This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
is true if and only if P. However, Mellor insists that the equivalence principle by itself cannot tell us what a proposition’s truthmaker is. Consider the propositions <Murder is wrong> and (alternatively, the Ramsey sentence is true of one and only one item, ).20 In the historical case being considered, the open sentence associated with the defi nite description (¶k)[Pk] has one and only one realizer, viz., the worldly entity <electron>. If no item realizes the open sentence then no denotation has been fi xed for the sole theoretical term. For example, in the Ramsey sentence version of the phlogiston theory, or the electromagnetic ether theory, the respective terms “phlogiston” or “electromagnetic ether” have no denotation fi xed for them. What can be said about electron theories? It will be argued that despite the fact that there is a definite description picking out the electron, if we were to use the full electron theories that many have proposed to provide a Ramsey–Lewis denotation fi xer, as in expressions akin to (4), then . Hence the use in the text of the corner brackets in the simple case of a 1-tuple.
14
Heather Dyke
Questions about truth, questions about reality, and questions that link the two domains are among the oldest and most venerable philosophical questions and yet remain some of the most vexed and perplexing of them all. A dominant theme of the chapters in this volume is that things can go badly wrong if we confuse a question about truth with a question about reality, or vice versa, and that this is surprisingly easy to do. Despite their unifying subject matter, the chapters in this volume tackle a broad range of issues concerning truth and reality and offer some new approaches to, and some new perspectives on, some of those old and venerable questions.
Part I
Truth
2
“Unsettledness” in a Bivalent Language A Modest, Nonepistemic Idea JC Beall
1
INTRODUCTION
In this chapter, I suggest a novel, nonepistemic approach towards accommodating the appearance of “unsettledness” or—in some sense—gaps in a bivalent language. Although the suggestion is modest, I think it to be quite plausible. My aim is simply to lay out the proposal, as concisely as possible, and briefly consider a few objections. Accordingly, §2 sets out some terminology. (I should warn that §2 simply marches through some definitions without pause. Readers might prefer to fi rst read §3 and, in turn, return to §2.) §3 sketches the main issue, namely, accommodating the appearance of (some sort of) “gaps” or “unsettledness” in our otherwise bivalent language. §4 and §5 present the main proposal. §6 briefly considers a few objections, and §7 offers a brief review and closing remarks.
2
TERMINOLOGY
The following defi nitions will be used throughout. Not all of the defi nitions are entirely standard or, perhaps, even ideal; however, they will serve to focus the issue. Throughout, we let D be our given (nonempty) domain, α and β any sentences—where sentences, throughout, are meaningful, declarative sentences—and φ(x) any unary predicate. (For simplicity, I put things in terms of unary predicates. The generalization to n-ary predicates should be straightforward but, for present purposes, needlessly multiplies symbols.) We let φ(x)+ and φ(x)- be the extension and antiextension of φ(x), respectively. (We also make the convenient assumption that y ∈ φ(x)- if y is not a sentence. This is not ultimately necessary, though dropping the assumption would require talk of “range of application” and so on, which is ultimately important but, I think, may be set aside here without loss.) Throughout, falsity is taken to be truth of negation: α is false just if ¬α is true. Finally, for convenience, I rely on context to clarify use-mention.
18 JC Beall Definition (Excluded Middle, LEM) α ∨ ¬α is logically true for all α. Definition (Bivalence, BIV) Tr(┌α┐) ∨ Tr(┌¬α┐) is logically true for all α. Definition (Bivalent Language) Language L is bivalent just if φ(x)+ ∪ φ(x)- = D for all objects x ∈ D and all L predicates φ(x). Definition (Standardly Gappy Language) L is standardly gappy just if φ(x)+ ∪ φ(x)- ≠ D for some x ∈ D and L predicate φ(x). Definition (Contraries) α and β are contraries just if α ∧ β is logically false. Definition (Subcontraries) α and β are subcontraries just if α ∨ β is logically true. Definition (Contradictories) α and β are contradictories just if they are both contraries and subcontraries. I should note that the point of setting out these defi nitions is not to foreshadow any sort of proof. Rather, setting out such terminology facilitates a concise presentation of the issue and, in turn, the chief proposal. I turn now to the issue at hand.
