconsistent with some findings on how people interpret conditionals and how they reason from them.

In Bayesian Rationality (Oaksford & Chater 2007, henceforth BR) the authors present a strong theoretical case for the “probabilistic turn” in the psychology of reasoning. I agree with much of the general thesis of the book: People often reason from uncertain information, and they do so by drawing on probabilistic information. Conditionals, which form the backbone of much of our knowledge, express conditional probabilities. I disagree with Oaksford & Chater (O&C), however, in details of their models of how people reason, and I am less sanguine about the evidence supporting these models. I focus on reasoning with conditionals. O&C’s model of reasoning from conditionals is based on a contingency table of the antecedent (A) and the consequent (C). One axiom of their model is that the marginal probabilities, P(A) and P(C), must be constant when the degree of belief in the conditional changes. This is an unfortunate assumption, for two reasons. First, it is implausible. Assume a new drug X is tested, and it turns out that it causes headaches. Thus, we increase our belief in “If a person takes X then they get a headache.” To accommodate the increase in P(headachejX) in one’s subjective contingency table, one can either revise P(headachej : X) BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

95

Commentary/Oaksford & Chater: Pre´cis of Bayesian Rationality down to hold P(headache) constant, or else revise P(headache) up while holding P(headachej : X) constant. The latter appears more reasonable – as drug X is taken by more people, the overall rate of headaches will increase, but the probability of headaches in those who refrain from taking X will not change. Revising P(headachej : X) down would lead to the absurd conclusion that, when many people take X, those who don’t will benefit because they get fewer headaches. Second, holding P(C) constant links the conditional to the probabilistic contrast, that is, the difference between P(CjA) and P(Cj :A). With P(C) held constant, every increase in belief in the conditional, that is, every increase in P(CjA), must be accompanied by a decrease in P(Cj : A), resulting in an increased probabilistic contrast. As a consequence, there is an ambiguity in O&C’s model on what a conditional means. Initially, O&C endorse “the Equation,” that is, the probability of “If A then C” equals P(CjA), and is independent of P(Cj :A). But later, O&C seem to endorse the view that a conditional is believable to the degree that the probabilistic contrast is high. For instance, they argue that “it is possible to believe a rule strongly that has many exceptions” (BR, p. 190), as long as the probabilistic contrast is high, such as “If a child walks home from school, it is abducted.” In line with this reasoning, O&C introduce the “independence model” as the alternative to a conditional hypothesis. The independence model means that P(CjA) ¼ P(C), which implies that the probabilistic contrast is zero. Since the independence model is meant to be the alternative to the conditional, they cannot both have high probability. If the conditional is defined by the Equation, however, P(CjA) can be high and at the same time be equal to P(C). For example, the probability of arriving safely on a flight, given one has a window seat, is very high, but not different from the unconditional probability of arriving safely. It follows that the independence model cannot, in general, be the alternative hypothesis to a conditional when the latter is defined by the Equation. My colleagues and I tested whether people interpret conditionals as simply expressing a high P(CjA) or as expressing a high probabilistic contrast. We found that people’s degree of belief in a conditional depended only on P(CjA), not on P(Cj : A), in agreement with the Equation but not with the probabilistic contrast model (Oberauer et al. 2007). This finding demands a revision of O&C’s argument in defence of the MPMT asymmetry (i.e., the finding that people endorse modus ponens more readily than modus tollens) and their explanation of the Wason selection task, which both assume that the independence model is the alternative to the conditional hypothesis. The evidence for O&C’s model of reasoning with conditionals is mixed at best. Evidence comes from three sources: (1) The model fits endorsement rates for the four basic inference forms. Fitting four data points with three free parameters is no convincing accomplishment, though. A richer database is provided by the frequencies of the 16 possible patterns of endorsement or rejection across the four inference forms. I applied seven formal models of reasoning to such pattern frequencies (Oberauer 2006). O&C’s model (Oaksford et al. 2000) provided fits that were worse than all competitors. (2) The model can explain established findings such as negation and suppression effects. Other theories, however, can also explain these effects (Evans & Handley 1999; Markovits & Barrouillet 2002). Therefore, these findings do not support O&C’s model over alternatives. (3) Direct manipulation of probabilities is arguably the most direct and stringent test, because no competing theory predicts the same effects as the O&C model. There are two series of experiments using this method. One provided support for the O&C model (Oaksford et al. 2000), whereas the other did not (Oberauer et al. 2004). O&C dismiss the latter evidence as difficult to interpret in light of “the large number of findings showing probabilistic effects in the conditional-inference task and in the selection task” (BR, p. 204). At least for the inference task, I fail to see this large number of confirmatory findings.

96

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

One difference between the experiments of Oaksford et al. (2000) and those of Oberauer et al. (2004) is that we used the standard deductive instruction, asking participants to judge whether the conclusion follows with logical necessity from the premises, whereas Oaksford et al. simply asked whether one can draw the conclusion. This difference points to a distinction between goals of reasoning, which I think is not sufficiently acknowledged by O&C. The goal of deductive reasoning is to evaluate whether an inference is valid, and in experiments investigating deduction people are instructed accordingly. Most experiments cited in support of probabilistic theories of reasoning, however, ask people to evaluate the soundness of inferences or the truth of conclusions. The (sparse) evidence from direct manipulations of probabilities suggests that people can ignore probabilities when asked to judge validity, whereas they draw on probabilities when asked to rate whether the conclusion is true. Such a modulation of reasoning processes by goals would be entirely rational. To conclude, the probabilistic view on human reasoning has high a priori plausibility, but the version fleshed out by O&C is conceptually ambiguous and not well supported by the data.

Human reasoning includes a mental logic doi:10.1017/S0140525X09000429 David P. O’Brien Department of Psychology, Baruch College, City University of New York, New York, NY 10010. [email protected]

Abstract: Oaksford & Chater (O&C) have rejected logic in favor of probability theory for reasons that are irrelevant to mental-logic theory, because mental-logic theory differs from standard logic in significant ways. Similar to O&C, mental-logic theory rejects the use of the material conditional and deals with the completeness problem by limiting the scope of its procedures to local sets of propositions.

In Bayesian Rationality (Oaksford & Chater 2007, henceforth BR) the authors reject the conception of human reasoning that focuses on logical inferences, arguing that probability theory should be used instead to account for rationality. Given space limitations, I here address only the two most prominent reasons Oaksford & Chater (O&C) present to reject logic, arguing that they fail to appreciate what mental-logic theory actually proposes. First, mental-logic theory (e.g., Braine 1990; Braine & O’Brien 1991; 1998; O’Brien 1993; 2004; O’Brien & Manfrinati, in press) consistently has proposed that mental logic differs from standard logic. O&C equate the logical view of conditionals with the truth table for the material conditional (if p then q is true unless p is true and q is false). Indeed, Oaksford and Chater (2003a) stated that the Braine and O’Brien theory includes the material conditional for if p then q. The problem with their criticism is that Braine and O’Brien consistently argued that the material conditional does not capture psychological reality. Our theory of conditionals consists instead of two schemas: one for modus ponens (MP) and another for conditional proof. The conditional proof schema states that to derive or evaluate if p then q, first suppose p; when q follows from the supposition of p together with other information assumed, one may assert if p then q. The schema is applied with a reasoning program that supposes p and then treats q as a tentative conclusion to be evaluated. When one evaluates a conditional if not p then not q from the premise p or q, one concludes that the conditional is false, even though this evaluation would not follow when treating if as the material conditional (because p might be false). Thus,

Commentary/Oaksford & Chater: Pre´cis of Bayesian Rationality O&C’s criticisms of the logical semantics of the material conditional do not apply to our mental-logic theory. Second, O&C reject logic as a model for human rationality because, they say, logic needs to provide a formal mechanism that could identify all and only those “common sense inferences that people endorse” (BR, pp. 89 – 90), and such a system would be either incomplete or intractable. (See Rosser [1939] on Go¨del’s 1931 theorems concerning incompleteness and undecidability.) After rejecting logic as a psychological model because of the incompleteness problem, O&C propose that cognitive science should instead rely on probability theory by applying it to particular domains (e.g., in reasoning about a game of dice), although they acknowledge that their suggested solution does not resolve the problem of knowledge representation identified in logic systems, because the computations also would be intractable if extended beyond working out probabilities for small problem sets. Furthermore, they fail to consider whether a mental logic might be able to provide a similar solution to their Bayesian solution by limiting the need to make decidability judgments to small sets of propositions. Indeed, Braine and O’Brien (1991) described just such a mental-logic approach. To understand how mental-logic theory deals with which inferences are admissible in constructing lines of reasoning and which are not, consider the reasoning constraints described in Braine and O’Brien (1991). Unlike standard logic, in mental logic nothing follows from a contradiction except the realization that some assumption that led to the contradiction is wrong. Thus, one cannot suppose a proposition that contradicts other propositions already being used in an argument, and one cannot bring a proposition into an argument under a supposition that contradicts the supposition. The nub of the question concerns how one knows whether a supposition, or a proposition being imported into an argument under a supposition, is creating a contradiction (see Braine & O’Brien 1991, p. 184). Clearly, in making such decisions people do not check the whole set of propositions they carry in long-term memory. Further, it is not simply a matter of deciding whether a proposition being considered is true, but whether it still would be true in the argument being constructed under a supposition. O&C, on the one hand, and Braine and O’Brien, on the other, are in agreement that making a global judgment about the consistency of all possible propositions being held in long-term memory is intractable. The mental-logic proposal is to consider the matter only when preparing to introduce a proposition into a line of reasoning, in order to decide whether that proposition would still be true given the supposition. The resolution of this matter does not require one to consider all possible propositions that might be affected by the supposition, for example, to decide the details of the closest possible world, but rather, to consider only the relevant proposition one desires to import into the argument; and the decision criterion concerns whether the reasoner is satisfied that the proposition still would be true under the supposition. When two or more interlocutors are involved, the matter is up for negotiation. O&C come to their conclusion because they view logic in terms of the sorts of systems that were developed by professional logicians beginning with Hilbert, Frege, Russell, and so on, who attempted to ground number theory in logic. A perusal of a history of logic, such as Kneale and Kneale (1962), reveals a more varied sense of what constitutes logic. The founder of Stoic logic, Chrysippus, suggested that even his dog was able to make simple logic inferences; for example, knowing that when it followed a scent on a road and came to a fork, if the scent did not go down one path, it must go down the other. Surely Chrysippus’s dog did not compute all possible inferable propositions in a dog’s possible world to make this disjunction-elimination inference, nor must we infer that the dog was computing Bayesian probabilities. Mental-logic theory assumes that the human logical reasoning repertory developed through bioevolutionary history because our hunter/gatherer ancestors

gained an advantage in making simple, direct logic inferences that did not include assessing the consistency of the set of all possible propositions they could have held. An interest in the decidability of intractable sets of propositions awaited the advent of logic as a profession and is not part of the universal logic repertory.

Uncertain premises and Jeffrey’s rule doi:10.1017/S0140525X09000430 David E. Overa and Constantinos Hadjichristidisb a

Department of Psychology, Durham University, Durham DH1 3LE, United Kingdom; bFacolta` di Scienze Cognitive, Universita` degli Studi di Trento, 38068 Rovereto, Italy. [email protected] [email protected]

Abstract: Oaksford & Chater (O&C) begin in the halfway Bayesian house of assuming that minor premises in conditional inferences are certain. We demonstrate that this assumption is a serious limitation. They additionally suggest that appealing to Jeffrey’s rule could make their approach more general. We present evidence that this rule is not limited enough to account for actual probability judgements.

In Bayesian Rationality (Oaksford & Chater 2007, henceforth BR), Oaksford & Chater (O&C) propose an important probabilistic theory of conditional reasoning. They identify the probability of a conditional, P(if p then q), with the conditional probability, P(qjp). Following the normative literature, they call this identity the Equation, and we will also take it for granted. There is substantial direct evidence for the Equation as a description of people’s probability judgements (Over et al. 2007). O&C take an impressive indirect approach: They show how much they can explain in the psychology of conditional reasoning using the Equation. However, their main account occupies the halfway Bayesian house of assuming that contingent minor premises are certain. The simplest application of O&C’s main account is to Modus Ponens (MP): inferring q from the major premise if p then q and the minor premise p. They assume that the probability of the minor premise is 1. They thus hold that the probability of q after MP, P(q), is P(qjp), with P(p) ¼ 1. Their assumption is not fully Bayesian. For a full Bayesian, committed to strict coherence, a contingent proposition should not have a probability of 1. That would imply that it could never be disconfirmed. O&C also assume that the minor premise is certain in their treatment of other conditional inferences. To illustrate the limitations of this assumption, consider Modus Tollens (MT). This is inferring, for example, “Global warming will not continue” (not-p) from “If global warming continues then London will be flooded” (if p then q) as the major premise and “London will not be flooded” (not-q) as the minor premise. (Over et al. 2007, has evidence that conditionals like this satisfy the Equation.) Looking in detail at O&C’s description of MT in temporal reasoning, we begin (using, we hope, clearer symbols) with the prior conditional probability, Pold(qjp), and the prior marginal probabilities, Pold( p) and Pold(q). They derive a formula for MT, by elementary steps in probability theory, for inferring Pold(not-pjnot-q) from those prior probabilities. They suppose that we later “learn” the minor premise, not-q, and so acquire a new belief state. They take “learn” in the extremely strong sense of meaning that Pnew(not-q) ¼ 1. This allows them to identify the probability of the conclusion of MT, Pnew(not-p), with Pold(not-pjnot-q). They note that we may lower our belief in the major premise of MT when we learn not-q (Stevenson & Over 1995): learning the minor premise sometimes “alters” Pold(qjp) by lowering it. We use, they hold, this “modified” lower probability, along with Pold(p) and Pold(q), in their derived formula to recalculate Pold(not-pjnot-q), and that is BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

97

Commentary/Oaksford & Chater: Pre´cis of Bayesian Rationality then taken as a revised Pnew(not-p). However, when the probability of the minor premise, not-q, becomes 1, the probability of q becomes 0. The probability of the major premise, if p then q, should consequently collapse to 0 (because the conditional probability is then 0), and not merely be “modified” to varying relatively low values. Of course, there is no technical problem in putting Pold(qjp)¼0 into O&C’s formula to recalculate Pold(not-pjnotq). The problem is that making Pold(qjp)¼0 in every case is a serious limitation. (There is a technical problem if the probability of p becomes 0, since that makes the conditional probability undefined. We set this possibility aside, although it is a problem for O&C’s account of denying the antecedent.) Stevenson and Over (2001) found that people had more confidence in some minor premises than in others, depending on the level of expertise of the person who asserted the premise. Their results suggest that, even if people heard an expert on the Thames Barrier assert “London will not be flooded,” they would not assign it a probability of 1, and might not even judge it to have a very high probability. Of course, people sometimes claim that they are “certain” or “sure” of a contingent proposition, but we should interpret them as saying that their confidence in it is relatively high, and not that it can never be disconfirmed. They might, in turn, interpret an experimenter who tells them to assume that a minor premise is “certain” as merely indicating that its probability is high. But what confidence do people have in the conclusion of a conditional inference as they become more (or less) confident of the minor premise? Rightly starting to move on from their halfway house, O&C suggest that a way to answer this question is to adopt Jeffrey’s rule of conditionalization (Evans & Over 1996a; Jeffrey 1983). The relevant form of Jeffrey’s rule for MP is: Pnew(q) ¼ Pold(qjp)Pnew(p) þ Pold(qjnot  p)Pnew(not  p)

The normative status of this rule is debated (Howson & Urbach 1993), but to what extent do people generally conform to it? We and our collaborators investigated this question using categorical inferences (Evans et al. 2004; Hadjichristidis et al., in preparation). An example is inferring “Falcons have the neurotransmitter Dihedron” (q) from “Robins have the neurotransmitter Dihedron” (p). One way to look at these inferences is to see them as having an implicit conditional (if p then q) as a major premise. We used a plausible manipulation of Pnew(p), for example, “Scientists are 80% certain that Robins have the neurotransmitter Dihedron.” We analysed our data with a regression that aimed to predict judgements of Pnew(q) using as predictors Pold(qjp)Pnew(p) and Pold(qjnot-p)Pnew(not-p). Only Pold(qjp)Pnew(p) had a consistent influence on participants’ judgements about Pnew(q). These results imply that O&C are too limited, in their main account, in making Pnew(q) depend only on Pold(qjp). But the results also imply that they are not limited enough in later suggesting, more generally, that Pnew(q) will be determined by the full use of Jeffrey’s rule. The evidence so far is that Pnew(q) will primarily depend on Pold(qjp)Pnew(p) and so, on just the probabilities of the major and minor premises. However, these results are only for one kind of implicit MP inference. The descriptive adequacy of Jeffrey’s rule should be much more extensively investigated.

Mental probability logic doi:10.1017/S0140525X09000442 Niki Pfeifer and Gernot D. Kleiter Department of Psychology, University of Salzburg, A-5020 Salzburg, Austria. [email protected] http://www.users.sbg.ac.at/  pfeifern/ [email protected] http://www.users.sbg.ac.at/  gdkleiter/

98

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

Abstract: We discuss Oaksford & Chater’s (O&C’s) probabilistic approach from a probability logical point of view. Specifically, we comment on subjective probability, the indispensability of logic, the Ramsey test, the consequence relation, human nonmonotonic reasoning, intervals, generalized quantifiers, and rational analysis.

Probability logic investigates probabilistic inference and relates it to deductive and other inferential systems. It is challenging to relate human deductive reasoning to probability logic rather than to logic. Oaksford & Chater (O&C) were the first who systematically investigated human deductive reasoning within a probabilistic framework. Our approach to human reasoning is, in many respects, related to O&C’s. However, our approach is closer to probability logic, especially with respect to the analysis of the experimental tasks. In commenting on Bayesian Rationality (Oaksford & Chater 2007, henceforth BR), we discuss a selection of questions arising from these different perspectives.1 Common everyday reasoning requires us to process incomplete, uncertain, vague, and imprecise information. Artificial Intelligence (AI) has developed many different approaches for uncertain reasoning. Typical examples are belief functions, possibilistic models, fuzzy systems, probabilistic description languages, many-valued logic, imprecise probabilities, and conditional independence models. Of these approaches only conditional independence models are considered in the book. Why do O&C consider probability and not, say, belief functions as a normative reference system? The probabilities are interpreted as subjective probabilities. The theory of subjective probability was conceived by Bruno de Finetti (1974/1975), and further developed by Coletti and Scozzafava (2002), Gilio (2002), and many others. A central concept of subjective probability theory is coherence. A probability assessment is coherent if it cannot lead to sure losses. The theory does not require event structures that are closed under negation and conjunction. Conditional events are primitive. Conditional probabilities are not defined by absolute probabilities. When we talk about P(AjB), why should we assume that we also know P(A and B) and P(B)? These properties make the coherence based probability logic more appropriate as a framework for psychological research than other approaches, including the “pragmatic strategy of using the simplest probabilistic semantics” (BR, p. 75). O&C argue that classical logic should be replaced by probability theory as a framework for human reasoning. This position is too radical. We should not “throw out the baby with the bath water.” Probability theory presupposes logic for operations on propositions. Simple logical inferences like And-Introduction or Modus Ponens (MP) are endorsed by practically all subjects. We do not see a dichotomy between logic and probability. We fully support O&C’s hypothesis, that human subjects understand the indicative “if A, then B” in the sense of a conditional probability, P(BjA), and not as the probability of a material conditional, P(A . B). Many empirical studies corroborate this hypothesis (Evans & Over 2004). In subjective probability theory, conditional events are not truthfunctional. If the antecedent is false, then the truth value of the conditional is undetermined. This corresponds to what is called a “defective truth table.” Considering P(BjA), we would not say the Ramsey Test adds P(A)¼1 to one’s stock of belief. Rather, A is assumed to be true. Probability 1 and the truth value TRUE are not the same. Ramsey’s idea can be explained in the way Lindley (2006) and several others introduce conditional probability. Subjective probabilities are assessed relative to a knowledge base K. The absolute probability P(A) is shorthand for P(AjK), and P(BjA) is shorthand for P(BjA:K). The colon separates the supposition from the knowledge base. The change from supposition to fact does not change the conditional probability, P(BjAC:K) ¼ P(BjA:CK). The core of O&C’s analysis of the conditional inferences (Modus Ponens [MP], Modus Tollens [MT], Affirming the Consequent [AC], and Denying the Antecedent [DA]) is “that the probability of the conclusion . . . is equal to the conditional probability of the conclusion given the categorical premise” (BR, p. 119), P(conclusionjcategorical premise). Thus, the consequence relation

Commentary/Oaksford & Chater: Pre´cis of Bayesian Rationality (denoted by a horizontal line and three dots in the book) is probabilistic. In our approach the consequence relation is deductive. Each premise obtains a probability. The probabilities of the premises are propagated deductively to the conclusion. Because the MP, MT, AC, and DA arguments consist of only two premises, only an interval probability can be inferred (Pfeifer & Kleiter 2005b; 2006). A conclusion with a point probability would require three premises. In this case, however, the argument does not “mimic” the logical forms of MP, MT, AC, or DA. O&C mention probability intervals, but they do not use them. In Chapter 4, O&C oppose nonmonotonic reasoning and probabilistic reasoning, and advocate probabilities. We do not see why both approaches are incompatible. System P (Kraus et al. 1990) is a basic and widely accepted nonmonotonic reasoning system. Several probabilistic semantics were developed for System P (Adams 1975; Biazzo et al. 2005; Gilio, 2002; Hawthorne & Makinson 2007; Lukasiewicz 2005). We observed good agreement between the predictions of the coherence-based semantics and actual human inferences (Pfeifer & Kleiter 2005a; in press a; in press b). O&C were the first who realized the importance of generalized quantifiers in psychology. Murphree (1991) and Peterson (2000) developed logical systems for syllogisms that involve generalized quantifiers (see also Peters & Westersta˚hl, 2006). On p. 219 of BR, O&C note that “without a notion such as p-validity [not in Adams’ sense!], there is no way of defining the correct answers to these generalized syllogisms.” Peterson, however, investigates syllogisms that involve quantifiers like “Most,” “Few,” or fractionated quantifiers like “n/m.” The semantics of the quantifiers works by comparisons of the cardinalities of appropriate sets and by the use of relative frequencies. Thus, Peterson’s semantics can easily be related to a probabilistic interpretation. Moreover, Atmosphere or Matching are easily generalized within this framework. O&C use Bayesian networks to model syllogisms. Each vertex in the network corresponds to a term in the syllogism. The directed arcs represent conditional probability distributions. We are uncertain about the statement that in Bayesian networks the conditional independence is a “standard assumption” (BR, p. 222). Moreover, under the assumption of conditional independence (X independent Z, given Y), there exists only one probabilistic model. Three models (Figure 1, 3, and 4 in Fig. 7.3) are Markov equivalent; only the veestructure (Figure 2 in Fig. 7.3) has a different factorization. It encodes a marginal independence (X and Z are independent). Rational analysis puts the quest for cognitive processes and representations in the second line. This is fine if the normative models fit the empirical data. In this case the models are both normative and descriptive. A good theory requires: (1) a thorough task analysis, (2) a minimum of generality to avoid “adhoceries,” and (3) a connection with other theoretical concepts like language, memory, or attention. ACKNOWLEDGMENTS This work is supported by the European Science Foundation (EUROCORES program “LogICCC”) and by the Austrian Research Fonds (project P20209 “Mental Probability Logic”). NOTE 1. We shall not comment on a number of misprints in formulae or inconsistencies in notation.