3 THE ISSUE: BIVALENCE AND APPARENT UNSETTLEDNESS A challenge that confronts those, like myself, who endorse bivalence, is the apparent “unsettledness” of our language. In a nutshell, there appear to be “gaps,” in some sense. (In what sense there are “gaps” is part of the issue. I will return to this.) Vagueness is one of the driving phenomena behind the appearance of such “unsettledness” or “gaps.” Let Rφ be a “tolerance relation” for (vague) predicate φ(x), so that Rφ x, y only if x and y are relevantly similar. Bivalence—plus plausible other logical assumptions—has the (prima facie untoward) consequence of “sharp cutoffs” for φ(x), namely, SC. ∃x∃y(R φ(x, y) ∧ φ(x) ∧¬φ(y)). But philosophical intuition rails against such cutoffs. That there’s a moment t such that at t you’re a person but, for i as small as you like, at t-i you’re not a person strikes many philosophers as absurd or, at least, too hard to believe. Our usage, one is inclined to think, is simply not that sharp. The swift step, in turn, is to mollify “intuition” by acknowledging “gaps” in some sense. Indeed, be it the prima facie repugnance of SC or background
“Unsettledness” in a Bivalent Language
19
thoughts about meaning and use (and truth), many philosophers are inclined to accept the following, for some b. G. φ(x) is neither true nor false of b. Of course, on standard nonclassical schemes (e.g., Kleene), G itself is puzzling, at least if, as we’re assuming, falsity is truth of negation. Still, be it SC or something else, the appearance of some sort of “gappiness” in our language is a strong one. As such, those who endorse bivalence face the challenge of either accommodating some sort of “gappiness,” some sort of “unsettledness,” or explaining it away. This is the task to which my proposal is aimed. If, as I think, the logic of our language is fairly normal, SC is inevitable, given the “exhaustive,” LEM-satisfying “nature” of negation. The challenge, as above, is to make sense of the strong appearance of “unsettledness” or “gaps,” in some (clearly nonstandard) sense of “gaps.” «Parenthetical remark. In saying that SC is inevitable given the exhaustive, LEM-satisfying “nature” of negation, I am sneaking in a background assumption about truth, which I fully endorse but do not discuss in this chapter. In particular, I am assuming that truth is entirely transparent in the sense that Tr(┌α┐) and α are intersubstitutable in all (nonopaque) contexts, for all α in the language. Though van Fraassen (1966) showed that languages can enjoy LEM without thereby being bivalent, the former implies the latter where transparent truth is involved—at least as far as what can be (truly) said in the given language. Suppose, for example, that we have the validity of α ∨ ¬α. Assuming, as I do, that negation is a nonopaque context, the transparency of truth yields the equivalence of ¬α and Tr(┌ ¬α┐). Hence, α ∨ ¬α is equivalent to Tr(┌α┐) ∨ Tr(┌ ¬α┐), which is just to say that Excluded Middle implies bivalence. This assumption may be dropped without loss, although my remarks about “exhaustive” negation will have to be read so as to ensure not only bivalence, as a true principle expressible in the language but a bivalent language too. End parenthetical.» Of course, there are already a few well-known proposals that purport to accommodate apparent unsettledness in a bivalent language. In particular, classical epistemicist proposals, such as Sorensen’s (2001) and Williamson’s (1994), fi rmly accept—as I do—that our language is bivalent but nonetheless enjoys “gaps.” The “gaps,” on such accounts, are epistemic gaps, gaps in our knowledge, as opposed to “gaps” between truth and falsity (whatever, in the end, that might come to). On these approaches, SC is an inevitable consequence of the exhaustive nature of negation and the logical behavior of our other connectives. On this I agree.1 How, then, does our
1
This is not to say that I endorse the particular logic in question. Both Sorensen and Williamson endorse classical (or, at least, an “explosive”) logic, whereas I endorse a paraconsistent logic to accommodate the transparency of truth (given
20 JC Beall language happen to be “unsettled”? As above, the epistemicist proposal is that our “gaps” are not in the language; they are in our knowledge. In short, we simply do not know the particular point at which we become a person, even though there is one. Moreover, we cannot—we could not—know the given point. The “gaps” in question are not just actual; they’re necessary. My aim is not to evaluate epistemicist proposals or even discuss them in any serious fashion. I mention epistemicism only as one—perhaps the leading—approach to accommodating some sort of “gaps” in a bivalent language.2 I do not have a knockdown argument against epistemicism; I have only the familiar autobiographical reason for rejection, namely, that I fi nd it difficult to believe. Although I accept, with epistemicists, that there’s a point at which I’m a person and before which not, and also accept that we’re actually ignorant of the particular point, I fi nd it difficult to believe that we couldn’t know the given point, even with all actual information (including the sufficient condition for being a person). Autobiography is not argument, but I am motivated to fi nd an alternative course.
4
STARS: AN ABSTRACT OPTION
My aim is to suggest a nonepistemicist account of “gaps” in a bivalent language. I accept, as above, that SC is inevitable—at least given the sort of logic in the background. I also accept that the appearance of “unsettledness” is strong, that the appearance of “gaps,” in some sense, requires explanation. My aim is to offer a modest proposal towards accommodating such appearances, despite bivalence. Given the defi nitions in §2, it is obvious that our language is bivalent only if it is not standardly gappy, in the sense of §2. The task is not to fi nd a way of explaining how gaps, on the standard defi nition, might arise; we are fully embracing bivalence—fully embracing that our language is not standardly gappy.3 The task, rather, is to consider another way in which our language might enjoy “unsettledness” or “gaps,” in some other, suitably related sense of such terms. The terminology, in the end, is neither here nor there. The task, in the end, is to explore a possible phenomenon that,
bivalence). This difference is important in the broader scheme of issues, but it’s not particularly germane here, and so I leave the matter there. 2 Another approach is that of Weatherson (2005). Though I think that this is an interesting approach, as are the epistemicist approaches, I do not discuss it here, chiefly for space reasons. 3 As such, our task is different from that of Soames (1999), wherein “smidget” is introduced to illustrate how gaps, in the standard sense, might arise. That said, I embrace Soames’ basic suggestion (though not his overall account). I return to this in §5.
“Unsettledness” in a Bivalent Language
21
if actual, might well underwrite the appearance of unsettledness, or indeed be the phenomenon of “vagueness.” Before turning to the particular proposal, it’ll be useful to consider the matter in abstract. In short, suppose that our language L is bivalent in the sense in §2. Then there is no b ∈ D such that φ(b)+ ∪ φ(b)- ≠ D. The question at hand, as above, is whether there’s an alternative sense in which L might be unsettled, a sense that isn’t too far from the standard sense. I suggest that there’s a natural, modest sense in which L might nonetheless be “unsettled.” To simplify matters, suppose that L is devoid of “semantic” terms like “true” and the like.4 Suppose that, for every “positive atomic” predicate φ(x) in L, there is a “star mate” φ*(x), which—for simplicity—is also an atomic predicate, though not a “positive” one.5 Now, suppose that the relation between our positive atomics and their star mates is exclusive but not necessarily exhaustive, in the sense that, for all x, EXC. φ(x)+ ∩ φ*(x)+ = ∅ but we need not have EXH. φ(x)+ ∪ φ*(x)+ = D. In other words, we never have some b of which both φ(x) and φ*(x) are true, but we might have some b of which neither φ(x) nor φ*(x) is true. Notice that, given familiar assumptions about negation and validity, EXC will yield φ*(x) ⊢ ¬φ(x) but we will not get the converse, since φ*(x) is simply an atomic governed only by the minimal constraints above—in effect, only EXC. The idea, in abstract, may be put succinctly: for every “positive atomic” φ(x) in the language, there is an atomic contrary. « Parenthetical remark. I should note that it’s not obvious that one must treat the given “star mates” (or, for short, “stars”) as atomic predicates, but I do so for simplicity. If one were to treat * as some sort of connective, the story would be rather complicated. One route might be along da Costa’s “negation” lines, wherein * is interpreted in a noncompositional fashion (da Costa and
4
This isn’t ultimately necessary, but I do not go into the matter here. I leave the notion of a “positive predicate” at an intuitive level. For present purposes, predicates like “is a book,” “is bald,” “is sad,” and so on are one and all positive atomics. 5
22
JC Beall
Alves 1977). Ultimately, though, the end result will have to be such that φ*(x) is, in effect, atomic, and so I simply treat it as such.) End parenthetical.» Of course, given that negation itself is a contradictory-forming operator, in the sense that ¬α is the contradictory of α for all α (see §2), we have φ*(x) ∨ ¬φ*(x) for all given “stars.” But this is simply the result of negation doing its “exhaustive” job. As for our “stars” themselves, while we have it that φ(x) ∧ φ*(x) is logically false for all x and all φ(x), we—importantly—do not, as mentioned above, have that, for all x, the following holds: φ(x) ∨ φ*(x). In other words, given standard assumptions of validity, we will have ⊢ ¬(φ(x) ∧ φ*(x)) for any x, but, notably, ⊬ (φ(x) ∨ φ*(x)). This latter feature, I suggest, is at least one natural sort of “unsettledness” or “gaps,” however modest it may be. The question, of course, is how, if at all, it applies to our actual language.