Abstract: Severity of Test (SoT) is an alternative to Popper’s logical falsification that solves a number of problems of the logical view. It was presented by Popper himself in 1963. SoT is a less sophisticated probabilistic model of hypothesis testing than Oaksford & Chater’s (O&C’s) information gain model, but it has a number of striking similarities. Moreover, it captures the intuition of everyday hypothesis testing.

Popper’s philosophical principle of falsification is commonly represented as seeking and finding counter-examples to a hypothesis, as Oaksford & Chater (O&C) show in Bayesian Rationality (henceforth BR, Oaksford & Chater 2007, p. 167). Indeed, general laws can be definitively falsified, but not definitively confirmed, according to logic. Hence, counter-examples give us sure knowledge that the putative hypothesis is false. Large parts of Popper’s work are directed at uncovering the failure of confirmation theory. However, philosophers and psychologists rejected the principle of falsification as obviously incompatible with scientific practice and everyday reasoning. In a natural environment, inside or outside the scientific lab, we cannot live with falsified beliefs alone. Falsified ideas are not publishable and they give poor cues about how to cope with the everyday environment (Poletiek 1996; Poletiek & Berndsen 2000). However, as O&C show, old theories of confirmation have also failed to provide a rational account of hypothesis testing. It was not before the introduction of the Bayesian approach in confirmation theory that a coherent alternative to the falsification method could be advanced.1 O&C’s information gain model provides such an account for hypothesis testing that predicts test preferences in accordance with the information they can provide for the truth and falsity of hypotheses. There is, however, an interesting alternative version of the idea of falsification, hesitatingly proposed by Popper himself. It avoids the obvious problems of Popper’s logical falsification principle, and fits in with O&C’s Bayesian approach. (Poletiek 2001). This alternative definition of falsification is putting hypotheses to severe tests. I argue here that Severity of Test (SoT) provides an intuitive statistical model for what people do when they test hypotheses. As Figure 1 below shows, the formal measure for SoT 2 (Popper 1963/1978; for a discussion, see Poletiek 2001) for a piece of evidence e predicted by the hypothesis H is:

Figure 1. What distinguishes SoT from logical falsification, and to what behavior does it correspond? First, SoT is about evidence that confirms the hypothesis put to the test, and not evidence that is expected to falsify it. The severity of a test increases as the prior probability of confirmation e is low and its likelihood under assumption of the hypothesis is high. Minimizing the prior probability of a confirmation corresponds to a falsifying strategy. Second, it can be shown that the measure S is mathematically equivalent to the traditional Bayesian measure for degree of confirmation, that is, the revision of belief in H, after observing e 3 (see Poletiek 2001) (Fig. 2 below):

Popper’s Severity of Test as an intuitive probabilistic model of hypothesis testing doi:10.1017/S0140525X09000454 Fenna H. Poletiek Department of Psychology, Leiden University, 2300RB Leiden, The Netherlands. [email protected]

Figure 2. This equivalence is probably unacceptable for a logical falsificationist, but for modeling hypothesis-testing strategies it is quite helpful. Indeed, it implies that increasing the severity of test, BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

99

Commentary/Oaksford & Chater: Pre´cis of Bayesian Rationality prior to the actual test, will result in an equal amount of increased confirmation after e has been observed. Third, the definition of test severity shows that the probability of a predicted confirmation e and its severity S are inversely related. Hence, everything being equal, as a confirming prediction is less expected a priori, it will provide a better confirmation of the hypothesis. This is in line with the information gain perspective. But, of course, the risk of not finding that confirmation increases as well. Furthermore, if we assume a test with two possible outcomes (a confirmation e and a falsification), the probability of the falsifying outcome increases as P(e) decreases. This uncovers a Popperian paradox: maximizing the probability of a falsification necessarily minimizes its power to reject the hypothesis (Poletiek 2001). Finally, SoT models hypothesis testing without specifying the choices people make, because these choices are assumed to depend on utilities regarding the values of test outcomes. Let us suppose a hypothesis H that generates several predictions with likelihoods 1 (P(ejH) ¼ 1). SoT suggests that choosing a test involves solving the trade-off between the chance of observing a prediction and the degree of confirmation this evidence may provide for H. Depending on considerations that are external to the test, people may either prefer low degrees of confirmation and avoid the risk of getting a falsification, or go for “risky” tests because they need to be very sure about H’s truth. This representation of test choice has been shown to capture the intuition of everyday hypothesis testing (Poletiek & Berndsen 2000). As another example of this intuition, consider how SoT would account for the following adapted selection task. A participant is presented with four cards revealing an A, a K, a 2, and a 7, respectively. Every card has a number on one side and a letter on the other. She is asked to test the rule, if there is an A on one side of a card, then there is a 2 on the other side. In addition, she is told that the cards are sampled from a huge number of cards having either the letter A or the letter K on one side, and any number between 1 and 100 on the other. Which cards would she turn over? SoT predicts that she is interested in confirming results only (“A and 2” cards). The K and 7 cards are not selected. Under reasonable assumptions about the distribution of the letters and the numbers on the cards, it can be calculated that turning the A and finding a 2 card is a more severe test of the rule than turning the 2 and finding an A. But this difference is felt, as well. Indeed, finding a 2 (P(e)) is much less likely (given 100 different numbers) than finding the A, beforehand. Only if the rule is true would we expect a 2. Observing an A (given only two possibilities) on the 2 card, however, may just be chance. The the SoT model not only predicts that we should be more convinced by the first test outcome, we actually feel we should. In summary, O&C’s probabilistic model of hypothesis testing, which provides a sophisticated alternative for Popper’s problematic logical falsification theory of testing, has an interesting precursor developed by Popper himself. However, this alternative falsification theory got little attention, and has been somewhat played down, both by Popper himself and by other falsificationist philosophers of science. Probably because it turned out to resemble too much its rival: confirmation theory.

NOTES 1. See, however, the relevance ratio proposed by Carnap (1950), which basically expresses the Bayesian notion of revision of belief. 2. Notice that the probabilities are applied to evidence, not to the hypothesis, as can be expected in Popper’s frequentist view on probability. 3. The confirming value of a test result is expressed by Popper in his concept of “degree of corroboration” (Popper 1963/1978).

100

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

Let us not put the probabilistic cart before the uncertainty bull doi:10.1017/S0140525X09000466 Guy Politzera and Jean-Franc¸ois Bonnefonb a Institut Jean Nicod, Centre National de la Recherche Scientifique (CNRS), Ecole Normale Supe´rieure, 75005 Paris, France; bCognition, Langue, Langages, Ergonomie (CLLE), Centre National de la Recherche Scientifique (CNRS), Universite´ de Toulouse, 31058 Toulouse, Cedex 9, France. [email protected] http://www.institutnicod.org/notices.php?user¼Politzer top [email protected] http://w3.univ-tlse2.fr/ltc/bonnefon

Abstract: Although we endorse the primacy of uncertainty in reasoning, we argue that a probabilistic framework cannot model the fundamental skill of proof administration. Furthermore, we are skeptical about the assumption that standard probability calculus is the appropriate formalism to represent human uncertainty. There are other models up to this task, so let us not repeat the excesses of the past.

Although research on human reasoning has long been focused on formal tasks, the past 15 years have seen a shift towards more realistic tasks, closer to everyday life. These tasks feature uncertain premises, incomplete information, defeasible conclusions, and the exploitation of individual knowledge bases. Earlier theoretical approaches to deduction were not designed to function in this new paradigm. Oaksford & Chater’s (O&C’s) Bayesian Rationality (Oaksford & Chater 2007) provides a decisive contribution to the adoption of an alternative approach to the investigation of deductive reasoning, in a radical way, by eliminating deductive logic and adopting the Bayesian view of uncertainty. Although we welcome the paradigm shift, we take issue with O&C’s radical solution, on at least two fronts. O&C’s radical probabilistic approach is silent about a fundamental question: Where do conclusions come from? Given a set of premises, how do reasoners produce their own conclusion? Whatever their shortcomings, logic-based theories such as mental models and mental rules have answers to these questions. For mental models, the procedure consists of reading off the models propositions that contain new information. For mental rules, it consists of applying strategies and executing routines that yield a chain of inferences. O&C’s proposal does not offer any procedure that would play this role. The probabilistic model is apt to analyze how a proffered conclusion is evaluated, but not how it is produced in the first place. In other words, a purely probabilistic model lacks the power of executing one basic human skill, the generation of proofs. We believe with O&C that theories of human deduction cannot be purely logical, otherwise they could not account for the nonmonotonic nature of everyday inferences. But, as we just argued, no account of deduction can be purely focused on uncertainty, either, because it would miss fundamental aspects of deductive competence. We believe that the solution to this quandary is to be found in a mixed model, accommodating both uncertainty and deduction. Our second concern with O&C’s Bayesian approach revolves around the nature of human uncertainty. Human beings have the precious metacognitive ability of being aware of their own uncertainty, with respect to the conjectures they entertain. If we wish to develop models of how individuals make judgments and inferences from uncertain information, we need, first, to characterize their metacognitive representation of uncertainty. We believe this question has been largely ignored, because the answer is all too often assumed right from the start. The temptation is strong to take it for granted that standard probability calculus is the only candidate to constitute a normative psychological model of reasoning with uncertainty. Accordingly, the debate has shifted towards the question of the extent to which lay reasoners perform in agreement with the calculus.

Commentary/Oaksford & Chater: Pre´cis of Bayesian Rationality But should they? Consider the requirement of completeness. Are we ready to take it for granted that the credibilities of any two assertions can always be compared? Consider now complementarity. Can we take it for granted that anytime the credibility of a conjecture increases, the credibility of its negation decreases by the exact same amount? Finally, do we take it for granted that credibilities in the mind map onto real numbers on the [0,1] interval, rather than levels on an ordinal scale? The hurried adoption of the probability calculus as a normative model preempts the examination of these built-in assumptions. We are concerned that such a move is putting the probabilistic cart before the uncertainty bull. Important evidence is lacking before the probability calculus (Bayesian or not) can be deemed an adequate formalism for psychological modeling of reasoning with uncertainty, be it at the normative or at the descriptive level. We do not deny that the probabilistic framework is highly successful in many disciplines, but this success does not warrant electing the probability calculus as the adequate psychological model of uncertainty, without carefully examining other options, for there are many other options, as noted by O&C. Intriguingly, this multiplicity is framed as a shortcoming in Bayesian Rationality (p. 73), where the uniqueness of the probability calculus is favorably compared to the multiplicity of nonprobabilistic frameworks. But surely, every uncertainty formalism is unique in its own right, and it seems abusive to throw in the same “nonprobabilistic” bag alternative candidates such as, for example, belief functions (Shafer 1976), possibility measures (Dubois & Prade 1988), relative likelihood (Halpern 1997) and plausibility measures (Friedman & Halpern 1995). In Bayesian Rationality, the main argument for preferring probability theory to all other frameworks is practical: Most psychologists are only aware of probability theory as a model of uncertainty, probability theory fares reasonably well, therefore psychologists should not waste time getting acquainted with other formalisms, which are not expected to fare much better. This argument would make perfect sense in an applied setting, but its relevance is not as clear to fundamental research on human reasoning. First, we will not know whether probability theory fares reasonably well until we compare its performances to that of other uncertainty formalisms. Second, most scientists would probably be perplexed at the thought of adopting the first theory they encounter, without considering even briefly the merits of competing theories. Third, there actually exists a community of psychologists and experimental philosophers investigating the empirical plausibility of alternative uncertainty formalisms (e.g., Benferhat et al. 2005; Ford 2004; Pfeifer & Kleiter 2005). In brief, we know far too little about human uncertainty to definitely adopt a normative model; and this matter should be solved empirically, rather than a priori. In conclusion, we are worried that O&C may repeat within the uncertainty paradigm the double move that doomed the logical paradigm. Part of this move is to adopt an exclusive view, and to reject indispensable concepts that belong to the other paradigm; the other part is to adopt one well-known formalism (formerly the propositional calculus, and currently the probability calculus) and to establish it too hastily as a model of competence.

On is an ought: Levels of analysis and the descriptive versus normative analysis of human reasoning doi:10.1017/S0140525X09000478 Walter Schroyens Department of Psychology, University of Gent, B-1000 Gent, Belgium. [email protected] http://ppw.kuleuven.be/reason/schroyens

Abstract: Algorithmic-level specifications carry part of the explanatory burden in most psychological theories. It is, thus, inappropriate to limit a comparison and evaluation of theories to the computational level. A rational analysis considers people’s goal-directed and environmentally adaptive rationality; it is not normative. Adaptive rationality is by definition non-absolute; hence, neither deductive logic nor Bayesian probability theory has absolute normative status.

In Bayesian Rationality (Oaksford & Chater 2007, henceforth BR) Oaksford & Chater (O&C) present both positive arguments in support of a probabilistic approach and negative arguments against a non-probabilistic “deductive” approach to human reasoning. The negative arguments are comparative in nature and hinge on the presumed acceptability of a direct comparison between the theories’ computational-level descriptions, without taking account of their algorithmic-level specifications. I argue that this central premise is fundamentally mistaken: It is potentially misguiding to compare and evaluate theories on the basis of just their computational-level descriptions. Cognitive psychology is about explaining what people do. Ignoring the algorithmic level is only acceptable when there is a “triumphant cascade” through the levels of description (Dennett 1987, p. 227), that is, when there is a functional match between the input-output function computed at the computational level and the function computed at the algorithmic level. In these cases, there is no added explanatory value in going from the computational level to the algorithmic level and, thus, no loss in explanatory value by not considering the algorithmic level. Cognitive theories are, however, swamped with functional mismatches between the levels (Franks 1995; 1999). It follows that when one critiques the idealized computational-level description of such theories on the basis of their explanatory (in)adequacy, one creates a straw-man argument, that is, a negative argument aimed at an oversimplified strawman version of a theory (Schroyens, in press). One could try to justify a comparison at the computationallevel descriptions of theories by assuming that they are normative. The underlying assumption is that a theory can be normatively justified on the basis of what people actually do: “the apparent mismatch between normative theories and reasoning behaviour suggests that the wrong normative theories may have been chosen; or that the normative theories may have been misapplied” (BR, p. 30). O&C take the poor explanatory adequacy of deductive logic to indicate it provides the wrong norm. (Of course, when one does not consider that the normative theories may have been misapplied, one has jumped to a conclusion). There are several problems with the “is-ought” argumentation in O&C’s book. Given that cognitive psychology is not concerned with how people should reason, the “is-ought” analysis in BR signifies that O&C are not engaged in cognitive psychology, but are making an epistemological, philosophical argument. As an argument about normative rationality, the “is-ought” analysis is fallacious. “This fallacy consists in assuming that because something is now the practice, it ought to be the practice. Conversely, it consists in assuming that because something is not now the practice, it ought not to be the practice” (Damer 2005, p. 127). That is, one must be careful not to commit the “naturalistic fallacy” (to use G. E. Moore’s term). The “is-ought” analysis is also fallacious in its appeal to common opinion. “This fallacy consists in urging the acceptance of a position simply on the grounds that a large number of people accept it or in urging rejection of a position on the ground that very few people accept it” (Damer 2005). Virtually all people were once ignorant about the currently accepted fact that the earth is a round celestial object that revolves around the sun; and even today there is a large segment of the world population that remains unenlightened with regard to this. Does this mean that Copernicus was wrong, and was rightly excommunicated by the Church? One can have a personal opinion about how people should reason, and I think one actually should have BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

101

Commentary/Oaksford & Chater: Pre´cis of Bayesian Rationality such an opinion. However, when one is considering the psychology of human reasoning, it is irrelevant whether one can label 5% or 95% of the participants as “rational” because their behavior corresponds to some norm. Whether 5% or 95% of participants’ responses can be labeled “correct” is irrelevant to the task of describing and explaining what they are actually doing. The cognitive psychologist’s task is first and foremost that of developing descriptively adequate processing models that allow us to understand how people reason. All performance has to be accounted for, whether logically correct or incorrect. O&C give body to their “is-ought” ideology, by adopting Anderson’s (1990) rational analysis approach. However, it is a mistake to think that the “the essence of rational analysis [Anderson 1990] is to view cognitive processes as approximating some normatively justified standard of correct performance” (Oaksford & Chater 1998, p.18). As O&C stated more correctly elsewhere: “Anderson [1990] argues that we must distinguish normative from adaptive rationality” (Oaksford & Chater 1998, p. 174; see also, Chater et al. 2003). Indeed, “rationality is being used in two senses and that rationality in the adaptive sense, which is used here, is not rationality in the normative sense that is used in the studies of decision making and social judgement” (Anderson 1990, p. 31). There is no unconditional “should”: there are no absolute normative standards of correct performance. This goes against the very notion of adaptive rationality, which is relative and “defined with respect to a specific environment” (Anderson 1990, p. 35). O&C have fallen into the same conceptual trap many “mental logicians” fell into: a conceptual mistake – called “reasoning imperialism” by Rips (2002) – for which O&C, with much merit to their work, criticized such logicians. Bayesian Rationality provides prominent and valuable contributions to the literature that echoes Anderson’s (1990) idea that deductively irrational behavior might not be adaptively irrational. That is, the book successfully questions the normative status of logic within the rational analysis approach of Anderson (1990). However, instead of adopting and following the notion of adaptive rationality to its full extent, by questioning the feasibility of a single theory (whether probabilistic or logical) to have absolute normative status, they simply replace one normative theory with another. Adaptation is defined with respect to a specific environment and a true rational analysis accordingly takes account of this environment. It does not concern itself with absolute normative issues, because this denies the environmental relativity of adaptively rational behavior.

“Nonmonotonic” does not mean “probabilistic”

a probabilistic approach to cognition is the pervasiveness of nonmonotonicity in reasoning: almost any conclusion can be overturned, if additional information is acquired. They claim that nonmonotonic inferences fall outside the scope of logical methods, and that probability theory must be preferred. This judgment of the authors does not reflect the present state of play (see, e.g., Antoniou 1997, especially Part V). There are several good logical formalisms for nonmonotonic reasoning in existence, some of which even have a computational complexity that is vastly less than that of classical logic. More importantly, even in those cases where probabilistic modeling is in principle useful, one sometimes needs to supplement the model with a nonmonotonic logic to account for all the data. We illustrate this here by using an example from Chapter 5 of BR, not mentioned in the present Pre´cis, but relevant throughout: the suppression task (Byrne 1989). When subjects are presented with the modus ponens material “If she has an essay she studies late in the library. She has an essay,” they almost universally draw the conclusion: “She studies late in the library.” When instead they are presented with the same two premises plus the premise: “If the library is open she studies late in the library,” about half of them withdraw the inference. Logically speaking, we here have an instance of nonmonotonicity: addition of a premise leads to withdrawal of the original conclusion. O&C argue that their probabilistic model – which represents the conditional “if p, then q” as a conditional probability P(qjp) – can account for suppression effects. To explain the model and our criticism, we must first state the main Bayesian mechanism for updating probabilities, “Bayesian conditionalization”: (BaCo). The absolute subjective probability of event D given that evidence E has been observed is equal to the conditional probability P(DjE). Here E should encompass available evidence. In this context it is important to note that Bayesian conditionalization is a nonmonotonic principle: An extension of the evidence p may invalidate a previous posterior probability for q derived by (BaCo) applied to P(qjp). Informally, the O&C model works likes this: One is given the conditional probability P(qjp) with value, say, 1 – e for some positive but small e. If few exceptions are salient, e is small, and, given p, (BaCo) yields that q can be concluded with high probability (namely, 1 – e). The second conditional highlights a possible exception (the library being closed), which leads to an increase in e, and hence to a decrease in the a posteriori probability of q. But appealing though this picture is, it is not Bayesian (Stenning & van Lambalgen 2008b). Consider two putative Bayesian processes that can change the value of P(qjp) when new possible exceptions – say, not-r with probability 1 – P(r) – become salient. (1) The probability space is enlarged from f p, qg to fp, q, rg (where r stands for, say, “the library is open”), and this leads to a new representation of P(qjp). One may write

doi:10.1017/S0140525X0900048X Keith Stenninga and Michiel van Lambalgenb a Human Communication Research Center, University of Edinburgh, Edinburgh EH8 9LW, United Kingdom; bInstitute for Logic, Language and Computation, Department of Philosophy, University of Amsterdam, Amsterdam, 1012CP, The Netherlands. [email protected] www.inf.ed.ac.uk/people/staff/Keith_Stenning [email protected] http://staff.science.uva.nl/  michiell

Abstract: Oaksford & Chater (O&C) advocate Bayesian probability as a way to deal formally with the pervasive nonmonotonicity of common sense reasoning. We show that some forms of nonmonotonicity cannot be treated by Bayesian methods.

One argument that Oaksford & Chater (O&C) proffer in Bayesian Rationality (Oaksford & Chater 2007, henceforth BR) for

102

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

Figure 1. where the last equality follows under the plausible assumption that p and r are independent. O&C assume that the subject assigns a lower probability to P(qjp) in the enlarged representation P(qjp & r)P(r), suggesting that this is because the subject lowers P(r) from 1 to a value smaller than 1 when becoming aware of possible exceptions. In probability theory, P(qjp & r) P(r) would simply be a different representation of P(qjp), and the values of these expressions must be the same. There are no rationality principles governing changes in probabilities when enlarging the probability space, or rather there is one such

Commentary/Oaksford & Chater: Pre´cis of Bayesian Rationality principle, that P(qjp) remains the same when computed on an enlarged space. This is the only way in which one can guarantee that enlargements of the probability space in the limit lead to a coherent probability distribution – the starting point of Bayesian rationality. (2) An orthodox Bayesian alternative would be a construction in which the probability spaces remain the same (namely, the universal space based on all possible propositions), but the probability distributions change. In our toy world, the probability space is in both cases f p, q, rg, but one could assume that the probability distribution first assigns probability 0 to not-r, and, upon becoming aware of the second conditional “if r then q,” a nonzero probability. The trouble with such a suggestion is that from a Bayesian point of view, the transition from the a priori probability P(not-r)¼0 to the a posteriori P(not-r) . 0 is not allowed, because this cannot be achieved via (BaCo): conditionalizing on more evidence cannot make a null probability positive. One thus needs an additional rationality principle (beyond [BaCo]) governing such transitions. In the absence of such a principle, one has to assume that the probabilities of all nonsalient exceptions (such as not-r) are initially very small but nonzero. This increases the computational complexity of probabilistic reasoning enormously: One requires massive storage and intricate computations to maintain consistency of the probability assignment. These considerations show that in order to account for the data on the suppression task any probabilistic model needs to be supplemented with a theory about nonmonotonic and non-Bayesian, but still somehow rational, changes in degrees of belief. One may then question whether a probabilistic model is necessary at all; Stenning and van Lambalgen (2005; 2008a) provide a model cast entirely in terms of nonmonotonic logic.

The dynamics of development: Challenges for Bayesian rationality doi:10.1017/S0140525X09000491 Nils Straubinger, Edward T. Cokely, and Jeffrey R. Stevens Center for Adaptive Behavior and Cognition, Max Planck Institute for Human Development, Berlin, 14195, Germany. [email protected] [email protected] [email protected] http://www-abc.mpib-berlin.mpg.de/users/jstevens/

Abstract: Oaksford & Chater (O&C) focus on patterns of typical adult reasoning from a probabilistic perspective. We discuss implications of extending the probabilistic approach to lifespan development, considering the role of working memory, strategy use, and expertise. Explaining variations in human reasoning poses a challenge to Bayesian rational analysis, as it requires integrating knowledge about cognitive processes.

Bayesian rationality highlights the remarkable successes rather than failures of human reasoning by recasting seemingly erroneous reasoning in logical tasks using a probabilistic approach. However, in their book Bayesian Rationality (Oaksford & Chater 2007, henceforth BR), Oaksford & Chater (O&C) draw a rather static picture of human reasoning by focusing on typical patterns of responses from adults. We propose a more dynamic perspective, which considers that reasoning systematically varies within individuals over the lifespan (Howe & Rabinowitz 1996; Markovits & Barrouillet 2002) and between individuals with different levels of knowledge and expertise (Ericsson et al. 2006). Although O&C acknowledge the importance of considering reasoning data on individual differences (BR, p. 288) and on information processing capacities (p. 290),

they do not adequately account for how variation influences a Bayesian rational analysis of reasoning. Anderson (1991a) and others have pointed out that perhaps the major potential limitation, the “Achilles heel,” of rational analysis would be computational constraints that are too complex or arbitrary. We argue that our understanding of the mechanisms of change in reasoning can help us specify computational limitations for probabilistic modeling and assess whether a single model can capture the complexities of reasoning. Many important aspects of cognition change over the lifespan, and reasoning is no exception (Baltes et al. 1999). According to Piaget, both logical reasoning and probabilistic reasoning emerge from adolescence to young adulthood at the highest stage of cognitive development (Piaget & Inhelder 1975). Subsequent research, however, has qualified these findings, showing that younger children understand aspects of such reasoning (Falk & Wilkening 1998; Galotti et al. 1997). Furthermore, reasoning continues to develop during adulthood with performance in specific domains increasing as individuals gain reasoning knowledge and expertise (Ericsson & Lehmann 1996; Sternberg 1999). Yet, overall across the adult lifespan, abstract reasoning (measured by intelligence tests) declines with age (Verhaeghen & Salthouse 1997). Thus, reasoning is a dynamic aspect of cognition that varies with age and experience and results from the interplay of biological processes (e.g., brain maturation) and enculturation (e.g., education) (Baltes et al. 1999). A developmental perspective may inform Bayesian rational analysis by specifying computational limitations of the cognitive system. An important limitation faced by the human cognitive system is working memory capacity – a key determinant of reasoning performance (Kyllonen & Christal 1990). Like other cognitive capacities, working memory systematically changes across the lifespan by steadily increasing during childhood (Conlin et al. 2005) and declining across adulthood (Verhaeghen & Salthouse 1997). Working memory, therefore, poses a dynamic constraint on the rational analysis of reasoning. Although O&C are currently silent on the role of developmental changes in working memory and reasoning, they do note that individuals with higher working memory capacities tend to exhibit more logical reasoning. To illustrate, in the Wason selection task, a subgroup of individuals (ca. 10%) consistently chooses the logically correct combination of cards, indicating that although most seem to adopt a probabilistic model, others clearly do not. O&C suggest that this variation in behavior primarily reflects deliberative strategy use and educational (training) differences, which are “not indicative of individual differences in the nature of the fundamental principles of human reasoning” (BR, p. 288). This claim seems problematic given what we know about the interplay between strategy use, training, and basic cognitive mechanisms. Of course, cognitive capacities can constrain the strategies that people use; however, specific strategy use and training may shape the basic cognitive mechanisms, as well. Differences in memory strategies (e.g., rehearsal, chunking) can also alter basic mechanisms of working memory capacity and its relationship to cognitive performance (Cokely et al. 2006). In addition, both extensive practice with specific strategies and the acquisition of knowledge and expertise dramatically expand working memory (Ericsson & Kintsch 1995). Indeed, as training changes deliberative strategies to automatic processes, the cortex can undergo functional neuroanatomical reorganization (Dick et al. 2006). Thus, it is possible that deliberative strategy use and training may influence reasoning precisely because they alter underlying cognitive mechanisms such as working memory. Given the complex relationship between strategies, training, and cognitive mechanisms, it seems premature to dismiss individual differences in strategy use as not fundamental to reasoning. A comprehensive model of human reasoning must account for these differences. Variation in human reasoning has proven difficult to capture for probabilistic models (Shultz 2007), although recent research BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

103

Commentary/Oaksford & Chater: Pre´cis of Bayesian Rationality has made some progress applying probabilistic models to individual differences (e.g., category learning; Navarro et al. 2006) and cognitive development (e.g., causal reasoning; Sobel et al. 2004). This work represents a step in the right direction; however, we expect that no single model can predict reasoning performance equally well across age groups and levels of experience. Indeed, systematic variations in peoples’ behavior suggest that several different models (or modifications of a given model) may be required to explain developing behavior (Shultz 2007). Nevertheless, investigating differences between the models across age groups and skill levels may help us to understand more exactly “what differs” between and “what develops” within individuals. In closing, we must emphasize O&C’s comment that probabilistic models are often only functional level theories that should not be confused with algorithmic level theories (process models). Brighton and Gigerenzer (2008) have pointed out in their discussion of the limits of Bayesian models of cognition that the question of why the human mind does what it does (functional level) cannot be separated from the question of how the human mind does it (algorithmic level). Therefore, it is crucial that future Bayesian rational analyses specify how exactly their functional level models constrain theorizing about cognitive processes. This issue is especially relevant as the data connecting development, expertise, working memory, and reasoning imply that multiple strategies (and therefore processes) are at play. Though Bayesian rationality seems to provide a functional level account of prototypical adult reasoning, the development of cognitive capacities and expertise remains underappreciated. ACKNOWLEDGMENTS Preparation of this commentary was supported in part by a stipend from the International Max Planck Research School LIFE to NS. We are grateful to Henrik Olsson for comments on a previous draft.

How do individuals reason in the Wason card selection task? doi:10.1017/S0140525X09000508 Eric-Jan Wagenmakers Department of Psychological Methods, University of Amsterdam, Amsterdam 1018 WB, The Netherlands. [email protected] http://users.fmg.uva.nl/ewagenmakers/

Abstract: The probabilistic approach to human reasoning is exemplified by the information gain model for the Wason card selection task. Although the model is elegant and original, several key aspects of the model warrant further discussion, particularly those concerning the scope of the task and the choice process of individuals.

In the book Bayesian Rationality (Oaksford & Chater 2007, henceforth BR), Oaksford & Chater (O&C) present a summary and a synthesis of their work on human reasoning. The authors argue that formal logic and deduction do not explain how people reason in everyday situations. The deficiencies of the most simple forms of logic are obvious when one considers that they may assign “true” to absurd statements such as “if the moon is blue, than cows eat fish” (BR, p. 70). More importantly, the authors propose that, in contrast to formal logic, probability calculus does provide the right tools for an analysis of human reasoning. Thus, the authors argue that people solve deductive tasks by inductive methods. From this perspective, human reasoning can be characterized as Bayesian or rational. Consider the Wason card selection task discussed in Chapter 6. Participants are confronted with four cards, showing an A, a K, a 2, and a 7. Participants are told that each card has a number

104

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

on one side and a letter on the other. They are given a rule, “if there is an A on one side, then there is a 2 on the other side,” and subsequently, have to select those cards that need to be turned over to assess whether the rule holds true or not. A moment’s thought reveals that the cards that need to be turned over are the A card and the 7 card. Yet, the majority of participants do not choose the 7 card, but tend to choose the 2 card instead. O&C propose an elegant Bayesian model – the information gain model – to account for people’s performance in the Wason task. According to the model, people select the cards that reduce their expected uncertainty the most. Specific assumptions about the rarity of the information on the cards lead to the conclusion that selection of the 2 card might be rational after all. The information gain model has been subjected to intense scrutiny (e.g., Oberauer et al. 1999). For non-experts, the details of this discussion are somewhat difficult to follow. A useful guideline is that a model should only be abandoned when it can be replaced with something better. And – criticisms raised against the information gain model notwithstanding – I have not come across a model that does a better job explaining how people make their card selections. Despite its simplicity and elegance, some important details of the information gain model were not clear to me. First, O&C argue, on page 210, that their account only holds if participants regard the cards as a sample from a larger population. Perhaps the authors could spell out this argument in a bit more detail. Taking probability as a reflection of degree of belief, I did not immediately see what calculations are in need of adjustment. Second, the authors mention that participants who realize that the cards are not sampled from a larger population would always choose the A card and the 7 card. I do not know whether this prediction has been tested empirically, but I find it only slightly more plausible than cows eating fish. Note that in the Wason task a substantial proportion of participants do not even select the A card. Another issue that warrants closer examination is the way the model’s predictions relate to the data. In the information gain model, each card reduces the expected uncertainty to some extent. Why then does an individual participant not select all four cards, but generally only selects one or two? In other words, it was unclear to me how the model, from a consideration of expected uncertainty reduction, can predict card selections for an individual participant. A fourth point concerns the role of individual differences. As the authors discuss on page 211, a subgroup of undergraduate students with high intelligence (about 10%) do select the A card and the 7 card. This result strengthened my initial belief that a motivated, intelligent person would always choose the A and 7 cards, when given sufficient time. In the spirit of falsification, I then tested this assumption on a colleague, who of course immediately selected the A and 2 cards. Perhaps she was not sufficiently motivated to think the problem through carefully – would incentives of time or money increase the selection of the 7 card? O&C are to be admired for their principled approach to quantitative modeling, and for their courage to take on the unassailable dogma of human irrationality. It is unfortunate that much of the material in the book was already available elsewhere (e.g., Oaksford & Chater 2001; 2003b); therefore, it was not entirely clear to me what the book adds to our current knowledge base. One final comment. It strikes me as paradoxical that researchers who argue for a coherent, rational approach to human reasoning then proceed to apply an incoherent, irrational approach to the statistical analysis of their experimental data. Throughout the book, the authors renounce Popper’s stance on the importance of falsification, arguing that this is not how science works, nor how people reason. But then, in the very same work, the authors measure the validity of their models by means of pvalues, and include statements such as “the model could not be rejected.” Why?

Response/Oaksford & Chater: Pre´cis of Bayesian Rationality

Authors’ Response The uncertain reasoner: Bayes, logic, and rationality doi:10.1017/S0140525X0900051X Mike Oaksforda and Nick Chaterb a School of Psychology, Birkbeck College London, London, WC1E 7HX, United Kingdom; bDivision of Psychology and Language Sciences & ESRC Centre for Economic Learning and Social Evolution, University College London, London, WC1E 6BT, United Kingdom. [email protected] [email protected] www.bbk.ac.uk/psyc/staff/academic/moaksford www.psychol.ucl.ac.uk/people/profiles/chater_nick

Abstract: Human cognition requires coping with a complex and uncertain world. This suggests that dealing with uncertainty may be the central challenge for human reasoning. In Bayesian Rationality we argue that probability theory, the calculus of uncertainty, is the right framework in which to understand everyday reasoning. We also argue that probability theory explains behavior, even on experimental tasks that have been designed to probe people’s logical reasoning abilities. Most commentators agree on the centrality of uncertainty; some suggest that there is a residual role for logic in understanding reasoning; and others put forward alternative formalisms for uncertain reasoning, or raise specific technical, methodological, or empirical challenges. In responding to these points, we aim to clarify the scope and limits of probability and logic in cognitive science; explore the meaning of the “rational” explanation of cognition; and re-evaluate the empirical case for Bayesian rationality.

R1. Introduction Bayesian Rationality (Oaksford & Chater 2007, henceforth BR) proposed that human reasoning should be understood in probabilistic, not logical, terms. In Part I, we discussed arguments from the philosophy of science, artificial intelligence, and cognitive psychology, which indicate that the vast majority of cognitive problems (outside mathematics) involve uncertain, rather than deductively certain, reasoning. Moreover, we argued that probability theory (the calculus for uncertain reasoning) is a more plausible framework than logic (the calculus for certain reasoning) for modeling both cognition in general, and commonsense reasoning in particular. In Part II, we considered a strong test of this approach, asking whether the probabilistic framework can capture human reasoning performance even on paradigmatically “logical” tasks, such as syllogistic reasoning or conditional inference. The structure of this response is as follows. In section R2, we reflect on the ubiquity of uncertainty and address the theoretical attempts to preserve logic as a separate and core reasoning process. In section R3, we compare and evaluate Bayesian and logic-based approaches to human reasoning about uncertainty. Section R4 focuses on the methodology of rational analysis (Anderson 1990; 1991a; Oaksford & Chater 1998b) and its relationship to more traditional algorithmic and neuroscientific approaches. Section R5 discusses a variety of specific

issues in the empirical data from the psychology of reasoning, and the modeling of that data. Finally, section R6 concludes the case for a “Bayesian turn” in the brain and cognitive sciences in general, and for the understanding of human reasoning in particular.

R2. The ubiquity of uncertainty: Distinctions that might preserve logic Many commentators suggest ways to preserve a role for logic as a separate and core component in an account of human reasoning, despite the challenge provided by uncertainty (Allott & Uchida, Evans, Politzer & Bonnefon). We argue that logic does have an important role in modeling cognition; but we argue against the existence of cognitive processes dedicated to logical reasoning. R2.1. Rationality 1 versus Rationality 2

Evans suggests that a distinction should be drawn between two types of rationality (Evans & Over 1996a). Rationality 1 relates to implicit, possibly associative, processes, operating over world knowledge, which Evans also terms “ecological rationality.” This type of rationality arises from System 1 in Evans and Over’s (2004) Dual Process Theory (see also Evans & Frankish, in press; Sloman 1996; Stanovich & West 2000). Rationality 2 involves explicitly following normative rules, and is the type of rationality achieved by Evans and Over’s (2004) System 2. System 2 processes are logical, rule-governed, and conscious. Moreover, Evans has argued for a crucial asymmetry between the systems. It requires cognitive effort to ignore System 1, and to use System 2 for logical inference: that is, to infer only what follows from the structure of the given premises. The fundamental problem with this Dual Process view is that these two systems must interact – and if the systems obey fundamentally different principles, it is not clear how this is possible. Consider the familiar example of inferring that Tweety flies from the general claim that birds fly and the fact that Tweety is a bird. On the Dual Process view, this inference could be drawn logically from the premises given by System 2, from the assumption that birds fly is a true universal generalization; System 1, by contrast, might tentatively draw this conclusion by defeasible, associative processes, drawing on general knowledge. But a lack of synchrony between the two systems, presumed to operate by different rational standards, threatens to cause inferential chaos. Consider, for example, what happens if we consider the possibility that Tweety is an ostrich. If System 2 works according to logical principles, the clash of two rules threatens contradiction: we know that birds fly, but that ostriches do not. To escape contradiction, one of the premises must be rejected: most naturally, birds fly will be rejected as false. But we now have two unpalatable possibilities. On the one hand, suppose that this retraction is not transferred to general knowledge and hence is not assimilated by System 1. Then the two systems will have contradictory beliefs (moreover, if System 2 reasoning cannot modify general knowledge, its purpose seems unclear). On the BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

105

Response/Oaksford & Chater: Pre´cis of Bayesian Rationality other hand, if birds fly is retracted from world knowledge, along with other defeasible generalizations, then almost all of general knowledge will be stripped away (as BR notes, generalizations outside mathematics are typically defeasible), leading System 1 into inferential paralysis. The centrality of logic for a putative System 2 is also brought into doubt by considering that one of its main functions is to consciously propose and evaluate arguments. Yet, argumentation, that is, the attempt to persuade oneself or others of a controversial proposition, is uniformly agreed not to be a matter of formal logic (Walton 1989), although aspects of argumentation may naturally be modeled using probability theory (Hahn & Oaksford 2007). Thus, perhaps the core human activity for which a logic-based System 2 is invoked may, ironically, be better explained in probabilistic terms. People can, of course, be trained to ignore some aspects of linguistic input, and concentrate on others – for example, in the extreme, they can learn to translate natural language statements into predicate logic (ignoring further aspects of their content) and employ logical methods to determine what follows. But, for the psychology of reasoning, this observation is no more significant than the fact that people can learn the rules of chess and ignore most of the visual features of the pieces, the board, or indeed, their surroundings. Conscious application of logical principles is a learned skill built on top of non-logical machinery (and, indeed, is highly effortful, even for logicians); it does not involve, we suggest, tapping in to some underlying logical “engine.” It is this conscious application of logical concepts (and related notions from mathematics, philosophy, and computer science) that underpins, we suggest, the small but significant correlation between “logical” performance on some reasoning tasks (e.g., selecting the p and not-q cards, in Wason’s selection task) and IQ (Stanovich & West 2000). Logical reasoning is a late and cognitively challenging cultural innovation, rather than a core component of our mental machinery. Evans also expresses disappointment that we do not address individual differences (Stanovich 2008), which have been viewed as supporting a Dual Process account. But from the present perspective, individual differences concerning the application of learned logical rules are no different from individual differences in chess playing – that is, neither are directly relevant to the question of whether there are single or multiple reasoning systems. Indeed, we suggest that individual differences provide no stronger evidence that cognition involves core logical competence, than that cognition involves core chess-playing competence. It may turn out, indeed, that there is no real incompatibility between Stanovich’s account and ours. In particular, the distinction Stanovich draws between control processes and other autonomous systems is a distinction common to all theories of cognition (see Oaksford & Chater, in press). But as Kowalski’s (1979) classic equation, “Algorithm ¼ Logic þ Control,” reminds us, logic and control processes are very different (see, e.g., Anderson 1983). Hence, Stanovich may not really be committed to anything like Evans’ logically competent System 2. (A further complication is that a distinction between processes of logic 106

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

and control is now reflected in Evans [2007], who moots the possibility of a tri-process theory.)

R2.2. The split between semantics and pragmatics

Grice’s (1975) theory of conversational implicature originally attempted to split off a “stripped down” logicbased natural language semantics, from the complex, knowledge-rich processes of pragmatic interpretation involved in inferring a speaker’s intentions. In this way, he aimed to retain a logical core to semantics, despite apparently striking and ubiquitous clashes between the dictates of formal logic and people’s intuitions about meaning and inference. Within this type of framework, Allott & Uchida attempt to preserve the truth of potentially defeasible conditionals (if it’s a bird, then it flies, or, as above, birds fly) despite the ready availability of counterexamples. They suggest that this conditional is true in one model, but not in the model that is considered when an additional premise giving a counterexample is added (e.g., when we consider the possibility that Tweety is an ostrich). But in classical logic, only an inference that holds in all models is deductively valid, by definition. Thus, accepting that this inference holds only in some models implies accepting that the inference is uncertain (contra, e.g., O’Brien). Indeed, in BR, we argue uncertainty is ubiquitous in human reasoning; outside mathematics, deductive reasoning, which guarantees the truth of a conclusion given the premises, is, to a first approximation, never observed. Moreover, understanding reasoning involves working out pragmatic details about what default background assumptions are applicable in reasoning. Thus, for example, our accounts of specific reasoning phenomena, across conditional reasoning, the selection task, and syllogistic reasoning, involve default assumptions about the environment, for example, what is rare and what is common (cf. McKenzie; McKenzie et al. 2001) and when states are likely to be independent or conditionally independent. In this light, we agree with Stenning & van Lambalgen’s claim that “pure” Bayesian analysis, working from the premises alone, cannot capture suppression effects in conditional reasoning (see sect. R3.6) – we view this as illustrating the knowledge-rich character of reasoning, rather than challenging a Bayesian account. The ubiquity of uncertain, knowledge-rich inference, argues for an alternative to invoking the semantics/pragmatics distinction to maintain a logical semantics for natural language: namely, that natural language semantics may be probabilistic “all the way down.” Experiments in the psychology of reasoning, as reviewed in BR, find little support for the existence of a level of logic-based representation or inference. BR proposes a starting point for a probabilistic semantics: If p then q conditionals are assumed to express that the conditional probability P(qjp) is high (following Adams 1975; 1998; Bennett 2003; and Edgington 1995, among others); the quantifiers Some, Few, Most, All are similarly assumed to express constraints on probabilities (e.g., Some A are B is rendered as P(A, B) . 0; Most A are B claims that P(BjA) is high). Switching from a logical to a probabilistic semantics provides, we argue, a better fit with patterns of human reasoning. Of course, it remains possible that a logical core

Response/Oaksford & Chater: Pre´cis of Bayesian Rationality interpretation might be maintained – but it seems theoretically unparsimonious to do so (Edgington 1995). A shift from a logical to a probabilistic semantics for aspects of natural language may also allow a more integrated account of semantics and pragmatics. Indeed, McKenzie (e.g., Sher & McKenzie 2006) has powerfully demonstrated the importance of pragmatic factors, even within a purely probabilistic framework (but see, Hilton et al. 2005). Nonetheless, the core insight of Grice’s program remains: that splitting apart semantic factors (concerning meaning) and pragmatic factors (concerning inferences about speaker intentions) is a prerequisite for constructing a tractable semantic theory, whether that theory be based on logic (as Allott & Uchida argue) or probability (as BR proposes). R2.3. Proof and uncertainty and structure and strength

Politzer & Bonnefon argue that a key element missing from a purely probabilistic account is how premises can be used to construct proofs to derive conclusions. Thus, they argue that the probabilistic account allows the evaluation of the strength of the relationship between premises and conclusion, but not how the conclusion is generated in the first place. Note, though, that both logic and probability are theories of the nature of inferential relationships between propositions (Harman 1986). Neither specify how reasoning should be carried out, let alone how interesting conclusions should be generated. Moreover, for both logic and probability, a range of algorithms have been developed which can both evaluate given conclusions, and generate new conclusions (e.g., logic programming and Bayesian networks). From both perspectives, any set of information potentially generates an infinite set of possible conclusions; so that an immediate question is: What counts as an interesting conclusion? A natural suggestion from the probabilistic point of view is that conclusions with a low prior probability are, other things being equal, more surprising and hence more interesting (as employed in the account of syllogistic reasoning described in BR), although interesting logic-based measures of semantic information content have also been proposed (Johnson-Laird 1983). More generally, the probabilistic approach is just as able as logic-based approaches to serve as the basis for algorithmic models of thought. For example, Oaksford & Chater (in press) use a constraint satisfaction neural network implementation of the probabilistic approach. The links in the network captures the conditional and default assumptions about structural relations between variables (in the causal context, involving alternative causes and defeaters); and the strength of each link is captured by a weight. A similar distinction between structure and strength has been invoked in causal reasoning using Bayesian networks (Griffiths & Tenenbaum 2005) and applied in Hahn and Oaksford’s (2007) probabilistic account of argumentation. R3. Logic, probability, and the challenge of uncertain reasoning? In this section, we consider whether, as some commentators suggest, we have mischaracterized the scope of logic

or chosen the wrong alternative calculus in order to reason about uncertainty. We deal with logic and probability in turn. R3.1. How are logic and probability related?