5
PROPOSAL: “GAPS” VIA STAR MATES
The proposal is that our own language has such “stars.” Although our language is bivalent, there is also unsettledness in a nonepistemic sense; there are “gaps” in the sense that, for some b and vague φ(x), we have the following: ¬φ(b) ∧ ¬φ*(b). In other words, b falls into the “gap” between φ(x) and φ*(x), though—given bivalence—obviously not the (nonexistent) gap between φ(x) and ¬φ(x). We do not, of course, have predicates in English that are spelled with “stars.” What, then, plays the role of “star mates” in our language? My suggestion is that something along the lines of “non-φ(x)” is what does the “star” work in our (natural) language, though it may well be some deviant
“Unsettledness” in a Bivalent Language
23
usage of “not” that results in an atomic “not-φ(x).” For present purposes, I’ll assume the former (“non”) construction. In particular, for each positive atomic predicate φ(x), at least of our base “semantic-free” fragment, we have a star mate expressed by “non-φ(x).” The star mate, as in the abstract sketch above, is an atomic predicate, however much it may superficially look like a molecular predicate. « Parenthetical remark. I should emphasize, again, that this assumption is probably not necessary. If one took “non” to be an operator, it would have to be something other than negation, and, as above, likely have to be noncompositional. As such, it strikes me as simpler to just assume that such “star mates” are atomic predicates in their own right. End parenthetical. » The proposal, as in §4, has it that our star mates, expressed via some atomic “non-φ(x)” construction, are contraries of φ(x), with EXC being their fundamental constraint. The pressing question concerns the satisfaction conditions for such stars. What are they? There might be different, equally viable approaches to this question. For present purposes, I suggest a simple and familiar idea about vague predicates. In particular, setting our “star mates” aside for the moment, I adopt the idea, made explicit by Soames (1999), that “vague” or “unsettled” predicates are those for which our practice has established only a sufficient condition for satisfaction.6 The idea, in short, is that our (vague) positive atomics φ(x) enjoy only some sufficient satisfaction condition ψ(x) so that we have only something of the following sort governing φ(x): ψ(x) → φ(x). Soames’ well-known “smidget” example is intended to show how a language might come to have “gappy” predicates, in the standard sense of “gaps” (see §2). Soames suggests that, in addition to φ(x) getting only a sufficient condition, so too for the molecular, negation-ful predicate ¬φ(x). In particular, Soames’ smidget example has us giving a sufficient condition for Smidget(x), namely,7 x ≤ 20 → Smidget(x) and also a sufficient condition for the molecular (negation-ful) predicate ¬Smidget(x), namely, x ≥ 25 → ¬Smidget(x).
6 NB: Soames’ (contextual) development of the idea is not something to which I’m subscribing but, rather, only his initial setup—or, at least, part of his setup. 7 This is not exactly Soames’ example, but the idea is the same.
24
JC Beall
This latter step is curious (at best) on the current proposal, wherein we’re assuming bivalence, and hence that negation itself is fully exhaustive (i.e., that LEM holds). As such, there’s no need to give a sufficient condition for ¬φ(x), as this is already covered by the exhaustive “nature” of negation. Let me emphasize that this is not an objection to Soames’ account or even his “smidget” example. His example is intended to illustrate how we might “enjoy” a standardly gappy language in the sense of §2. For present purposes, I’m assuming that we do not have standard gaps but still think that Soames’ illustration is relevant—not for negation but for our “stars.” The natural suggestion for our “star mates,” which are similarly “vague,” is the same: we have only some sufficient condition for φ*(x): µ(x) → φ*(x). In turn, so long as we do not have ψ(x) ∨ µ(x), we need not have φ(x) ∨ φ*(x). In cases—perhaps “precise” ones—in which we do have ψ(x) ∨ µ(x), then we’ll have φ(x) ∨ φ*(x). But, again, there’s no reason to think that we’d have such “star exhaustion” (i.e., “star excluded middle”) for all such predicates. The proposal, in short, is that we can understand the apparent “unsettledness” of our (bivalent) language along “star” lines. Given bivalence (and other logical features of the language), we do not have “gaps” in the standard sense. Still, our practice, governing “vague” predicates, leaves lots of “star gaps” (as it were); for many (vague) predicates φ(x), there are objects y such that neither φ(x) nor φ*(x) is true of y. This, I think, is a modest way in which “unsettledness” arises in our (bivalent) language.