Pfeifer & Kleiter observe that probability theory already includes classical propositional logic as a special case. Thus, one way of understanding the approach outlined in BR is as enriching conventional logic to give an inductive logic – a system of logic that extends deduction to less-than-certain inferences (Hawthorn 2008). To a good approximation, modern inductive logic just is Bayesian probability (Chater et al., in press; Earman 1992), with some additional discussion of the measure of the confirmation relation (see later discussion of Poletiek and Nelson). Since Carnap (1950), this Bayesian inductive logic includes classical logic – if a statement has a probability of 1, then any logical consequence of that statement also has a probability of 1. Similarly, if a statement has an implication with a probability of 0, then that statement has a probability of 0 (note, however, that probability theory does not readily represent the internal structure of atomic propositions, and has no general theory of, for example, quantification or modality). The Bayesian inductive perspective is required not because classic logic is incorrect, but because, outside mathematics, it rarely, if ever, applies (Oaksford & Hahn 2007) – inferential relations between propositions are relentlessly uncertain (Jeffrey 1967). R3.2. Is relevance relevant?

O’Brien proposes a different enrichment of logic, drawing on his important work with Braine on mental logics (Braine & O’Brien 1991), which aims to capture a notion of relevance between antecedent and consequent (i.e., so that conditionals such as if 2 is odd, then the sky is purple are no longer automatically true, just in virtue of the false antecedent). Thus, Braine and O’Brien’s work aims to go beyond the material conditional, which BR ascribed to mental logic as a whole (e.g., Rips 1994). Adding a condition of relevance, while potentially important, does not help deal with the problem of uncertain reasoning, however. Indeed, O’Brien’s account of conditionals is, instead, a strictly deductive version of the Ramsey test (like, e.g., Ga¨rdenfors 1986) – conditionals are only asserted if the consequent, q, follows with certainty from the antecedent p (and background knowledge). Thus, Braine and O’Brien’s (1991) logical interpretation of the conditional suffers the same fundamental problem as material implication: an inability to capture the fact that generalizations outside mathematics are inevitably uncertain.1 Moreover, despite Braine and O’Brien’s intentions, their system does not seem to enforce relevance between antecedent and consequent, either. The introduction rule for if p then q, used by O’Brien, and described in Braine and O’Brien (1991), states that if p then q can be inferred if q follows from the supposition of p together with background knowledge, B. If we know p is false (i.e., background knowledge B implies not-p), then supposing p and B implies p & not-p, which is a contradiction, from which any conclusion follows – including q. So conditionals BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

107

Response/Oaksford & Chater: Pre´cis of Bayesian Rationality such as if 2 is odd, then the sky is purple can be asserted, after all. Similarly, any conditional whose conclusion is known to be true (i.e., B implies q) will automatically meet the condition that p & B implies q (because this is a monotonic logic – adding premises can never remove conclusions). Hence, conditionals such as if the sky is purple, then 2 is even, will also be asserted – again violating intuitions of relevance. R3.3. Uncertain reasoning via nonmonotic logic?

Stenning & van Lambalgen argue that we misrepresent the scope of current logical methods, noting that a range of nonmonotonic logics, in which adding a premise may require withdrawing a previously held conclusion, might meet the challenge of uncertainty. As noted in BR, and elsewhere (e.g., Oaksford & Chater 1991; 2002), there are, however, fundamental problems for nonmonotonic logics in the crucial case where different “lines of argument” clash. Thus, if it is sunny, John goes to the park, and it’s sunny appears to provide a powerful argument that John goes to the park. But adding the premise, John is arrested by the police in a dawn raid, together with background knowledge, appears to yield the conclusion that John does not go to the park. From the perspective of classical logic, this situation is one of contradiction – and what is needed is a way of resolving which premise should be rejected. For example, one might claim that the conditional if it’s sunny, John goes to the park is false, precisely because of the possibility of, among other things, arrest. But, as noted in section R2.1, it is then difficult to avoid the conclusion that all conditionals, outside mathematics, are false, because the possibility of counterexamples always exists. Reasoning from premises known to be false is not, of course, justified, whether in logic, or any other standard framework, and hence, the logical analysis of the original argument collapses. The strategy of nonmonotonic logic attempts to solve this problem by treating the conditional as a default rule, which holds, other things being equal. Indeed, outside mathematics, almost all rules are default rules. Indeed, the implicit rule that allows us to infer that being arrested is incompatible with a trip to the park is itself a default rule, of course – for example, arrest may be extremely brief, or perhaps the police station is itself in the park. Thus, from this viewpoint, uncertain reasoning centrally involves resolving clashes between default rules. In BR, we argue that resolving such clashes is not typically possible by looking only at the structural features of arguments. Instead, it is crucial to differentiate stronger and weaker arguments, and degrees of confidence in the premises of those arguments. Logical methods provide no natural methods for expressing such matters of degree; but dealing with degrees of belief and strength of evidence is the primary business of probability theory. R3.4. Is logic relevant to cognition?

Several commentators suggest that the powerful machinery of logic should not be jettisoned prematurely (Allott & Uchida, De Neys, O’Brien, Politzer & Bonnefon, Stenning & van Lambalgen). As we noted in section R3.1, probability theory (i.e., modern inductive logic) is 108

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

a generalization of logic, allowing degrees of uncertainty. However, it is a generalization that is presently limited in scope. This is because how probability interacts with richer representations involving, for example, relations, quantification, possibility, deontic claims, tense and aspect, and so on, is yet to be worked out. BR has, as we have mentioned, some preliminary suggestions about the probabilistic representation of individual connectives (if. . .then. . .) and quantifiers (Most, Few, Some, etc.). But this is very far from a full probabilistic generalization of, for example, the predicate calculus, the workhorse of classical logic and natural language semantics. The formal challenges here are substantial. Nonetheless, much progress has been made, in a number of directions, in fusing together probabilistic and logical methods (e.g., see papers in Williamson & Gabbay 2003), thus advancing Carnap’s (1950) program of building an inductive logic. Pfeifer & Kleiter apply logic in an interesting, but distinct, way: as providing a machinery for reasoning about probability, rather than using probability to generalize logic. According to De Neys, concentrating on the computational level means that BR underplays the role of logic in human reasoning. De Neys argues that latency and brain imaging studies, investigating the mental processing involved in reasoning, rather than just the output of these processes, consistently reveal a role for logic. Yet all the cases that De Neys cites involve a conflict between belief and logic such that prior belief suggests one response, but logical reasoning from the given premises suggests another. However, the Bayesian approach can explain at the computational level why such conflicts might arise and therefore why inhibitory processes might need to be invoked (De Neys et al. 2008; Houde´ et al. 2000). Oaksford and Hahn (2007) point out that probabilistic validity of an argument and its inductive strength can conflict. So, for example, Modus Ponens (MP) is probabilistically valid. However, if the probability of the conditional is low, then the inductive strength of the argument, that is, the probability of the conclusion given the premises, will also be low. The right computational level analysis may, therefore, remove the need to propose two special purpose cognitive systems operating according to different principles. This view seems consistent with the current state of imaging studies, which provide little evidence for a dedicated logical reasoning module (Goel 2007). O’Brien describes Chrysippus’ dog’s ability to follow a scent down one path in a fork in the road, having eliminated the other as an application of the logical law of disjunction elimination – and hence, suggests that logic is cognitively ubiquitous. However, this logical law cannot uncritically be imported into a theory of canine cognition. For one thing, such patterns of behavior are at least as well modeled in probabilistic (Toussaint et al. 2006), as in logical, terms. Indeed, probabilistic methods are crucial in planning tasks in uncertain environments, which is, of course, the normal case, outside mathematically specified game-playing environments. In any case, just because a behavior can be described in logical or probabilistic terms does not directly imply that it is governed by logical or probabilistic processes. The issues here are complex (see the excellent introductory chapter to Hurley & Nudds 2006) and many possibilities would

Response/Oaksford & Chater: Pre´cis of Bayesian Rationality need to be ruled out before abandoning Lloyd Morgan’s canon: that lower-level explanations of animal behavior should be preferred. In short, we believe that cognitive science ignores logic at its peril – logic provides powerful and much needed tools, just as do other branches of mathematics. It does not, however, readily capture patterns of human reasoning, or, we suggest, cognition at large, unless generalized into a probabilistic form able directly to deal with uncertainty.

R3.5. Why probability rather than other numerical measures?

Danks & Eberhardt and Politzer & Bonnefon ask why we use probability, rather than other numerical measures of degrees of belief, such as confidence intervals, Dempster-Shafer belief functions (Dempster 1968; Shafer 1976), or fuzzy logic (Zadeh 1975). In BR, our primary motivation is practical: Bayesian probabilistic methods provide a natural way to capture human reasoning data; and more generally, Bayesian methods have swept through the brain and cognitive sciences, from understanding neural coding (Doya et al. 2007), through vision, motor control, learning, language processing, and categorization. Even within research on reasoning, Bayesian methods have proved central to understanding inductive inference (Griffiths & Tenenbaum 2005; Tenenbaum et al. 2007), causal reasoning (Sloman 2005; Tenenbaum & Griffiths 2001), and argumentation (e.g., Hahn & Oaksford 2007), as well as the primarily deductive reasoning problems considered in BR.2 Moreover, probabilistic methods connect with rich literatures concerning computational inference methods (e.g., based on graphical models, Lauritzen & Spiegelhalter 1988; Pearl 1988), machine learning (e.g., Jacobs et al. 1991), and normative theories of reasoning about causality (Pearl 2000). Finally, probability also has deep relationships to other powerful concepts in the brain and cognitive sciences, including information theory (e.g., Blakemore et al. 1991) and simplicity, for example, as captured by Kolmogorov complexity theory (e.g., Chater 1996; Chater & Vita´nyi 2002). Thus, our focus on probability is primarily pragmatic rather than, for example, depending on a priori justifications. Danks & Eberhardt focus, nonetheless, on justification, arguing that doubt can be cast on justifications such as the Dutch Book argument and long run convergence theorems. We see the project of rational analysis as a user of probability, on a par with the rest of science, for example, statistical mechanics, Bayesian image restoration, or economics. We only need to be as concerned about justification as these other endeavors. Danks & Eberhardt’s worries are analogous to Berkeley’s objections to Newton’s infinitesimals: of considerable conceptual importance, but with little direct impact on the practical conduct of science. Nonetheless, probability is at least better justified than alternative formalisms for modeling uncertainty. Politzer & Bonnefon and Danks & Eberhardt raise the possibility that the assumptions of the probabilistic approach may be too strong. We instead believe that they are, if anything, too weak; that is, they define

minimal coherence conditions on beliefs, which need to be supplemented with richer formalisms, including, as noted in section R3.4, the ability to represent relations and quantification, and to represent and manipulate causal relations (e.g., Pearl 2000). R3.6. Are we Bayesian enough?

Other commentators (Over & Hajichristidis, Pfeifer & Kleiter, Stenning & van Lambalgen) have the opposite concern: that BR is not Bayesian enough. Over & Hadjichristidis argue that in conditional inference, not only is the conditional premise (e.g., if p then q) uncertain, but so is the categorical premise, p. In BR (p. 121), we mention this general case (implying Jeffrey’s rule [Jeffrey 1983]), but point out that this extra element of uncertainty appears unnecessary to capture the conditional reasoning data. Stenning & van Lambalgen and Pfeifer & Kleiter also argue, in different ways, that we are insufficiently Bayesian. Stenning & van Lambalgen argue that our account of suppression effects is not Bayesian because coherent Bayesian revision of the probability space assumes “rigidity”: that is, the conditional probability P(qjp) remains the same if we learn the truth of a categorical premise: p, q, not-p, or not-q (and no other information). We agree. But this does not imply that P(qjp) remains the same if we are told about that p, because pragmatic factors allow us to infer a great deal of additional information; and this information can legitimately change P(qjp). It is this latter case that is relevant for reasoning with verbal materials. Thus, suppose I believe if the key is turned, the car starts; and I am told: “the car didn’t start this morning.” This would be a pragmatically pointless remark, if the key had not been turned. I therefore infer that the key was turned, and the car didn’t start for some other reason. Thus, I revise down the probability of the relevant conditional P(car startsjkey turned) dramatically. So the violation of rigidity, notably in this type of Modus Tollens (MT) inference, does not violate Bayesian precepts, but merely applies them to the pragmatics of utterances (see BR, pp. 126– 128; Sobel 2004; Sober 2002). Pfeifer & Kleiter suggest that inference can proceed locally and deductively in a mental probability logic. In such a logic, the precise probability of a conclusion cannot typically be deduced from the probabilities of the premises – but a probability interval can be. We adopted a similar approach to probabilistic validity for syllogisms where, according to our probabilistic semantics, quantifiers describe probability intervals. Nonetheless, in line with Stanovich and West’s (2000) “fundamental computational bias,” we believe that people spontaneously contextualize and elaborate verbal input, by adding information from world knowledge. Indeed, it takes substantial cognitive effort not to do this. Consequently, we think it unlikely that people reason deductively about probability intervals. R3.7. Measuring confirmation

People are not merely passive observers. They can actively search for information to help test hypotheses, or to BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

109

Response/Oaksford & Chater: Pre´cis of Bayesian Rationality achieve specific goals. In BR, we outline “rational” accounts for both cases. Where people test between hypotheses, a natural objective is to search for data in order to maximize the expected amount of information that will be gained in the task (Shannon & Weaver 1949). This is “disinterested” inquiry. Where people gain information to help achieve specific goals, then a natural objective is to choose information to maximize expected utility (balancing costs of information search with the improved choices that may result from new information). This is “goal-directed” inquiry. In BR, we note that different variations of Wason’s selection task are appropriately captured by versions of one or other model. In particular, we showed how attending to the goal-directed case avoids the postulation of specific machinery, such as “cheaterdetection” modules (e.g., Cosmides 1989), to explain patterns of experimental data (e.g., BR, pp. 191 –98). Focusing on disinterested inquiry, Nelson notes that a wide range of normative and descriptive proposals for assessing the strength of information in a piece of data have been proposed. In testing these models against a wide range of psychological data (Nelson 2005), he finds that the information-theoretic measure implicit in our analysis stands up well against competitors, although it is not picked out uniquely by the empirical data. Poletiek notes a further interesting link to philosophy of science, noting that Popper’s measure of severity of test is equivalent to P(ejH)/P(e), for data e and hypothesis H. And the logarithm of this quantity just is the amount of information carried by the evidence e about H – the quantity which we use in our model of disinterested inquiry in the selection task. This quantity is also used as a measure of the degree to which a theory is confirmed by the data in confirmation theory (Milne 1996). This is, as Poletiek notes, particularly interesting, given that Popper’s measure of severity of test is part of a theoretical framework which aims to entirely avoid the notion of confirmation (see also Milne 1995). Thus, our account of the selection task could be recast, from a Popperian standpoint, as a rational analysis in which people attempt to choose data to provide the more severe possible tests for their hypotheses.

R4. Rational analysis, algorithmic processes, and neural implementation BR is primarily concerned with the rational analysis of human reasoning (e.g., Anderson 1990; 1991a; Chater & Oaksford 2008a; Oaksford & Chater 1998b). In this section, we consider the role of rational analysis in the brain and cognitive science and whether this style of explanation is fundamentally flawed.

R4.1. The power of rational analysis

Hahn notes that the shift away from considerations of algorithms and representations, encouraged by rational analysis, has led to a substantial increase in explanatory power in cognitive science, in a number of domains. Where the underlying explanation for an aspect of cognition arises from the rational structure of the problem being solved, there focusing on specific algorithmic and 110

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

neural mechanisms may be unhelpful. Therefore, building specific algorithmic models (e.g., connectionist networks) of a phenomenon may replicate the phenomenon of interest (by virtue of being an adaptive solution to the “rational” problem in hand), but may throw little light on why it occurs. R4.2. Normativity and rational analysis

Evans and Schroyens are concerned about the normative aspect of rational analysis. Evans questions whether normativity is a proper part of a computational-level analysis of human reasoning, and by implication, cognition in general, and recommends a switch to an ecological notion of rationality. He suggests rationality should concern how well people are adapted to their environment, which may not require following the prescriptions of any normative theory of reasoning (cf. Gigerenzer & Goldstein 1996). We suggest, however, that ecological rationality does not replace, but rather, complements normative rationality. Normative considerations are still required to explain why a particular algorithm works, given a particular environment; indeed, this is precisely the objective of rational analysis. Thus, for example, in arguing for the ecological rationality of various fast and frugal heuristics (Gigerenzer et al. 1999), Gigerenzer and colleagues appeal to a Bayesian analyses to explore the type of environmental structure for which their algorithms succeed (e.g., Martignon & Blackmond-Laskey 1999). Thus, rational analysis cannot be replaced by, but seeks to explain, ecological rationality. Note, too, that rational analysis is goal-relative: it specifies how best to achieve a given goal, in a given environment, with given constraints (Anderson 1990; Oaksford & Chater 1998b). So, if your goal is to land a rocket on the moon, your guidance system ought to respect classical physics; if your goal is to avoid contradictions, you ought to reason according to standard logic; and if your goal is to avoid accepting bets that you are bound to lose, you ought to follow the rules of probability theory (see Ch. 2 of BR). Ignoring the goal-relativity of rational analysis leads Schroyens to suggest that we have fallen into Moore’s (1903) “naturalistic fallacy” in ethics: that we have attempted to derive an “ought” from an “is.” Moore’s concern is that no facts about human behavior, or the world, can justify an ethical theory. Ethics is concerned with non-relative notions of “ought”: the aim is to establish universal principles of right behavior. But the goal-relativity of rational analysis makes it very different from the domain of ethics, because it is conditional. Rational analysis considers: if you have objective O, given an environment E, and constraints C, then the optimal action is A. Ethics, by contrast, considers whether O is a justifiable objective. And the nature of the solution to a well-specified optimization problem is itself firmly in the domain of facts. Indeed, were Schroyens’ concern valid, then its consequences would be alarming, sweeping away functional explanation in biology and rational choice explanation in economics. Yet in all cases, rational/optimality explanations are used to derive empirical predictions; and, as in any scientific enterprise, the assumptions of the rational/optimality accounts are adjusted, where

Response/Oaksford & Chater: Pre´cis of Bayesian Rationality appropriate, to give a better fit with empirical predictions. Specifically, empirical data lead to revision of empirical assumptions in the rationality/optimality analysis – the empirical data does not lead to a revision of the laws of logic, probability, or any other rational theory. Khalil raises the opposite concern: that we use rational explanation too narrowly. He argues that the style of optimality explanation that we advocate applies just as well in the explanation of non-cognitive biological structures as it does to cognitive processes – he argues that, in the sense of rationality used in BR, stomachs are just as rational as cognitive mechanisms. This concern appears purely terminological; we reserve “rationality” for information processing systems. But rational analysis is, indeed, parallel to optimality explanation in biology (Chater et al. 2003).

R4.3. Relevance of the algorithmic level

McKenzie and Griffiths note, however, that advocating rational analysis does not make the challenges concerning algorithmic, and indeed neural, implementation, disappear. Moreover, the mapping between levels of explanation need not necessarily be straightforward, so that a successful probabilistic rational analysis of a cognitive task does not necessarily require that the cognitive system be carrying out probabilistic calculations – any more than the bird is carrying out aerodynamic calculations in growing a wing perfectly adapted for flight. Nonetheless, in many contexts, it is natural to see cognition as carrying out probabilistic calculations; and a prior rational analysis (or, in Marr’s [1982] terms, computational level of explanation) is extremely valuable in clarifying what calculations need to be carried out. Without a “rational analysis” for arithmetic calculations (i.e., a mathematical theory of elementary arithmetic), understanding which algorithms might be used by a pocket calculator, let alone how those algorithms might be implemented in silicon, would be impossible. Griffiths outlines key challenges for creating an algorithmic-level theory of cognition, viewed from a Bayesian perspective; and this perspective dovetails nicely with work viewing neural machinery as carrying out Bayesian inference (e.g., Ma et al. 2006; Rao et al. 2002), which we consider briefly further on. BR is largely focused on rational level explanation (Anderson 1990; 1991a). Indeed, following Marr (1982), we argued that understanding the rational solution to problems faced by the cognitive system crucially assists with explanation in terms of representations and algorithms, as stressed by Hahn and Griffiths. In BR, this is illustrated by our model of syllogistic reasoning, which proposes a set of “fast and frugal” heuristics (Gigerenzer & Goldstein 1996) for generating plausible conclusions, rooted in a Bayesian rational analysis (Chater & Oaksford 1999b). More recently, we have suggested methods for causal and conditional reasoning, based on “mental mechanisms” (Chater & Oaksford 2006; Ali et al., in press) directly building on rational and algorithmic models inspired by the literature on Bayesian networks (Glymour 2001; Pearl 1988; 2000). Moreover, an explicit algorithmic implementation of our probabilistic account of conditional inference has been constructed using a constraint satisfaction neural network (Oaksford & Chater, in press). Moreover, there

is a significant movement in current cognitive science that focuses on developing and employing Bayesian machine learning techniques to model cognition at both the rational and algorithmic levels (e.g., Griffiths et al. 2007; Kemp & Tenenbaum 2008). Evans’ concern that we ignore the algorithmic level is therefore puzzling. He worries that BR recommends that one should “observe some behaviour, assume that it is rational, find a normative theory that deems it to be so, and then . . . nothing else, apparently.” We assume that the ellipsis should, in Evans’ view, be fleshed out with an algorithmic, process-based explanation, which should then be subject to rigorous empirical test. The abovementioned list of algorithmic level proposals inspired by Bayesian rational analysis, both in the domain of reasoning and in cognitive science more generally, gives grounds for reassurance. Moreover, the extensive empirical testing of these models (Green & Over 1997; 2000; McKenzie & Mikkelsen 2000; 2007; McKenzie et al. 2001; Nelson 2005; Oaksford & Moussakowski 2004; Oaksford & Wakefield 2003; Oaksford et al. 1999; 2000; Tenenbaum 1999) should allay concerns that rational analysis provides no testable predictions. Ironically, the only theories in the psychology of reasoning that have been algorithmically specified, aside from those within the Bayesian tradition, are directly based on another rational level theory: logic (Johnson-Laird 1992; Rips 1994). Theorists who have instead focused primarily on heuristics for reasoning have couched their explanations in purely verbal terms (Evans 1989; Evans & Over 2004). This indicates, we suggest, that rational analysis assists, rather than impedes, algorithmic explanation. R4.4. Relevance of neural implementation

Bayesian rational analysis is, moreover, appealing because it appears to yield algorithms that can be implemented in the brain. In BR (Ch. 4), we observed that the Bayesian approach was sweeping across cognitive psychology. We might also have added that its influence in computational neuroscience is at least as significant (Friston 2005). Although our Bayesian analyses of higher-level reasoning do not directly imply Bayesian implementations at the algorithmic level, it is intriguing that influential theorists (Doya et al. 2007; Friston 2005; Ma et al. 2006) view Bayesian inference as providing the driving computational principle for neural information processing. Such models, using population codes (Ma et al. 2006), which avoid treating the brain as representing probabilities directly on a numerical scale, can model simple perceptual decision tasks (Gold & Shadlen 2000). Such convergence raises the possibility that Bayesian rational analyses of reasoning may one day find rather direct neural implementations. De Neys specifically appeals to the implementation level in commenting on BR. He draws attention to imaging studies of reasoning that suggest a role for the anterior cingulate cortex in detecting conflict and inhibiting responses. As we have seen (sect. R3.4), such a role is entirely consistent with Bayesian approaches. Indeed, more broadly, imaging work on human reasoning, pioneered by Goel (e.g., Goel 2007), is at an exploratory stage, and currently provides few constraints on theory. Moreover, as we have seen, where cognitive BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

111

Response/Oaksford & Chater: Pre´cis of Bayesian Rationality neuroscientists concentrate on what computations the brain performs rather than where, the emerging answer is Bayesian. R4.5. Optimality and rational analysis

A range of commentators (e.g., Brighton & Olsson, Danks & Eberhardt, Evans, and Schroyens) argue that the methodology of rational analysis faces conceptual problems. Our general response to these concerns is pragmatic. As with any methodology, we see rational analysis, using probabilistic methods or otherwise, as primarily to be judged by its results. Anderson’s path-breaking work (1990; 1991a), and the huge literature on Bayesian models across the brain and cognitive sciences, of which BR is a part, is therefore, in our view, the best argument for the value of the approach. Parallels with closely related work in behavioral ecology and rational choice explanation in economics give further weight to the view that a “rational” style of explanation can yield considerable insights. But, like any style of explanation, rational analysis has its limits. Just as, in biology, some behaviors or structures are products of “history” rather than adaptation (Carroll 2005), and some economic behaviors are the product of cognitive limitations (e.g., Ariely et al. 2003; Thaler 2005), so in the brain and cognitive sciences, we should expect some phenomena to arise from specific aspects of algorithms/representations or neural implementation. We are therefore happy to agree with commentators who suggest that there are cognitive phenomena for which purely rational considerations provide an incomplete, or indeed incorrect, explanation (e.g., Brighton & Olsson, Evans). We also agree that rational analysis is challenged where there are many, perhaps very different, near-optimal rational solutions (Brighton & Olsson). In such situations, rational analysis provides, at best, a range of options – but it does not provide an explanation of why one has been chosen. Nonetheless, these issues often cause few problems in practice, as the results in BR and in the wider program of rational explanation illustrate. We agree, moreover, with concerns that finding exactly the optimal solution may be over-restrictive (Brighton & Olsson, Evans). Consider the case of perceptual organization, where the cognitive system must decide between multiple interpretations of a stimulus (Gregory 1970; von Helmholtz 1910/1925). Accounts based on Bayesian probability and on the closely related idea of maximizing simplicity (Chater 1996; Hochberg & McAlister 1953; Leeuwenberg & Boselie 1988) adopt the perspective of rational analysis, but they do so comparatively. That is, the perceptual system is presumed to choose interpretation A, rather than interpretation B, if A is more likely than B (or, in simplicity-based formulations, if it provides a simpler encoding of the sensory input). Neither the likelihood nor the simplicity principles in perceptual organization are presumed to imply that the perceptual system can optimize likelihood/simplicity – and indeed, in the general case, this is provably impossible (see Chater 1996, for discussion). Indeed, we suspect that rational analysis will, in many cases, primarily be concerned with providing a measure of the relative “goodness” of different cognitive processes or behaviors; and it is explanatory to 112

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

the degree to which the “good” mechanisms are more prevalent than the “bad.” The parallel with evolutionary explanation seems to be exact here: Inclusive fitness provides a crucial explanatory measure in explaining the evolution of biological structures, but the explanatory “bite” is comparative (i.e., in a certain environment, a flipper yields greater fitness than a leg). There is no assumption that biological evolution, in any context, reaches a state of completely optimized perfection; indeed, quite the reverse (Jacob 1977). Thus, Evans’ emphasis on satisficing rather than optimizing, and Brighton & Olsson’s focus on relative rationality, seem to us entirely consistent with BR. Note, too, that in modeling many aspects of cognition, a full-scale rational analysis (specifying a task, environment, and computational limitations) may not be required. For example, conditional inference can be modeled in Bayesian terms, assuming only a probabilistic interpretation of the premises, and the requirement of maintaining consistent degrees of belief. The success of the probabilistic, rather than a logical, interpretation of the premises can be assessed by comparing the predictions of both approaches to data on human reasoning, as well general philosophical principles. Brighton & Olsson also raise a different concern: that the specific sets of probabilistic assumptions (such as the independence assumptions embodied in naı¨ve Bayes) may sometimes be justified not by rational analysis, but instead in the light of their general, formal properties, combined with empirical success in solving some externally defined task (e.g., estimating the relative sizes of German cities, Gigerenzer & Goldstein 1996). For example, a model such as naı¨ve Bayes, they note, may be effective because it has few parameters and hence avoids over-fitting. We suggest, however, that this is not a separate type of explanation of inferential success, distinct from Bayesian rational analysis. Instead, the justification for preferring simple models can, itself, be provided in terms of Bayesian reasoning, and closely related formalisms, including minimum description length (Chater & Oaksford 2008b; MacKay 2003; Rissanen 1989; Vita´nyi & Li 2000). R4.6. Need rational explanation be causal?

Brighton & Olsson, together with Danks & Eberhardt, raise the fundamental concern that rational explanation does not provide a causal explanation of behavior. We agree. Rational explanation is teleological (Fodor 1968) – it explains by reference to purpose, rather than cause. In particular, rational explanation does not require that the rational analysis is itself represented in the mind of the agent, and does not, therefore, imply that behavior is governed by any such representation. Aerodynamics may provide an optimality-based explanation of the shape of the bird’s wing; but aerodynamic calculations by the bird (or any other agent) are not causally responsible for the wing’s shape. Similarly, delineating the circumstances in which algorithms such as naı¨ve Bayes (Brighton & Olsson; Domingos & Pazzani 1997), Take the Best (Gigerenzer & Goldstein 1996; Martignon & Hoffrage 1999), or unitweighted regression (Dawes 1979) are reliable may require highly sophisticated rational explanation. Yet a

Response/Oaksford & Chater: Pre´cis of Bayesian Rationality cognitive system that employs such models may know nothing of such rational explanations – and indeed, these rational assumptions typically play no causal role in determining the behavior. Thus, in behavioral ecology, for example, the strategies animals use in foraging, mate selection, and so on, are typically explained using optimality explanations; but animals are not assumed to carry out optimality calculations to validate their behavioral strategies. Danks & Eberhardt suggest that there is a “requirement for a teleological explanation that the normative principle must have played a causal role – ontogenetic, phylogenetic, or both – in the behavior’s existence or persistence. ‘Origin stories’ are required for teleological explanation.” We find this claim puzzling: normative principles, and rational explanations in general, are abstract – they are not part of the causal realm. Thus, a Bayesian rational analysis can no more cause a particular piece of behavior or reasoning, than the principles of arithmetic cause a calculator to display a particular number. Teleological explanations are distinctively non-causal, and necessarily so. In this section, we have considered concerns about the general project of rational analysis. We now turn to consider specific issues relating to the rational models and empirical data presented in BR.

R5. Reconsidering models and data Even if the broad sweep of arguments from the preceding sections is endorsed, there remain doubts about the details of the particular models described in BR and their ability to account for human reasoning data. Indeed, in the commentaries, issues of detail emerge most often between researchers who otherwise are in broad agreement. It is in this light that we consider the comments of Liu, Oberauer, Over & Hadjichristidis, and Wagenmakers. We also consider here Halford’s comments on syllogistic reasoning, drawn from a different framework.

R5.1. Conditional inference

Liu, Oberauer, and Over & Hadjichristidis, who have also advocated a probabilistic approach (in particular, to conditional inference), have concerns about our specific model. We addressed, in section R3.6, Over & Hadjichristidis’s argument that we are not Bayesian enough, and that we should employ Jeffrey’s rule to deal with uncertainty in the categorical premise of conditional inference. We pointed out that we too explicitly adopted Jeffrey’s rule in BR. They also cite some unpublished results apparently showing that people have an imperfect understanding of Jeffrey’s rule. These results are intriguing and suggest that more extensive empirical testing of this rule is required.3 Oberauer argues that our models of conditional inference and data selection may lead to absurdity. He argues that if the marginals, P(p) and P(q), remain fixed, which he describes as “axiomatic” in our theory,4 then if one increases the probability that someone gets a headache, given they take drug X, then those who don’t take X will get fewer headaches. This apparent absurdity stems from a conflation in Oberauer’s description between the factual and the epistemic/doxastic: Changing this

conditional degree of belief does not mean that these people actually achieve these benefits. In ignorance of the real conditional probability, but knowing the values of the marginals, I should revise my degree of belief that not taking this drug leads to fewer headaches. Yet this will only be appropriate when the marginals are known – which is clearly inappropriate in Oberauer’s example. Oberauer also perceives an inconsistency between our adoption of The Equation – P(if p then q) ¼ P(qjp) – and our use of a contingency table to represent the conditional hypothesis in data selection. However, by The Equation there is only sufficient information in the premises of a conditional inference to draw MP by Bayesian (or Jeffrey) conditionalization (at least a point value). The remaining inferences can only be drawn on the assumption that people use the marginals to calculate the relevant conditional probabilities, for example, P( : qj : p) for Denying the Antecedent (DA). Once P(qjp) and the marginals are fixed, the contingency table is determined. Knowing the meaning of a statement is often equated with knowing the inferences that a statement licenses (Dowty et al. 1981). According to The Equation, the conditional only licenses “probabilized” MP. Probabilistically, to draw further inferences requires more information to be drawn from world knowledge. Hence, there is no inconsistency. Moreover, in the selection task, people are presented with an array of possible evidence types that makes the marginals relevant in the same way as presenting more than just MP in the conditional inference task. The degree of belief that is modified by selecting data is in the conditional and the marginals, which constitute the dependence and independence models. Thus, Oberauer’s concerns can be readily addressed. Oberauer also suggests that contingency tables are consistent with a probabilistic contrast approach, that is, the measure of the strength of an argument, for example, MP, is P(qjp) – P(qj : p). It is for this reason that we believe that argument strength may indeed be two-dimensional (Oaksford & Hahn 2007). The conditional probability alone can mean that a good argument leads to no increase in the degree of belief in the conclusion, for example, for MP when P(qjp) ¼ P(q) ¼ 1. The probabilistic contrast (and other measures; see, e.g., Nelson, Poletiek, and Oaksford & Hahn 2007) captures the change in the probability of the conclusion brought about by an argument. Oberauer suggests that there is no evidence for people’s use of the probabilistic contrast. Yet Over et al. (2007) found significant sensitivity to P(qj : p), consistent with some use of the probabilistic contrast or a related measure of change, and the evidence is currently equivocal. Oberauer also raises two concerns over evidence for our model of conditional inference. First, fitting a model with two free parameters to four data points “is no convincing accomplishment.” Even so, as Hahn observes, the move to detailed model fitting of quantitative data represents significant progress in the psychology of reasoning (for early examples, see Krauth [1982] and Klauer [1999]). Moreover, in BR (pp. 146 –49) we fitted the model to the 32 data points produced in Oaksford et al.’s (2000) Experiment 1 using only nine parameters, collapsing far more degrees of freedom than the model fitting reported in Oberauer (2006). Although Oberauer (2006) found poorer fits for our model than alternative theories, BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

113

Response/Oaksford & Chater: Pre´cis of Bayesian Rationality Oaksford and Chater (2008) found that the revised model presented in BR may provide better fits to Oberauer’s data. Second, Oberauer argues that the most relevant empirical evidence comes from studies where probabilities were directly manipulated, of which he mentions two, Oaksford et al. (2000) and Oberauer et al. (2004). Moreover, he argues that their results are equivocal. However, several other studies have manipulated probabilities in conditional inference and found evidence in line with a probabilistic account (George 1997; Liu 2003; Liu et al. 1996; Stevenson & Over 1995). Oberauer also leaves aside the many studies on data selection showing probabilistic effects (see BR, Ch. 6). Liu’s arguments about second-order conditionalization point, we think, to an important factor that we have yet to consider in reasoning, that is, the effects of context. Liu has found that people often endorse the conclusion that, for example, Tweety flies on being told that Tweety is a bird in the absence of the conditional premise (reduced problems). This occurs because they fill in this information from world knowledge. However, Liu also found that endorsements increase when the conditional premise is added (complete problems). In BR, we argued that this occurs because people take the conditional premise as evidence that the conditional probability is higher (an inference that may arise from conversational pragmatics). Liu argues that our account implies that manipulations affecting reduced problems should also affect complete problems and provides evidence against this. Yet context, both cognitive and physical, may explain these differences in a way similar to recent studies of decision-making (Stewart et al. 2006). For example, suppose one is told about two swanneries, both containing the same number of swans. In one, 90% of swans are black (P(blackjswan) ¼ .9); in the other, 90% of swans are white (P(whitejswan) ¼ .9). On being told that Tweety is a swan, presumably one would only endorse Tweety is white at .5. This is because conversational pragmatics and world knowledge indicate that Tweety is in one of the just mentioned swanneries, but the dialogue up to this point does not indicate which one.5 However, the addition of the conditional premise if a bird is a swan it is white immediately indicates which swannery is being talked about, that is, the one in which P(whitejswan) is high, and now endorsements should increase to .9. Clearly, although manipulations of the relative number of swans in each swannery might affect the reduced problem, they should not affect the complete problem. So if the swannery in which most swans are black were one tenth of the size of the other swannery, then, given natural sampling assumptions, endorsements for the reduced problem should increase to .83, but endorsements of the complete problem should remain the same. R5.2. Data selection

Wagenmakers raises a variety of concerns about our optimal data selection model. First, why do we concede that people should select the standard “logical” A card and 7 card choices, if the rule only applies to the four cards? In BR (p. 210), we argue that people rarely use conditionals to describe just four objects – they assume that the cards are drawn from a larger population. 114

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

Consequently, we quite explicitly do not make the counterintuitive prediction that Wagenmakers ascribes to us. Second, Wagenmakers wonders why – when all cards carry some information – do participants not select all the cards, if they are maximizing information gain? We assume that the pragmatics of the task suggests to participants that they should select some cards, but not others (BR, pp. 200 –201). Third, Wagenmakers suggests that incentivized individuals with more time might make the logical response. Work on individual differences (e.g., Stanovich & West 2000) is consistent with the view that logical competence is learned, either directly (e.g., studying logic or math) or indirectly (e.g., learning to program or learning conventional, non-Bayesian statistics); such logical competence is a prerequisite for “logical” responses, and covaries with IQ as measured in University populations. Wagenmakers also remarks that, as Bayesians, we should avoid null hypothesis testing in statistically assessing our models. This choice is purely pragmatic: it conforms to the current demands of most journals. R5.3. Syllogisms and development

Halford argues that mental models theory and a relational complexity measure fit the data as well as the probability heuristics model (PHM), conceding, however, that only PHM generalizes to most and few. Copeland (2006) has also recently shown that PHM provides better fits than mental models and mental logic for extended syllogisms involving three quantified premises. Halford also suggests that basing confidence in the conclusion on the least probable premise, as in our max-heuristic, is counterintuitive. He proposes that confidence should instead be based on relational complexity, which covaries with the least probable premise. But perhaps Halford’s intuition goes the wrong way: the least probable premise is the most informative; and surely the more information you are given, the stronger the conclusions you can draw? De Neys and Straubinger, Cokely, & Stevens (Straubinger et al.) both argue that there are important classes of evidence that we do not address. De Neys argues that attention to latency data and imaging studies provides a greater role for logic, a claim we disputed earlier. Note, also, that the algorithmic theory in PHM has been applied to latency data and accounts for the data, as well as mental models (Copeland & Radvansky 2004). Straubinger et al. are concerned that we ignore developmental data. In particular, they view the findings on the development of working memory as providing a particular challenge to a Bayesian approach. They do, however, acknowledge that in different areas (e.g., causal reasoning), Bayesian ideas are being successfully applied to developmental data (Navarro et al. 2006; Sobel et al. 2004). Straubinger et al.’s emphasis on working memory provides good reason to believe that our particular approach to deductive reasoning may extend to development. Copeland and Radvansky (2004) explicitly related working-memory limitations to PHM, finding that it provided as good an explanation as mental models theory of the relationship between working-memory capacity and reasoning performance. This result provides some indication that, at least for syllogistic reasoning, developmental trajectories explicable by mental models may be similarly amenable

References/Oaksford & Chater: Pre´cis of Bayesian Rationality to explanation in terms of probability heuristics. Our approach also provides a natural way in which experience, leading to the learning of environmental statistics, might influence reasoning development. Exploring these possibilities must await future research. R6. The Bayesian turn BR is part of a larger movement across the brain and cognitive sciences – a movement which sees cognition as centrally concerned with uncertainty; and views Bayesian probability as the appropriate machinery for dealing with uncertainty. Probabilistic ideas have become central to theories of elementary neural function (Doya et al. 2007), motor control (Ko¨rding & Wolpert 2004), perception (Knill & Richards 1996), language processing (Manning & Schu¨tze 1999), and high-level cognition (Chater & Oaksford 2008a; Chater et al. 2006). They also cut across Marr’s (1982) computational (Anderson 1990; Pearl 2000), algorithmic (Jacobs et al. 1991), and implementational (Doya et al. 2007) levels of explanation. In arguing that commonsense reasoning should be understood in terms of probability, we are merely recasting Laplace’s (1814/1951) classic dictum concerning the nature of probability theory: “The theory of probabilities is at bottom nothing but common sense reduced to calculus.”

NOTES 1. Although Braine and O’Brien (1991) explicitly reject the use of relevance logic (Anderson & Belnap 1975), this does provide an interesting possible route for developing these ideas. In particular, interpretations of the semantics of relevance logics as a ternary relation between possible worlds, or from an information-theoretic perspective, as a ternary relation between a source, a receiver, and a channel (Restall 1996), may provide interesting connections with nonmonotonic reasoning. 2. By contrast, we know of just one paper in the psychology of reasoning discussing Dempster-Shafer belief functions, namely, George (1997). 3. Its normative status has also been questioned for many years (see, e.g., Field 1978). 4. This is despite the fact that they were not fixed in Oaksford and Chater (1994). 5. Of course, different assumptions would yield different results. For example, if the previous dialogue had been talking about the swannery, where most swans are black, just before introducing Tweety, the assumption may be that Tweety comes from that swannery and so Tweety is white might only be endorsed at .1.

References [The letters “a” and “r” before author’s initials stand for target article and response references, respectively] Adams, E. W. (1975) The logic of conditionals: An application of probability to deductive logic. Reidel. [arMO, NP] (1998) A primer of probability logic. CLSI Publications. [arMO] Ali, N., Schlottmann, A., Shaw, A., Chater, N. & Oaksford, M. (in press) Causal discounting and conditional reasoning in children. In: Cognition and conditionals: Probability and logic in human thought, ed. M. Oaksford & N. Chater. Oxford University Press. [rMO] Anderson, A. & Belnap, N. D. (1975) Entailment: The logic of relevance and necessity, vol. 1. Princeton University Press. [rMO] Anderson, J. R. (1983) The architecture of cognition. Harvard University Press. [rMO]

(1990) The adaptive character of thought. Erlbaum. [TLG, UH, ELK, CRMM, arMO, WS] (1991a) Is human cognition adaptive? Behavioral and Brain Sciences 14:471– 84; discussion 485 –517. [HB, arMO, NS] (1991b) The adaptive nature of human categorization. Psychological Review 98:409 – 29. [aMO] Anderson, J. R. & Matessa, M. (1998) The rational analysis of categorization and the ACT-R architecture. In: Rational models of cognition, ed. M. Oaksford & N. Chater, pp.197 – 217. Oxford University Press. [aMO] Anderson, J. R. & Milson, R. (1989) Human memory: An adaptive perspective. Psychological Review 96:703– 19. [aMO] Anderson, J. R. & Schooler, L. J. (1991) Reflections of the environment in memory. Psychological Science 2:396– 408. [aMO] Antoniou, G. (1997) Nonmonotonic reasoning. MIT Press. [KS] Ariely, D., Loewenstein, G. & Prelec, D. (2003) Coherent arbitrariness: Stable demand curves without stable preferences. Quarterly Journal of Economics 118:73 – 105. [rMO] Aristotle (1980) Nicomachean ethics, trans. W. D. Ross. Clarendon Press. [aMO] Ball, L. J., Philips, P., Wade, C. N. & Quayle, J. D. (2006) Effects of belief and logic on syllogistic reasoning: Eye-movement evidence for selective processing models. Experimental Psychology 53:77 – 86. [WDN] Baltes, P. B., Staudinger, U. M. & Lindenberger, U. (1999) Lifespan psychology: Theory and application to intellectual functioning. Annual Review of Psychology 50:471 – 507. [NS] Baron, J. (1981) An analysis of confirmation bias. Paper presented at the 22nd Annual Meeting of the Psychonomic Society. 6 – 8 November, 1981, Philadelphia, PA. [aMO] (1985) Rationality and intelligence. Cambridge University Press. [JDN, aMO] (2004) Normative models of judgment and decision making. In: Blackwell handbook of judgment and decision making, ed. D. J. Koehler & N. Harvey, pp. 19– 36. Blackwell. [JDN] Barwise, J. & Cooper, R. (1981) Generalized quantifiers and natural language. Linguistics and Philosophy 4:159– 219. [aMO] Becker, G .S. (1976) The economic approach to human behaviour. University of Chicago Press. [ELK] Benferhat, S., Bonnefon, J. F. & Da Silva Neves, R. (2005) An overview of possibilistic handling of default reasoning, with experimental studies. Synthese 146:53 – 70. [GP] Bennett, J. (2003) A philosophical guide to conditionals. Oxford University Press. [arMO] Bernoulli, J. (1713/2005) Ars conjectandi [The art of conjecture], trans. E. D. Sylla. Johns Hopkins University Press. [aMO] Biazzo, V., Gilio, A., Lukasiewicz, T. & Sanfilippo, G. (2005) Probabilistic logic under coherence: Complexity and algorithms. Annals of Mathematics and Artificial Intelligence 45(1 – 2):35 – 81. [NP] Blakemore, C., Adler, K. & Pointon, M. (1990) Vision: Coding and efficiency. Cambridge University Press. [rMO] Bookstaber, R. & Langsam, J. (1985) On the optimality of coarse behavior rules. Journal of Theoretical Biology 116:161– 93. [HB] Boole, G. (1854/1958) An investigation of the laws of thought. Macmillan/Dover. (Reprinted by Dover, 1958). [aMO] Bovens, L. & Hartmann, S. (2003) Bayesian epistemology. Clarendon Press. [aMO] Braine, M. D. S. (1978) On the relation between the natural logic of reasoning and standard logic. Psychological Review 85:1 – 21. [aMO] (1990) The natural logic approach to reasoning. In: Reasoning, necessity, and logic: Developmental perspectives, ed. W. F. Overton. Erlbaum. [DPO] Braine, M. D. S. & O’Brien, D. P. (1991) A theory of if : A lexical entry, reasoning program, and pragmatic principles. Psychological Review 98:182 – 203. [DPO, rMO] Braine, M. D. S. & O’Brien, D. P., eds. (1998) Mental logic. Erlbaum. [DPO] Brighton, H. & Gigerenzer, G. (2008) Bayesian brains and cognitive mechanisms: Harmony or dissonance? In: The probabilistic mind: Prospects for Bayesian cognitive science, ed. N. Chater & M. Oaksford, pp. 189 – 208, Oxford University Press. [HB, NS] Bullinaria, J. A. & Chater, N. (1995) Connectionist modelling: Implications for neuropsychology. Language and Cognitive Processes 10:227 – 64. [UH] Byrne, R. M. J. (1989) Suppressing valid inferences with conditionals. Cognition 31:61– 83. [KS] Byrne, R. M. J., Espino, O. & Santamaria, C. (1999) Counterexamples and the suppression of inferences. Journal of Memory and Language 40:347 – 73. [I-mL] Carnap, R. (1950) The logical foundations of probability. University of Chicago Press. [rMO, FHP] Carroll, S. (2005) Endless forms most beautiful. W. W. Norton. [rMO] Chater, N. (1995) Neural networks: The new statistical models of mind. In: Connectionist models of memory and language, ed. J. P. Levy, D. Bairaktaris,

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

115

References/Oaksford & Chater: Pre´cis of Bayesian Rationality J. A. Bullinaria & P. Cairns, pp. 207 – 27. University College London Press. [UH] (1996) Reconciling simplicity and likelihood principles in perceptual organization. Psychological Review 103:566 –81. [rMO] Chater, N., Crocker, M. & Pickering, M. (1998) The rational analysis of inquiry: The case of parsing: In: Rational models of cognition, ed. M. Oaksford & N. Chater, pp. 441 – 68. Oxford University Press. [aMO] Chater, N. & Manning, C. D. (2006) Probabilistic models of language processing and acquisition. Trends in Cognitive Sciences 10:335 – 44. [TLG, aMO] Chater, N. & Oaksford, M. (1999a) Ten years of the rational analysis of cognition. Trends in Cognitive Science 3:57 –65. [TLG] (1999b) The probability heuristics model of syllogistic reasoning. Cognitive Psychology 38:191 – 258. [arMO] (2006) Mental mechanisms: Speculations on human causal learning and reasoning. In: Information sampling and adaptive cognition, ed. K. Fiedler & P. Juslin, pp. 210 – 38. Cambridge University Press. [rMO] eds. (2008a) The probabilistic mind: Prospects for Bayesian cognitive science. Oxford University Press. [UH, rMO] (2008b) The probabilistic mind: Where next? In: The probabilistic mind: Prospects for Bayesian cognitive science, ed. N. Chater & M. Oaksford, pp. 501 – 14. Oxford University Press. [rMO] Chater, N., Oaksford, M., Heit, E. & Hahn, U. (in press) Inductive logic and empirical psychology. In: The handbook of philosophical logic, vol. 10, ed. S. Hartmann & J. Woods. Springer. [rMO] Chater, N., Oaksford, M., Nakisa, R. & Redington, M. (2003) Fast, frugal and rational: How rational norms explain behavior. Organizational Behavior and Human Decision Processes 90:63– 86. [JDN, rMO, WS] Chater, N., Tenenbaum, J. B. & Yuille, A. eds. (2006) Probabilistic models of cognition: Where next? Trends in Cognitive Sciences 10:335 – 44. [Special Issue.] [UH, rMO ] Chater, N. & Vita´nyi, P. (2002) Simplicity: A unifying principle in cognitive science? Trends in Cognitive Sciences 7:19 – 22. [rMO] Cheng, P. W. & Holyoak, K. J. (1985) Pragmatic reasoning schemas. Cognitive Psychology 17:391 – 416. [aMO] Chomsky, N. (1957) Syntactic structures. Mouton. [aMO] (1965) Aspects of the theory of syntax. MIT Press. [aMO] Chou, T. H. (2007) The mechanism of suppression effects in conditional reasoning. Unpublished Doctoral dissertation, National Chung-Cheng University, Department of Psychology, Chia-Yi, Taiwan. [I-mL] Cohen, L. J. (1981) Can human irrationality be experimentally demonstrated? Behavioral and Brain Sciences 4:317– 70. [JDN, aMO] Cokely, E. T., Kelley, C. M. & Gilchrist, A. L. (2006) Sources of individual differences in working memory: Contributions of strategy to capacity. Psychonomic Bulletin and Review 13:991 – 97. [NS] Coletti, G. & Scozzafava, R. (2002) Probabilistic logic in a coherent setting. Kluwer. [NP] Conlin, J. A., Gathercole, S. E. & Adams, J. W. (2005) Children’s working memory: Investigating performance limitations in complex span tasks. Journal of Experimental Child Psychology 90:303 – 17. [NS] Cooper, G. F. (1990) The computational complexity of probabilistic inference using Bayesian belief networks. Artificial Intelligence 42:393 – 405. [HB] Copeland, D. E. (2006) Theories of categorical reasoning and extended syllogisms. Thinking and Reasoning 12:379 – 412. [rMO] Copeland, D. E. & Radvansky, G. A. (2004) Working memory and syllogistic reasoning. Quarterly Journal of Experimental Psychology 57A:1437 – 57. [arMO] Cosmides, L. (1989) The logic of social exchange: Has natural selection shaped how humans reason? Studies with the Wason selection task. Cognition 31:187 – 276. [arMO] Cosmides, L. & Tooby, J. (1996) Are humans good intuitive statisticians after all? Rethinking some conclusions from the literature on judgment under uncertainty. Cognition 58:1 – 73. [JDN] (2000) Evolutionary psychology and the emotions. In: Handbook of emotions, 2nd edition, ed. M. Lewis & J. M. Haviland-Jones, pp. 91– 115. Guilford. [aMO] Courville, A. C., Daw, N. D. & Touretzky, D. S. (2006) Bayesian theories of conditioning in a changing world. Trends in Cognitive Sciences 10:294 – 300. [aMO] Damer, E. T. (2005) Attacking faulty reasoning: A practical guide to fallacy-free arguments, 5th edition. Thomson Wadsworth. [WS] Danks, D. (2008) Rational analyses, instrumentalism, and implementations. In: The probabilistic mind: Prospects for Bayesian cognitive science, ed. N. Chater & M. Oaksford. Oxford University Press. [HB] Daston, L. (1988) Classical probability in the enlightenment. Princeton University Press. [aMO] Davidson, D. (1984) Inquiries into truth and interpretation. Oxford University Press. [aMO]

116

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

Dawes, R. M. (1979) The robust beauty of improper linear models in decision making. American Psychologist 34:571 – 82. [rMO] de Finetti, B. (1974/1975) Theory of probability, vols. 1 & 2. Wiley. [NP] Dempster, A. P. (1968) A generalization of Bayesian inference. Journal of the Royal Statistical Society, Series B 30:205 – 47. [rMO] Dennett, D. (1987) The intentional stance. MIT Press. [WS] De Neys, W. & Franssens, S. (2007) The nature of belief inhibition during thinking: How reasoning impairs memory. Proceedings of the Annual Meeting of the Cognitive Science Society 29:8. [WDN] De Neys, W. & Glumicic, T. (2008) Conflict monitoring in dual process theories of thinking. Cognition 106:1248 – 99. [WDN] De Neys, W., Vartanian, O. & Goel, V. (2008) Smarter than we think: When our brains detect that we are biased. Psychological Science 19:483 – 89. [WDN, rMO] Dick, F., Leech, R., Moses, P. & Saccuman, M. C. (2006) The interplay of learning and development in shaping neural organization. Developmental Science 9:14 – 17. [NS] Domingos, P. & Pazzani, M. (1997) On the optimality of the simple Bayesian classifier under zero-one loss. Machine Learning 29:103 –30. [HB, rMO] Dowty, D. R., Wall, R. E. & Peters, S. (1981) Introduction to Montague semantics. Springer. [rMO] Doya, K., Ishii, S., Rao, R. P. N. & Pouget, A., eds. (2007) The Bayesian brain: Probabilistic approaches to neural coding. MIT Press. [rMO] Dubois, D. & Prade, H. (1988) Possibility theory. Plenum. [GP] Earman, J. (1992) Bayes or bust? MIT Press. [arMO] Edgington, D. (1995) On conditionals. Mind 104:235 – 329. [UH, arMO] Elqayam, S. (2008) Normative rationality and the is-ought fallacy. In: Proceedings of the Second European Cognitive Science Conference, ed. S. Vosiniadou, D. Kayser & A. Protopaps, pp. 249 –99. Psychology Press. [JStBTE] Ericsson, K. A., Charness, N., Hoffmann, R. R. & Feltovich, P. J. (2006) Cambridge handbook of expertise and expert performance. Cambridge University Press. [NS] Ericsson, K. A. & Kintsch, W. (1995) Long-term working memory. Psychological Review 102:211– 45. [NS] Ericsson, K. A. & Lehmann, A. C. (1996) Expert and exceptional performance: Evidence of maximal adaptation to task constraints. Annual Reviews in Psychology 47:273 – 305. [NS] Evans, J. St. B. T. (1972) Reasoning with negatives. British Journal of Psychology 63:213 – 19. [aMO] (1977) Linguistic factors in reasoning. Quarterly Journal of Experimental Psychology 29A:297 – 306. [I-mL] (1989) Bias in human reasoning: Causes and consequences. Erlbaum. [rMO] (2002) Logic and human reasoning: An assessment of the deduction paradigm. Psychological Bulletin 128:978– 96. [JStBTE] (2007) Hypothetical thinking: Dual processes in reasoning and judgement. Psychology Press. [JStBTE, rMO] (2008) Dual-processing accounts of reasoning, judgment and social cognition. Annual Review of Psychology 59:255– 78. [JStBTE] Evans, J. St. B. T. & Frankish, K., eds. (in press) In two minds: Dual processes and beyond. Oxford University Press. [rMO] Evans, J. St. B. T. & Handley, S. J. (1999) The role of negation in conditional inference. Quarterly Journal of Experimental Psychology 52A:739 – 69. [KO, aMO] Evans, J. St. B. T., Handley, S. J., Harper, C. N. J. & Johnson-Laird, P. N. (1999) Reasoning about necessity and possibility: A test of the mental model theory of deduction. Journal of Experimental Psychology: Learning, Memory, and Cognition 25:1495– 1513. [aMO] Evans, J. St .B. T., Handley, S. J. & Over, D. E. (2003) Conditionals and conditional probability. Journal of Experimental Psychology: Learning, Memory and Cognition 29321 – 55. [I-mL, aMO] Evans, J. St. B. T. & Lynch, J. S. (1973) Matching bias in the selection task. British Journal of Psychology 64:391– 97. [aMO] Evans, J. St. B. T., Newstead, S. E. & Byrne, R. J. (1993) Human reasoning. Erlbaum. [aMO] Evans, J. St. B. T. & Over, D. E. (1996a) Rationality and reasoning. Psychology Press. [JStBTE, DEO, arMO] (1996b) Rationality in the selection task: Epistemic utility versus uncertainty reduction. Psychological Review 103:356 – 63. [aMO] (2004) If. Oxford University Press. [JStBTE, UH, arMO, NP] Evans, J. St. B. T., Stevenson, R. J., Over, D. E., Handley, S. J. & Sloman, S. A. (2004) ESRC end of grant report, No. R00239074. The Economic and Social Research Council. [DEO] Falk, R. & Wilkening, F. (1998) Children’s construction of fair chances: Adjusting probabilities. Developmental Psychology 34:1340– 57. [NS] Field, H. (1978) A note on Jeffrey conditionalization. Philosophy of Science 45:361– 67. [rMO] Fodor, J. A. (1968) Psychological explanation. Random House. [rMO] (1983) The modularity of mind. MIT Press. [UH, aMO]

References/Oaksford & Chater: Pre´cis of Bayesian Rationality Fodor, J. A. & Pylyshyn, Z. (1988) Connectionism and cognitive architecture: A critical analysis. Cognition 28:183 –204. [UH] Ford, M. (2004) System LS: A three tiered nonmonotonic reasoning system. Computational Intelligence 20:89 – 108. [GP] Franks, B. (1995) On explanation in the cognitive sciences: Competence, idealization, and the failure of the classical cascade. British Journal of the Philosophy of Science 46(4):475– 502. [WS] (1999) Idealizations, competence and explanation: A response to Patterson. British Journal of the Philosophy of Science 50:735 – 46. [WS] Friedman, M. (1953) Essays in positive economics. University of Chicago Press. [ELK] Friedman, N., Getoor, L., Koller, D. & Pfeffer, A. (1999) Learning probabilistic relational models. In: Proceedings of the 16th International Joint Conference on Artificial Intelligence (IJCAI), ed. T. Dean, pp. 1300 – 309. Morgan Kaufmann. [TLG] Friedman, N. & Halpern, J. Y. (1995) Plausibility measures: A user’s guide. In: Proceedings of the eleventh conference on uncertainty in AI, ed. P. Besnard & S. Hanks, pp. 175 – 84. Morgan Kaufmann. [GP] Friston, K. (2005) A theory of cortical responses. Philosophical Transactions of the Royal Society B 360:815 – 36. [rMO] Galotti, K. M., Komatsu, L. K. & Voelz, S. (1997) Children’s differential performance on deductive and inductive syllogisms. Developmental Psychology 33:70– 78. [NS] Ga¨rdenfors, P. (1986) Belief revisions and the Ramsey test for conditionals. Philosophical Review 95:81 – 93. [rMO] George, C. (1997) Reasoning from uncertain premises. Thinking and Reasoning 3:161– 90. [rMO] Geurts, B. (2003) Reasoning with quantifiers. Cognition 86:223 – 51. [aMO] Gigerenzer, G. (1996) On narrow norms and vague heuristics: A reply to Kahneman and Tversky. Psychological Review 103:592– 96. [CRMM] (2004) Fast and frugal heuristics: The tools of bounded rationality. In: Blackwell handbook of judgment and decision making, ed. D. J. Koehler & N. Harvey, pp. 62 – 88. Blackwell. [JStBTE] Gigerenzer, G. & Goldstein, D. (1996) Reasoning the fast and frugal way: Models of bounded rationality. Psychological Review 103:650– 69. [arMO] Gigerenzer, G. & Hoffrage, U. (1995) How to improve Bayesian reasoning without instruction: Frequency formats. Psychological Review 102(4):684 –704. [JDN, aMO] Gigerenzer, G., Swijinck, Z., Porter, T., Daston, L., Beatty, J. & Kruger, L. (1989) The empire of chance. Cambridge University Press. [aMO] Gigerenzer, G., Todd, P. & the ABC Research Group. (1999) Simple heuristics that make us smart. Oxford University Press. [HB, arMO] Gilio, A. (2002) Probabilistic reasoning under coherence in System P. Annals of Mathematics and Artificial Intelligence 34:5 –34. [NP] Gilovich, T., Griffin, D. & Kahneman, D. (2002) Heuristics and biases: The psychology of intuitive judgement. Cambridge University Press. [JStBTE, CRMM] Glymour, C. (2001) The mind’s arrow. MIT Press. [rMO] Goel, V. (2007) The anatomy of deduction. Trends in Cognitive Science 11:435 – 41. [rMO] Gold, J. I. & Shadlen, M. N. (2000) Representation of a perceptual decision in developing oculomotor commands. Nature 404:390 – 94. [rMO] Good, I. J. (1950) Probability and the weighing of evidence. Griffin. [JDN] (1975) Explicativity, corroboration, and the relative odds of hypotheses. Synthese 30:39– 73. [JDN] Green. D. W. & Over, D. E. (1997) Causal inference, contingency tables and the selection task. Current Psychology of Cognition 16:459 – 87. [arMO] (2000) Decision theoretical effects in testing a causal conditional. Current Psychology of Cognition 19:51 – 68. [arMO] Gregory, R. L. (1970) The intelligent eye. Weidenfeld & Nicolson. [rMO] Grice, H. P. (1975) Logic and conversation. In: The logic of grammar, ed. D. Davidson & G. Harman, pp. 64 – 75. Dickenson. [rMO] Griffin, D. R. (1992) Animal minds. University of Chicago Press. [ELK] Griffiths, T. L. & Ghahramani, Z. (2006) Infinite latent feature models and the Indian buffet process. In: Advances in neural information processing systems, vol. 18, ed. Y. Weiss, B. Scholkopf & J. Plaut, pp. 475 – 82. MIT Press. [TLG] Griffiths, T. L., Steyvers, M. & Tenenbaum, J. B. (2007) Topics in semantic representation. Psychological Review 114:211 – 44. [rMO] Griffiths, T. L. & Tenenbaum, J. B. (2005) Structure and strength in causal induction. Cognitive Psychology 51:354– 84. [TLG, arMO] Hacking, I. (1975) The emergence of probability. Cambridge University Press. [aMO] (1990) The taming of chance. Cambridge University Press. [aMO] Hadjichristidis, C., Sloman, S. A. & Over, D. E. (in preparation) Revision of beliefs from probabilistic categorical arguments. [DEO] Hahn, U. & Oaksford, M. (2007) The rationality of informal argumentation: A Bayesian approach to reasoning fallacies. Psychological Review 114:704 – 32. [rMO]

Halford, G. S., Cowan, N. & Andrews, G. (2007) Separating cognitive capacity from knowledge: A new hypothesis. Trends in Cognitive Sciences 11(6):236 – 42. [GSH] Halpern, J. Y. (1997) Defining relative likelihood in partially-ordered preferential structures. Journal of Artificial Intelligence Research 7:1– 24. [GP] Harman, G. (1986) Change in view: Principles of reasoning. MIT Press. [rMO] Hattori, M. (2002) A quantitative model of optimal data selection in Wason’s selection task. Quarterly Journal of Experimental Psychology 55A:1241 – 72. [aMO] Hawks, J., Wang, E. T., Cochran, G. M., Harpending, H. C. & Moyzis, R. K. (2007) Recent acceleration of human adaptive evolution. Proceedings of the National Academy of Sciences USA 104(52):20753– 58. [JDN] Hawthorn, J. (2008) Inductive logic. In: Stanford Encyclopedia of Philosophy. Available at: http://plato.stanford.edu/entries/logic-inductive/. [rMO] Hawthorne, J. & Makinson, D. (2007) The quantitative/qualitative watershed for rules of uncertain inference. Studia Logica 86:247 – 97. [NP] Henle, M. (1978) Foreword. In: Human reasoning, ed. R. Revlin & R. E. Mayer. Winston. [aMO] Hilton, D. J., Kemmelmeier, M. & Bonnefon, J-F. (2005) Putting Ifs to work: Goalbased relevance in conditional directives. Journal of Experimental Psychology: General 134:388– 405. [rMO] Hochberg, J. & McAlister, E. (1953) A quantitative approach to figure “goodness.” Journal of Experimental Psychology 46:361 – 64. [rMO] Hoffman, D. (in press) The interface theory of perception: Natural selection drives true perception to swift extinction. In: Object categorization: Computer and human vision perspectives, ed. S. Dickinson, M. Tarr, A. Leonardis & B. Schiele. Cambridge University Press. [JDN] Hoffrage, U. (2000) Why the analyses of cognitive processes matter. Behavioral and Brain Sciences 23:679 –80. [WDN] Horwich, P. (1982) Probability and evidence. Cambridge University Press. [aMO] Houde´, O., Zago, L., Mellet, E., Moutier, S., Pineau, A., Mazoyer, B. & TzourioMazoyer, N. (2000) Shifting from the perceptual brain to the logical brain: The neural impact of cognitive inhibition training. Journal of Cognitive Neuroscience 12:721 – 28. [WDN, rMO] Howe, M. L. & Rabinowitz, F. M. (1996) Reasoning from memory: A lifespan inquiry into the necessity of remembering when reasoning about class inclusion. Journal of Experimental Child Psychology 61:1 – 42. [NS] Howson, C. & Urbach, P. (1993) Scientific reasoning: The Bayesian approach, 2nd edition. Open Court. [DEO, aMO] Hurley, S. & Nudds, M., eds. (2006) Rational animals? Oxford University Press. [rMO, ELK] Inhelder, B. & Piaget, J. (1955) De la logique de l’enfant a` la logique de l’adolescent. Presses Universitaires de France. (English version: The growth of logical thinking from childhood to adolescence. Routledge, 1958). [aMO] Jacob F. (1977) Evolution and tinkering. Science 196:1161– 66. [rMO] Jacobs, R. A., Jordan, M. I., Nowlan, S. & Hinton, G. E. (1991) Adaptive mixtures of local experts. Neural Computation 3:1– 12. [rMO] Jeffrey, R. C. (1967) Formal logic: Its scope and limits, 2nd edition. McGraw-Hill. [rMO] (1983) The logic of decision, 2nd edition. University of Chicago Press. [DEO, arMO] Johnson-Laird, P. N. (1983) Mental models. Cambridge University Press. [arMO] (1992) Syllogs (computer program). Available at: http://webscript.princeton. edu/~mentmod/models.php. [rMO] Johnson-Laird, P. N. & Byrne, R. M. J. (1991) Deduction. Erlbaum. [aMO] (2002) Conditionals: A theory of meaning, pragmatics, and inference. Psychological Review 109:646– 78. [aMO] Juhl, C. F. (1993) Bayesianism and reliable scientific inquiry. Philosophy of Science 60:302 – 19. [DD] Kahneman, D., Slovic, P. & Tversky, A., eds. (1982) Judgment under uncertainty: Heuristics and biases. Cambridge University Press. [CRMM, aMO] Kahneman, D. & Tversky, A. (2000) Choices, values, and frames. Cambridge University Press. [CRMM] Kakade, S. & Dayan, P. (2002) Acquisition and extinction in autoshaping. Psychological Review 109:533– 44. [aMO] Kant, E. (1787/1961) Critique of pure reason, First edition, second impression, trans. N. K. Smith. Macmillan. [aMO] Kearns, M., Mansour, Y., Ng, A. Y. & Ron, D. (1997) An experimental and theoretical comparison of model selection methods. Machine Learning 27:7 – 50. [HB] Kemp, C. & Tenenbaum, J. B. (2008) The discovery of structural form. Proceedings of the National Academy of Sciences USA 105:10687– 92. [rMO] Kirby, K. N. (1994) Probabilities and utilities of fictional outcomes in Wason’s four card selection task. Cognition 51:1– 28. [aMO] Klauer, K. C. (1999) On the normative justification for information gain in Wason’s selection task. Psychological Review 106:215– 22. [arMO]

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

117

References/Oaksford & Chater: Pre´cis of Bayesian Rationality Klauer, K. C., Stahl, C. & Erdfelder, E. (2007) The abstract selection task: New data and an almost comprehensive model. Journal of Experimental Psychology: Learning, Memory and Cognition 33:680 – 703. [UH] Klayman, J. & Ha, Y.-W. (1987) Confirmation, disconfirmation, and information. Psychological Review 94:211 – 28. [JDN] Kneale, W. & Kneale, M. (1962) The development of logic. Oxford University Press. [DPO] Knill, D. & Richards, W., eds. (1996) Perception as Bayesian inference. Cambridge University Press. [arMO] Koehler, D. J. & Harvey, N. (2004) Blackwell handbook of judgement and decision making. Blackwell. [JStBTE] Ko¨rding, K. P. & Wolpert, D. (2004) Bayesian integration in sensorimotor learning. Nature 427:244– 47. [rMO] Kowalski, R. (1979) Algorithm ¼ Logic þ Control. Communications of the Association for Computing Machinery 22:424 – 36. [rMO] Kraus, S., Lehmann, D. & Magidor, M. (1990) Nonmonotonic reasoning, preferential models and cumulative logics. Artificial Intelligence 44:167 – 207. [NP] Krauth, J. (1982) Formulation and experimental verification of models in propositional reasoning. Quarterly Journal of Experimental Psychology 34:285 – 98. [rMO] Kreps, D. M. (1990) A course in microeconomic theory. Princeton University Press. [ELK] Krueger, J. I. & Funder, D. C. (2004) Towards a balanced social psychology: Causes, consequences, and cures for the problem-seeking approach to social behavior and cognition. Behavioral and Brain Sciences 27(3):313– 76. [CRMM] Kuhn, T. (1962) The structure of scientific revolutions. University of Chicago Press. [aMO] Kuncheva, L. I. (2006) On the optimality of Naı¨ve Bayes with dependent binary features. Pattern Recognition Letters 27:830 – 37. [HB] Kyllonen, P. C. & Christal, R. E. (1990) Reasoning ability is (little more than) working-memory capacity?! Intelligence 14:389 – 433. [NS] Lakatos, I. (1970) Falsification and the methodology of scientific research programmes. In: Criticism and the growth of knowledge, ed. I. Lakatos & A. Musgrave, pp. 91– 196. Cambridge University Press. [aMO] Laplace, P. S. (1951) A philosophical essay on probabilities, trans. F. W. Truscott & F. L. Emory. Dover. (Original work published 1814). [rMO] Lauritzen, S. & Spiegelhalter, D. (1988) Local computations with probabilities on graphical structures and their application to expert systems. Journal of the Royal Statistical Society B 50:157 – 224. [rMO] Leeuwenberg, E. & Boselie, F. (1988) Against the likelihood principle in visual form perception. Psychological Review 95:485 –91. [rMO] Levi, I. (1988) The demons of decision. Monist 70:193– 211. [DD] (2002) Money pumps and diachronic books. Philosophy of Science 69:S235 – 47. [DD] Lindley, D. V. (1956) On a measure of the information provided by an experiment. Annals of Mathematical Statistics 27:986– 1005. [JDN, aMO] (2006) Understanding uncertainty. Wiley. [NP] Lipsey, R. G. & Lancaster, K. (1956) The general theory of second best. Review of Economic Studies 24:11 – 32. [HB] Liu, I.-M. (2003) Conditional reasoning and conditionalization. Journal of Experimental Psychology: Learning, Memory, and Cognition 29:694 – 709. [I-mL, rMO] Liu, I.-M. & Chou, T. H. (2008) A developmental analysis of conditional reasoning. Unpublished manuscript, National Chung-Cheng University, Chia-Yi, Taiwan. [I-mL] Liu, I. M., Lo, K. C. & Wu, J. T. (1996) A probabilistic interpretation of “If-Then”. The Quarterly Journal of Experimental Psychology 49A:828 – 44. [I-mL, rMO] Lukasiewicz, T. (2005) Weak nonmonotonic probabilistic logics. Artificial Intelligence 168:119– 61. [NP] Ma, W. J., Beck, J., Latham, P. & Pouget, A. (2006) Bayesian inference with probabilistic population codes. Nature Neuroscience 9:1432 – 38. [rMO] MacKay, D. J. C. (2003) Information theory, inference, and learning algorithms. Cambridge University Press. [rMO] Maher, P. (1992) Diachronic rationality. Philosophy of Science 59:120 – 41. [DD] Manktelow, K. I. & Over, D. E. (1987) Reasoning and rationality. Mind and Language 2:199 – 219. [aMO] (1991) Social roles and utilities in reasoning with deontic conditionals. Cognition 39:85 –105. [aMO] Manktelow, K. I., Sutherland, E. J. & Over, D. E. (1995) Probabilistic factors in deontic reasoning. Thinking and Reasoning 1:201– 20. [aMO] Manning, C. & Schu¨tze, H. (1999) Foundations of statistical natural language processing. MIT Press. [rMO] Markovits, H. & Barrouillet, P. (2002) The development of conditional reasoning: A mental model account. Developmental Review 22:5 – 36. [KO, NS]

118

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

Marr, D. (1982) Vision: A computational investigation into the human representation and processing of visual information. Freeman. [ELK, CRMM, JDN, rMO] Martignon, L. & Blackmond-Laskey, K. (1999) Bayesian benchmarks for fast and frugal heuristics. In: Simple heuristics that make us smart, ed. G. Gigerenzer, P. M. Todd & the ABC Research Group, pp. 169 – 88. Oxford University Press. [rMO] McCarthy, J. & Hayes, P. J. (1969) Some philosophical problems from the standpoint of artificial intelligence. In: Machine intelligence, vol. 4, ed. B. Meltzer & D. Michie. Edinburgh University Press. [aMO] McClelland, J. L. (1998) Connectionist models and Bayesian inference. In: Rational models of cognition, ed. M. Oaksford & N. Chater, pp. 21– 53. Oxford University Press. [aMO] McKenzie, C. R. M. (2003) Rational models as theories – not standards – of behavior. Trends in Cognitive Sciences 7(9):403 – 406. [JDN] (2004) Framing effects in inference tasks – and why they are normatively defensible. Memory and Cognition 32:874 – 85. [CRMM] McKenzie, C. R. M. & Amin, M. B. (2002) When wrong predictions provide more support than right ones. Psychonomic Bulletin and Review 9:821– 28. [CRMM] McKenzie, C. R. M., Ferreira, V. S., Mikkelsen, L. A., McDermott, K. J. & Skrable, R. P. (2001) Do conditional statements target rare events? Organizational Behavior and Human Decision Processes 85:291 – 309. [CRMM, arMO] McKenzie, C. R. M. & Mikkelsen, L. A. (2000) The psychological side of Hempel’s paradox of confirmation. Psychonomic Bulletin and Review 7:360 – 66. [CRMM, arMO] (2007) A Bayesian view of covariation assessment. Cognitive Psychology 54:33– 61. [CRMM, arMO] McKenzie, C. R. M. & Nelson, J. D. (2003) What a speaker’s choice of frame reveals: Reference points, frame selection, and framing effects. Psychonomic Bulletin and Review 10:596 – 602. [CRMM] Milch, B., Marthi, B. & Russell, S. (2004) BLOG: Relational modeling with unknown objects. In: ICML 2004 workshop on statistical relational learning and its connections to other fields, Banff, Alberta, Canada, ed. T. Dietterich, L. Getoor & K. Murphy, pp. 67– 73. Available at: www.cs.umd.edu/projects/ srl2004/srl2004_complete.pdf [TLG] Milne, P. (1995) A Bayesian defence of Popperian science? Analysis 55:213– 15. [rMO] (1996) log[P(hjeb)/P(hjb)] is the one true measure of confirmation. Philosophy of Science 63:21 – 26. [rMO] Moore, G. E. (1903) Principia ethica. Cambridge University Press. [rMO] Murphree, W. A. (1991) Numerically exceptive logic: A reduction of the classical syllogism. Peter Lang. [NP] Navarro, D. J., Griffiths, T. L., Steyvers, M. & Lee, M. D. (2006) Modeling individual differences using Dirichlet processes. Journal of Mathematical Psychology 50:101 – 22. [rMO, NS] Nelson, J. D. (2005) Finding useful questions: On Bayesian diagnosticity, probability, impact, and information gain. Psychological Review 112(4):979 – 99. [JDN, arMO] (2008) Towards a rational theory of human information acquisition. In: The probabilistic mind: Prospects for rational models of cognition, ed. M. Oaksford & N. Chater, pp. 143 – 63. Oxford University Press. [JDN] Nelson, J. D., McKenzie, C. R. M., Cottrell, G. W. & Sejnowski, T. J. (submitted) Experience matters: Information acquisition optimizes probability gain. [JDN] Newell, A., Shaw, J. C. & Simon, H. A. (1958) Chess-playing programs and the problem of complexity. IBM Journal of Research and Development 2:320– 25. [aMO] Newell, A. & Simon, H. A. (1972) Human problem solving. Prentice-Hall. [aMO] Newstead, S. E., Handley, S. J. & Buck, E. (1999) Falsifying mental models: Testing the predictions of theories of syllogistic reasoning. Memory and Cognition 27:344 – 54. [aMO] Nickerson, R. S. (1996) Hempel’s paradox and Wason’s selection task: Logical and psychological puzzles of confirmation. Thinking and Reasoning 2:1 – 32. [JDN, aMO] Novick, L. R. & Cheng, P. W. (2004) Assessing interactive causal influence. Psychological Review 111:455– 85. [aMO] Oaksford, M. (2004a) Conditional inference and constraint satisfaction: Reconciling probabilistic and mental models approaches? Paper presented at the 5th International Conference on Thinking, July 22– 24, 2004. University of Leuven, Leuven, Belgium. [aMO] (2004b) Reasoning. In: Cognitive psychology, ed. N. Braisby & A. Gellatly, pp. 418 – 55. Oxford University Press. [aMO] Oaksford, M. & Brown, G. D. A., eds. (1994) Neurodynamics and psychology. Academic Press. [UH] Oaksford, M. & Chater, N. (1991) Against logicist cognitive science. Mind and Language 6:1– 38. [arMO]

References/Oaksford & Chater: Pre´cis of Bayesian Rationality (1994) A rational analysis of the selection task as optimal data selection. Psychological Review 101:608– 31. [JDN, arMO, UH] (1996) Rational explanation of the selection task. Psychological Review 103:381 – 91. [aMO] (1998a) Rationality in an uncertain world. Psychology Press. [aMO, WS] eds. (1998b) Rational models of cognition. Oxford University Press. [UH, arMO] (2001) The probabilistic approach to human reasoning. Trends in Cognitive Sciences 5:349– 57. [E-JW] (2002) Common sense reasoning, logic and human rationality. In: Common sense, reasoning and rationality, ed. R. Elio, pp. 174 –214. Oxford University Press. [rMO] (2003a) Conditional probability and the cognitive science of conditional reasoning. Mind and Language 18:359 – 79. [aMO, DPO] (2003b) Optimal data selection: Revision, review, and reevaluation. Psychonomic Bulletin and Review 10:289 – 318. [aMO, E-JW] (2007) Bayesian rationality: The probabilistic approach to human reasoning. Oxford University Press. [NA, HB, DD, WDN, JStBTE, TLG, UH, GSH, ELK, I-mL, CRMM, JDN, KO, arMO, DPO, DEO, NP, FHP, GP. KS, NS, WS, E-JW] (2008) Probability logic and the Modus Ponens – Modus Tollens asymmetry in conditional inference. In: The probabilistic mind: Prospects for Bayesian cognitive science, ed. N. Chater & M. Oaksford, pp. 97– 120. Oxford University Press. [arMO] (in press) Conditionals and constraint satisfaction: Reconciling mental models and probabilistic approaches? In: Cognition and conditionals: Probability and logic in human thought, ed. M. Oaksford & N. Chater. Oxford University Press. [rMO] Oaksford, M., Chater, N. & Grainger, B. (1999) Probabilistic effects in data selection. Thinking and Reasoning 5:193– 244. [arMO] Oaksford, M., Chater, N., Grainger, B. & Larkin, J. (1997) Optimal data selection in the reduced array selection task (RAST). Journal of Experimental Psychology: Learning, Memory and Cognition, 23:441 – 58. [aMO] Oaksford, M., Chater, N. & Larkin, J. (2000) Probabilities and polarity biases in conditional inference. Journal of Experimental Psychology: Learning, Memory and Cognition 26:883 – 89. [KO, arMO] Oaksford, M. & Hahn, U. (2007) Induction, deduction and argument strength in human reasoning and argumentation. In: Inductive reasoning, ed. A. Feeney & E. Heit. pp. 269 – 301. Cambridge University Press. [rMO] Oaksford, M. & Moussakowski, M. (2004) Negations and natural sampling in data selection: Ecological vs. heuristic explanations of matching bias. Memory and Cognition 32:570 –81. [arMO] Oaksford, M., Roberts, L. & Chater, N. (2002) Relative informativeness of quantifiers used in syllogistic reasoning. Memory and Cognition 30:138 –49. [aMO] Oaksford, M. & Stenning, K. (1992) Reasoning with conditionals containing negated constituents. Journal of Experimental Psychology: Learning, Memory and Cognition 18:835 – 54. [aMO] Oaksford, M. & Wakefield, M. (2003) Data selection and natural sampling: Probabilities do matter. Memory and Cognition 31:143 – 54. [arMO] Oberauer, K. (2006) Reasoning with conditionals: A test of formal models of four theories. Cognitive Psychology 53:238– 83. [UH, KO, rMO] Oberauer, K., Weidenfeld, A. & Fischer, K. (2007) What makes us believe a conditional? The roles of covariation and causality. Thinking and Reasoning 13:340 – 69. [KO] Oberauer, K., Weidenfeld, A. & Ho¨rnig, R. (2004) Logical reasoning and probabilities: A comprehensive test of Oaksford and Chater (2001) Psychonomic Bulletin and Review 11:521 – 27. [KO, arMO] Oberauer, K. & Wilhelm, O. (2003) The meaning(s) of conditionals: Conditional probabilities, mental models and personal utilities. Journal of Experimental Psychology: Learning, Memory, and Cognition 29:680 – 93. [I-mL, aMO] Oberauer, K., Wilhelm, O. & Diaz, R. R. (1999) Bayesian rationality for the Wason selection task? A test of optimal data selection theory. Thinking and Reasoning 5:115– 44. [aMO, E-JW] O’Brien, D. P. (1993) Mental logic and irrationality: We can put a man on the moon, so why can’t we solve those logical reasoning problems? In: Rationality: Psychological and philosophical perspectives, ed. K. I. Manktelow & D. E. Over. Routledge. [DPO] (2004) Mental-logic theory: What it proposes, and reasons to take this proposal seriously. In: The nature of reasoning, ed. J. Leighton & R. J. Sternberg, pp. 205 – 33. Cambridge University Press. [DPO] O’Brien, D. P. & Manfrinati, A. (in press) The mental logic theory of conditional propositions. In: The psychology of conditionals, ed. M. Oaksford & N. Chater. Oxford University Press. [DPO] Osherson, D. N., Stob, M. & Weinstein, S. (1988) Mechanical learners pay a price for Bayesianism. Journal of Symbolic Logic 53:1245 – 51. [DD]

Over, D. E. & Evans, J. St. B. T (1994) Hits and misses: Kirby on the selection task. Cognition 52:235– 43. [aMO] Over, D. E., Hadjichristidis, C., Evans, J. St. B. T., Handley, S. J. & Sloman, S. A. (2007) The probability of causal conditionals. Cognitive Psychology 54:62– 97. [I-mL, DEO, rMO] Over, D. E. & Jessop, A. (1998) Rational analysis of causal conditionals and the selection task. In: Rational models of cognition, ed. M. Oaksford & N. Chater, pp. 399 – 414. Oxford University Press. [aMO] Pearl, J. (1988) Probabilistic reasoning in intelligent systems. Morgan Kaufmann. [arMO] (2000) Causality: Models, reasoning and inference. Cambridge University Press. [arMO] Perham, N. & Oaksford, M. (2005) Deontic reasoning with emotional content: Evolutionary psychology or decision theory? Cognitive Science 29:681 – 718. [aMO] Peters, S. & Westersta˚hl, D. (2006) Quantifiers in language and logic. Oxford University Press. [NP] Peterson, P. L. (2000) Intermediate quantifiers. Logic, linguistics, and Aristotelian semantics. Ashgate. [NP] Pfeifer, N. & Kleiter, G. D. (2005a) Coherence and nonmonotonicity in human reasoning. Synthese 146(1 – 2):93 –109. [GP, NP] (2005b) Towards a mental probability logic. Psychologica Belgica 45(1):71 – 99. Updated version available at: http://www.users.sbg.ac.at/ ~pfeifern/problog.htm [NP] (2006) Inference in conditional probability logic. Kybernetika 42:391 – 404. [NP] (in press a) Framing human inference by coherence based probability logic. Journal of Applied Logics. [NP] (in press b) The conditional in mental probability logic. In: Cognition and conditionals: Probability and logic in human thoughts, ed. M. Oaksford & N. Chater. Oxford University Press. [NP] Piaget, J. (1957) Logic and psychology. Basic Books. [GSH] Piaget, J. & Inhelder, B. (1975) The origin of the idea of chance in children. W. W. Norton. [NS] Pitt, M. A., Myung, I. J. & Zhang, S. (2002) Toward a method of selecting among computational models of cognition. Psychological Review 109:472– 91. [HB] Poletiek, F. H. (1996) Paradoxes of falsification. Quarterly Journal of Experimental Psychology 49A:447 – 62. [FHP] (2001) Hypothesis-testing behavior. Psychology Press. [FHP] Poletiek, F. H. & Berndsen, M. (2000) Hypothesis testing as risk behavior with regard to beliefs. Journal of Behavioral Decision Making 13:107 – 23. [FHP] Politzer, G. & Braine, M. D. (1991) Responses to inconsistent premises cannot count as suppression of valid inferences. Cognition 38:103 – 08. [aMO] Popper, K. R. (1935/1959) The logic of scientific discovery. Basic Books. [aMO] (1963/1978) Conjectures and refutations, 4th edition. Routledge and Kegan Paul. [FHP] Putnam, H. (1974) The “corroboration” of theories. In: The philosophy of Karl Popper, vol. 2, ed. A. Schilpp. Open Court. [aMO] Pylyshyn, Z., ed. (1987) The robot’s dilemma: The frame problem in artificial intelligence. Ablex. [aMO] Quine, W. V. O. (1953) From a logical point of view. Harvard University Press. [aMO] Raby, C. R., Alexis, D. M., Dickinson, A. & Clayton, N. S. (2007) Planning for the future by Western scrub-jays. Nature 445:919– 21. [ELK] Ramsey, F. P. (1931/1990a) General propositions and causality. In: Philosophical papers, ed. D. H. Mellor, pp. 145 – 63. Cambridge University Press. [I-mL] (1931/1990b) The foundations of mathematics and other logical essays. Routledge and Kegan Paul. [aMO] Rao, R. P. N., Olshausen, B. A. & Lewicki, M. S., eds. (2002) Probabilistic models of the brain: Perception and neural function. MIT Press. [rMO] Reichenbach, H. (1938) Experience and prediction. University of Chicago Press. [TLG] Reiter, R. (1980) A logic for default reasoning, Artificial Intelligence 13:81 – 132. [aMO] Restall, G. (1996) Information flow and relevant logics. In: Logic, language and computation, ed. J. Seligman, & D. Westersta˚hl, pp. 463 – 78 CSLI Publications. [rMO] Rips, L. J. (1983) Cognitive processes in propositional reasoning. Psychological Review 90:38– 71. [aMO] (1994) The psychology of proof. MIT Press. [arMO] (2002) Reasoning imperialism. In: Common sense, reasoning, and rationality, ed. R. Elio, pp. 215 – 35. Oxford University Press. [WS] Rissanen, J. J. (1989) Stochastic complexity and statistical inquiry. World Scientific. [rMO] Rosch, E. (1975) Cognitive representation of semantic categories. Journal of experimental psychology: General 104:192 – 233. [aMO] Rosser, J. B. (1939) An informal exposition of proofs of Go¨del’s theorem and Church’s theorem. Journal of Symbolic Logic 4:53– 60. [DPO]

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

119

References/Oaksford & Chater: Pre´cis of Bayesian Rationality Sanborn, A. N., Griffiths, T. L. & Navarro, D. J. (2006) A more rational model of categorization. In: Proceedings of the 28th Annual Conference of the Cognitive Science Society. Erlbaum. [TLG] Savage, L. J. (1972) Foundations of statistics. Dover. [DD] Schroyens, W. (in press) Logic and/in psychology: The paradoxes of material implication and psychologism in the cognitive science of human reasoning. In: Cognition and conditionals: Probability and logic in human thoughts, eds. M. Oaksford & N. Chater. Oxford University Press. [WS] Schroyens, W. & Schaeken, W. (2003) A critique of Oaksford, Chater and Larkin’s (2000) conditional probability model of conditional reasoning. Journal of Experimental Psychology: Learning, Memory and Cognition 29:140 – 49. [UH, aMO] Schulte, O. (1999) The logic of reliable and efficient inquiry. Journal of Philosophical Logic 28:399 – 438. [DD] Shafer, G. (1976) A mathematical theory of evidence. Princeton University Press. [rMO, GP] Shannon, C. E. & Weaver, W. (1949) The mathematical theory of communication. University of Illinois Press. [rMO] Shepard, R. N. (1987) Towards a universal law of generalization for psychological science. Science 237:1317 – 23. [TLG] (1995) Mental universals: Toward a twenty-first century science of mind. In: The science of the mind: 2001 and beyond, ed. R. L. Solso & D. W. Massaro, pp. 50 – 62. Oxford University Press. [TLG] Sher, S. & McKenzie, C. R. M. (2006) Information leakage from logically equivalent frames. Cognition 101:467– 94. [CRMM, rMO] (2008) Framing effects and rationality. In: The probabilistic mind: Prospects for Bayesian cognitive science, ed. N. Chater & M. Oaksford, pp. 79– 96. Oxford University Press. [CRMM] Shi, L., Feldman, N. & Griffiths, T. L. (2008) Performing Bayesian inference with exemplar models. In: Proceedings of the Thirtieth Annual Conference of the Cognitive Science Society, ed. B. Love, K. McRae & V. Sloutsky, pp. 745 – 50. Cognitive Science Society. [TLG] Shultz, T. R. (2007) The Bayesian revolution approaches psychological development. Developmental Science 10:357 –64. [NS] Skov, R. B. & Sherman, S. J. (1986) Information-gathering processes: Diagnosticity, hypothesis-confirmatory strategies, and perceived hypothesis confirmation. Journal of Experimental Social Psychology 22:93– 121. [JDN] Sloman, S. A. (1996) The empirical case for two systems of reasoning. Psychological Bulletin 119:3– 22. [rMO] (2005) Causal models. Oxford University Press. [rMO] Slowiaczek, L. M., Klayman, J., Sherman, S. J. & Skov, R. B. (1992) Information selection and use in hypothesis testing: What is a good question, and what is a good answer? Memory and Cognition 20:392 –405. [JDN] Smolensky, P. (1990) Tensor product variable binding and the representation of symbolic structures in connectionist networks. Artificial Intelligence 46:159 – 216. [UH] Sobel, D. M., Tenenbaum, J. B. & Gopnik, A. (2004) Children’s causal inferences from indirect evidence: Backwards blocking and Bayesian reasoning in preschoolers. Cognitive Science 28:303 –33. [rMO, NS] Sobel, J. H. (2004) Probable modus ponens and modus tollens and updating on uncertain evidence. Unpublished manuscript, Department of Philosophy, University of Toronto, Scarborough. Available at: www.scar.toronto.ca/ ~sobel/Conf/Disconf.pdf. [rMO] Sober, E. (2002) Intelligent design and probability reasoning. International Journal for Philosophy of Religion 52:65– 80. [arMO] Stanovich, K. E. (1999) Who is rational? Studies of individual differences in reasoning. Elrbaum. [JStBTE] (2008) Individual differences in reasoning and the algorithmic/intentional level distinction in cognitive science. In: Reasoning: Studies of human inference and its foundations, ed. L. Rips & J. Adler, pp. 414 – 36. Cambridge University Press. [rMO] Stanovich, K. E. & West, R. F. (2000) Individual differences in reasoning: Implications for the rationality debate? Behavioral and Brain Sciences 23:645 – 65. [rMO] Stenning, K. & van Lambalgen, M. (2005) Semantic interpretation as reasoning in non-monotonic logic: The real meaning of the suppression task. Cognitive Science 29(6):919 – 60. [KS] (2008a) Human reasoning and cognitive science. MIT Press. [KS]

120

BEHAVIORAL AND BRAIN SCIENCES (2009) 32:1

(2008b) Logic in a noisy world. In: Cognition and conditionals: Probability and logic in human thought, eds. M. Oaksford & N. Chater. Oxford University Press. [KS] Sternberg, R. J. (1999) Intelligence as developing expertise. Contemporary Educational Psychology 24:359– 75. [NS] Stevenson, R. J. & Over, D. E. (1995) Deduction from uncertain premises. The Quarterly Journal of Experimental Psychology 48A:613 – 43. [DEO, rMO] (2001) Reasoning from uncertain premises. Effects of expertise and conversational context. Thinking and Reasoning 7:367– 90. [DEO] Stewart, N., Chater, N. & Brown, G. D. A. (2006) Decision by sampling. Cognitive Psychology 53:1 – 26. [arMO] Taplin, J. E. (1971) Reasoning with conditional sentences. Journal of Verbal Learning and Verbal Behavior 10:219 – 25. [I-mL] Taplin, J. E. & Staudenmayer, H. (1973) Interpretation of abstract conditional sentences in deductive reasoning. Journal of Verbal Learning and Verbal Behavior 12:530 –42. [I-mL] Teller, P. (1973) Conditionalization and observation. Synthese 26:218 – 38. [DD] Tenenbaum, J. B. (1999) A Bayesian framework for concept learning. Doctoral dissertation, Brain and Cognitive Sciences Department, MIT. [rMO] Tenenbaum, J. B. & Griffths, T. L. (2001) Structure learning in human causal induction. In: Advances in neural information processing systems, vol. 13, ed. T. K. Keen, T. G. Dietterich & V. Tresp, pp. 59 –65. MIT Press. [rMO] Tenenbaum, J. B., Griffiths, T. L. & Kemp, C. (2006) Theory-based Bayesian models of inductive learning and reasoning. Trends in Cognitive Science 10:309 – 18. [TLG] Tenenbaum, J. B., Kemp, C. & Shafto, P. (2007) Theory based Bayesian models of inductive reasoning. In: Inducive reasoning, ed. A. Feeney & E. Heit, pp. 167 – 204. Oxford University Press. [UH, rMO] Thaler, R. H. (2005) Advances in behavioral finance, Vol. II. Princeton University Press. [rMO] Todd, P. M. & Gigerenzer, G. (2003) Bounding rationality to the world. Journal of Economic Psychology 24:143 – 65. [HB] Toussaint, M., Harmeling, S. & Storkey, A. (2006) Probabilistic inference for solving (PO)MDPs. Technical Report EDI-INF-RR-0934, University of Edinburgh. [rMO] van Fraassen, B. C. (1980) The scientific image. Oxford University Press. [ELK] Verhaeghen, P. & Salthouse, T. A. (1997) Meta-analyses of age-cognition relations in adulthood: Estimates of linear and nonlinear age effects and structural models. Psychological Bulletin 122:231 –49. [NS] Vermeij, G. J. (2004) Nature: An economic history. Princeton University Press. [ELK] Viswanathan, G. M., Buldyrev, S. V., Havlin, S., da Luz, M. G., Raposo, E. P. & Stanley, H. E. (1999) Optimizing the success of random searches. Nature 401:911 –14. [JDN] Vita´nyi, P. M. B. & Li, M. (2000) Minimum description length induction, Bayesianism, and Kolmogorov complexity. IEEE Transactions on Information Theory IT-46446 – 64. [rMO] von Helmholtz, H.(1910/1925) Physiological optics. Volume III. The theory of the perception of vision. Dover. (Translated from 3rd German edition, 1910). [rMO] Wagner, C. G. (2004) Modus tollens probabilized. British Journal for Philosophy of Science 55:747 – 53. [aMO] Walton, D. N. (1989) Informal logic. Cambridge University Press. [rMO] Wang, Z.-J. (1999) A two-component model of conditional reasoning. Unpublished Master’s Thesis, National Chung-Cheng University, Chia-Yi, Taiwan. [I-mL] Wells, G. L. & Lindsay, R. C. L. (1980) On estimating the diagnosticity of eyewitness nonidentifications. Psychological Bulletin 88:776 – 84. [JDN] Williamson, J. & Gabbay, D., eds. (2003) Special issue on combining logic and probability. Journal of Applied Logic 1:135– 308. [rMO] Yama, H. (2001) Matching versus optimal data selection in the Wason selection task. Thinking and Reasoning 7:295 – 311. [aMO] Yuille, A. L. & Kersten, D. (2006) Vision as Bayesian Inference: Analysis by Synthesis? Trends in Cognitive Sciences 10:301 – 08. [aMO] Zadeh, L. A. (1975) Fuzzy logic and approximate reasoning. Synthese 30:407– 28. [rMO] Zeelenberg, M., Van Dijk, W. W., Manstead, A. S. R. & Van der Pligt, J. (2000) On bad decisions and disconfirmed expectancies: The psychology of regret and disappointment. Cognition and Emotion 14:521 – 41. [aMO]

Behavioral and Brain Sciences Instructions for Authors and Commentators http://www.editorialmanager.com/bbs/account/BBS_ifc.pdf Behavioral and Brain Sciences (BBS) is a unique scientific communication medium, providing the service of Open Peer Commentary for reports of significant current work in psychology, neuroscience, behavioral biology or cognitive science. If a manuscript is judged by BBS referees and editors to be appropriate for Commentary (see Criteria below), it is circulated electronically to a large number of commentators selected (with the aid of systematic bibliographic searches and e-mail Calls for Commentators) from the BBS Associateship and the worldwide biobehavioral science community, including individuals recommended by the author. If you are not a BBS Associate and wish to enquire about joining, please see the instructions for associate membership at http://www.editorialmanager.com/bbs/account/BBS_ifc.pdf Once the Commentary stage of the process has begun, the author can no longer alter the article, but can respond formally to all commentaries accepted for publication. The target article, commentaries, and authors' responses then co-appear in BBS. (Note: Continuing Commentary submissions are no longer being accepted.) Criteria for acceptance: To be eligible for publication, a paper should not only meet the standards of a journal such as Psychological Review or the International Review of Neurobiology in terms of conceptual rigor, empirical grounding, and clarity of style, but the author should also offer an explicit 500 word rationale for soliciting Commentary, and a list of suggested commentators (complete with e-mail addresses). A BBS target article an be: (i) the report and discussion of empirical research that the author judges to have broader scope and implications than might be more appropriately reported in a specialty journal; (ii) an unusually significant theoretical article that formally models or systematizes a body of research; or (iii) a novel interpretation, synthesis, or critique of existing experimental or theoretical work. Occasionally, articles dealing with social or philosophical aspects of the behavioral and brain sciences will be considered. The service of Open Peer Commentary will be primarily devoted to original unpublished manuscripts written specifically for BBS treatment. However, a recently published book whose contents meet the standards outlined above spontaneously and multiply nominated by the BBS Associateship may also be eligible for Commentary. In such a BBS Multiple Book Review, a comprehensive, article-length précis by the author is published together with the commentaries and the author's response. In special cases, Commentary will also be extended to a position paper or an already published article that deals with particularly influential or controversial research or that has itself proven to be especially important or controversial. In normal cases however, BBS submissions may not be already published (either in part or whole) or be under consideration for publication elsewhere and submission of an article is considered expressly to imply this. Multiple book reviews and previously published articles appear by invitation only. Self-nominations cannot be considered, neither can non-spontaneous (i.e. author elicited) nominations. However, the BBS Associateship and professional readership of BBS are encouraged to nominate current topics, books and authors for Commentary; e-mail [email protected] In all the categories described, the decisive consideration for eligibility will be the desirability of Commentary for the submitted material. Controversiality simpliciter is not a sufficient criterion for soliciting Commentary: a paper may be controversial simply because it is wrong or weak. Nor is the mere presence of interdisciplinary aspects sufficient: general cybernetic and "organismic" disquisitions are not appropriate for BBS. Some appropriate rationales for seeking Open Peer Commentary would be that: (1) the material bears in a significant way on some current controversial issues in behavioral and brain sciences; (2) its findings substantively contradict some well-established aspects of current research and theory; (3) it criticizes the findings, practices, or principles of an accepted or influential line of work; (4) it unifies a substantial amount of disparate research; (5) it has important cross-disciplinary ramifications; (6) it introduces an innovative methodology or formalism for broader consideration; (7) it meaningfully integrates a body of brain and behavioral data; (8) it places a hitherto dissociated area of research into an evolutionary or ecological perspective; etc. In order to assure communication with potential commentators (and readers) from other BBS specialty areas, all technical terminology must be clearly defined or simplified, and specialized concepts must be fully described. In case of doubt of appropriateness for BBS Commentary, authors should submit a detailed target article proposal using the new BBS Editorial Manager site at http://www.editorialmanager.com/bbs/. After evaluating the proposal, the Editors will encourage or discourage formal target article submission. A note on commentaries: The purpose of the Open Peer Commentary service is to provide a concentrated constructive interaction between author and commentators on a topic judged to be of broad significance to the biobehavioral science community. Commentators should provide substantive criticism, interpretation, and elaboration as well as any pertinent complementary or supplementary material, such as illustrations; all original data will be refereed in order to assure the archival validity of BBS commentaries. Commentaries and articles should be free of hyperbole and remarks ad hominem. Please refer to and follow exactly the BBS Instructions for Commentators at http://www.editorialmanager.com/bbs/account/BBS_ifc.pdf before submitting your invited commentary. Style and format for target articles: Target Articles must not exceed 14,000 words (and should ordinarily be considerably shorter); commentaries should not exceed

1,000 words, excluding references. Spelling, capitalization, and punctuation should be consistent within each article and commentary and should follow the style recommended in the latest edition of A Manual of Style, The University of Chicago Press. It is advisable to examine a recent issue of BBS as a model. Target articles should be submitted in MSWord format to the new Editorial Manager site at http://www.editorialmanager.com/bbs/. Figures should appear in the body of the text, not at the end of the paper, and should also be supplied as separate TIFF, EPS, JPEG, or GIF files. However, if your article is accepted, TIFF or EPS format will be requested for publication since printing requires resolutions of at least 1100dpi. (Please note that costs for color figure reproduction will be passed along to the author. Color printing is expensive, and authors are encouraged to find alternative methods for presentation of their argument.) Once accepted, a Call for Commentators will be sent to thousands of BBS Associates and readers. The Call letter includes a link to the pre-copyedited final draft archived publicly for potential commentators. The copyedited final draft will only be posted for the invited commentators. Please make sure your target article file has ALL of the following in this order: Four Separate Word Counts (for the abstract, main text, references, and entire text – total + addresses etc.), an Indexable Title, Full Name(s), Institutional Address(es), E-mail Address(es) and Homepage URL(s) for all authors (where available), Short Abstract (100 words), Long Abstract (250 words), 5–10 Keywords (in alphabetical order), approx. 12,000 word Main Text (with paragraphs separated by full blank lines, not tab indents), and Alphabetical Reference List. Target article authors must also provide numbered headings and subheadings to facilitate cross-reference by commentators. Tables and figures (i.e., photographs, graphs, charts, or other artwork) should be numbered consecutively, and should appear in its appropriate location. Every table should have a title; every figure, a caption. Endnotes and appendices should be grouped together at the end of the paper and should ideally be locally linked to in the text to facilitate the reader (and of course the referee’s task). Acknowledgements should be placed at the end of the paper. The short abstract will appear by way of an advertisement, one issue in advance of the publication issue. The long abstract will be circulated to referees and then potential commentators should the paper be accepted, and will appear with the printed article. BBS’s rigorous timetable constraints (requiring the coordination of target articles, commentaries and author’s responses within the publishing queue) make it extremely difficult for us to process follow-up drafts of your submission. Please make sure that the paper you submit is the carefully checked final draft to which you wish the referees to address. Please also ensure that your submission has been proof-read by a native English speaker before submission. This, of course, greatly improves its chances at the refereeing stage. References: Bibliographic citations in the text must include the author’s last name and the date of publication and may include page references. Complete bibliographic information for each citation should be included in the list of references. Please also include and link to the WWW URL for any paper for which it exists. Examples of correct styles are: Brown (1973); (Brown 1973); Brown 1973; 1978); (Brown 1973; Jones 1976); (Brown & Jones 1978); (Brown et al. 1978). References should be in alphabetical order in the style of the following examples. Do not abbreviate journal titles: Freeman, W. J. (1958) Distribution in time and space of prepyriform electrical activity. Journal of Neurophysiology 2:644–66. http://cogprints.soton.ac.uk/abs/ neuro/199806009 Dennet, D. C. (1991) Two contrasts: Folk craft versus folk science and belief versus opinion. In: The future of folk psychology: Intentionality and cognitive science, ed. J. D. Greenwood, pp. 26–7. Cambridge University Press. http:// cogprints.soton.ac.uk/abs/phil/199804005 Bateson, P.P.G. & Hinde, R.A., eds. (1978) Growing points in ethology. Cambridge University Press. Editing: The publishers reserve the right to edit and proof all articles and commentaries accepted for publication. Authors of target articles will be given the opportunity o review the copy-edited manuscript and page proofs. Commentators will be asked to review copy-editing only when changes have been substantial; commentators will not see proofs. Both authors and commentators should notify the editorial office of all corrections within 48 hours or approval will be assumed. Author response to commentaries: All invited commentaries received before the deadline are only accessible to the Authors and Editors. Please note that no commentary is officially accepted until the Editor in charge has formally reviewed it and notified both the authors and the Editorial Administrator. Please refer to and follow exactly the BBS Commentary Response Instructions at http://www.editorialmanager.com/ bbs/account/BBS_ifc.pdf before submitting your response. Authors of target articles receive 50 offprints of the entire treatment, and can purchase additional copies. Commentators will also be given an opportunity to purchase offprints of the entire treatment.

In this issue Offprints of the following forthcoming BBS treatments can be purchased for educational purposes if they are ordered well in advance. For ordering information, please write to Journals Department, Cambridge University Press, 32 Avenue of the Americas, New York, NY 10013-2473.

Sex, attachment, and the development of reproductive strategies Marco Del Giudice

Précis of Bayesian Rationality: The Probabilistic Approach to Human Reasoning Mike Oaksford and Nick Chater

To appear in Volume 32, Number 2 (2009)

The propositional nature of human associative learning Chris J. Mitchell, University of New South Wales, Jan De Houwer, Ghent University & Peter F. Lovibond, University of New South Wales Many learning theorists believe that associative learning results from the operation of a mechanism in which links are formed automatically between mental representations. We show that the evidence supports an alternative view: that associative learning is the consequence of propositional reasoning processes that cooperate with the low-level unconscious processes involved in memory retrieval and perception. We argue that this new conceptual framework allows many of the important, recent advances in associative learning theory to be retained but recast in a model that provides a firmer foundation for both immediate application and future research.

With commentary from F Baeyens, D Vansteenwegen & D Hermans; AG Baker, I Baetu & RA Murphy; T Beckers & B Vervliet; E Bliss-Moreau & L Feldman Barrett; RA Boakes; L Castro & EA Wasserman; N Chater; ME Dawson & AM Schell; DM Dwyer, M Le Pelley, DN George, M Haselgrove & RC Honey; B Gawronski & GV Bodenhausen; A Gopnik; JD Greenwood; G Hall; DA Lagnado; N Li; EJ Livesey & JA Harris; H Lyn & DM Rumbaugh; G Mandler; H Matute & MA Vadillo; IPL McLaren; JD Miles, RW Proctor & EJ Capaldi; E Morsella, TA Riddle & JA Bargh; BR Newell; R Nolan; A Olsson & A Öhman; DC Penn, PW Cheng, KJ Holyoak, JE Hummel & DJ Povinelli; NA Schmajuk & GM Kutlu; H Schultheis & H Lachnit; DR Shanks; DA Sternberg & JL McClelland; JS Uleman; JE Witnauer, GP Urcelay & RR Miller

How we know our own minds: The relationship between mindreading and metacognition Peter Carruthers, University of Maryland A number of accounts of the relationship between third-person mindreading and first-person metacognition are compared and evaluated. Three of these accounts endorse the existence of introspection for propositional attitudes, but the fourth (defended here) claims that our knowledge of our own attitudes results from turning our mindreading capacities upon ourselves. The different types of theory are developed and evaluated, and multiple lines of evidence are reviewed, including evolutionary and comparative data, evidence of confabulation when self-attributing attitudes, phenomenological evidence of “unsymbolized thinking,” data from schizophrenia, and data from autism.

With commentary from ML Anderson & D Perlis; BJ Baars; C Buckner, A Shriver, S Crowley & C Allen; AC Catania; JJ Couchman, MVC Coutinho, MJ Beran & JD Smith; JStBT Evans; C Fernyhough; B Fiala & S Nichols; K Frankish; O Friedman & AR Petrashek; M Hernik, P Fearon & P Fonagy; B Huebner & DC Dennett; RT Hurlburt; N Kornell, BL Schwartz & LK Son; P Langland-Hassan; C Lewis & JIM Carpendale; MV Lombardo, B Chakrabarti & S Baron-Cohen; RW Lurz; CM Mills & JH Danovitch; D Murphy; D Pereplyotchik; RE Petty & P Briñol; J Proust; P Robbins; P Rochat; B Wiffen & A David; D Williams, SE Lind & F Happé; A Zinck, S Lodahl & CD Frith

Among the articles to appear in forthcoming issues of BBS: C. J. Mitchell, J. De Houwer & P. F. Lovibond, “The propositional nature of human associative learning” P. Carruthers, “How we know our own minds: The relationship between mindreading and metacognition” J. Archer, “Does sexual selection explain human sex differences in aggression?” J. Vigil, “A sociorelational framework of sex differences in the expression of emotion”

Cambridge Journals Online For further information about this journal please go to the journal website at: journals.cambridge.org/bbs

http://www.bbsonline.org/Preprints/Mitchell-08092007/Referees http://www.bbsonline.org/Preprints/Carruthers-03112008/Referees http://www.bbsonline.org/Preprints/Archer-05142007/Referees http://www.bbsonline.org/Preprints/Vigil-02212008/Referees