6
OBJECTIONS AND REPLIES
In this section, I briefly consider a few objections to the proposal. For convenience, I put the objections as if directed against me, so that “you” denotes me. Objection. Your proposal still leaves SC intact and hence fails to remove the untoward consequence of sharp cutoffs. But it’s precisely such “sharp cutoffs” that many want to avoid in accepting that our language is “unsettled.” At the very least, it isn’t clear how you explain—or, perhaps, explain away—why so many fi nd SC to be repugnant. Reply. The answer, in the end, invokes confl ation. Of course we have SC, which, as said, is inevitable given our negation (and other features of our logic). Given bivalence, reflection makes it clear that—assuming other logical features—there’s always a “cutoff” when negation is around
“Unsettledness” in a Bivalent Language
25
to do the cutting; that’s just how negation is, and the job that negation does. Usage may have—and, I think, has—left many “gaps” for predicates, but the gaps are quickly closed by negation. The “gaps” of our (vague) predicates are not standard gaps (as in §2); rather, they are “star gaps,” as it were, that arise from the distinctive suffi cient-condition-only practice that governs vague terms. If we reflect on the role and standard-gap-closing features of negation (which I’m assuming), we needn’t scream in the face of SC. The proposal is that our screams against SC are actually screams against a different but perhaps superfi cially similar principle, namely, SC*. ∃x∃y(R φ(x, y) ∧ φ(x) ∧ φ*(y)). The unsettledness of our language essentially involves our star mates—our atomic “non-φ(x)” constructions—and, so, we would rightly scream if SC* were true. But SC* is not true. Objection. Your proposal has some similarities to the supervaluationist response.8 It too is committed to • It is true that there is a point at which φ(n) ∧ ¬φ(n + 1). But they too seek to avoid this seemingly untoward consequence by asserting “the confusion hypothesis.” They charge that the truth to which they are committed is confused with another, namely • There is a point for which it is true that φ(n) ∧ ¬φ(n + 1). Only the latter is taken to imply the existence of a “sharp boundary,” but they take themselves to only be committed to the supposedly distinct former claim. The question for you: why not, then, simply go with supervaluationism? Reply. I am sympathetic with supervaluationism and certainly have no knockdown argument against it. My chief reason for pursuing a different course is that, as mentioned, I’m committed to bivalence—not simply LEM—and the whole point of supervaluationism is to enjoy a lot of “classical” negation behavior while nonetheless rejecting bivalence.9 What fuels my commitment to bivalence is a background theory of truth
8 This objection is directly from Dominic Hyde (correspondence), and the reply is also indebted to Hyde. 9 Of course, one could have bivalence on a subvaluationist approach, with which I’m also sympathetic; however, this approach will similarly have a notion of truth with which I’m not sympathetic. (And see Hyde, 1997, and Varzi, 1999, for subvaluationism.)
26
JC Beall
and commitment to the “exhaustive,” LEM-satisfying behavior of negation. In particular, as in the parenthetical remark in §3, I think that truth is fundamentally a fully transparent device; it is a device Tr(x) such that Tr(x) and x are intersubstitutable in all (nonopaque) contexts, for all sentences x in the language. Given LEM and transparency of truth, we quickly get Tr(┌ α ┐) ∨ Tr(┌ ¬α ┐), which is the principle of bivalence (at least given, as I assume, that falsity is truth of negation). Other “transparent truth theorists,” such as Field (2008), go on to recognize stronger notions of truth in addition to our fundamental transparent one, but I do not, at least as yet, see the need to do so. The reason that such theorists pursue stronger notions of truth arises from a prior rejection of bivalence—for transparent truth theorists, a prior rejection of LEM. If there were no way to accommodate the appearance of “unsettledness” (in some sense of the term) given bivalence, then I would likely go the route of Field or, perhaps, something along the lines of McGee (1991). The current chapter, however, offers a proposal on how we might enjoy “unsettledness” (in a nonstandard sense) while also enjoying bivalence. Objection. An obvious problem is that you still get “sharp borders” with your “star mates” (i.e., the atomic contraries). In particular, any soritical series will be cut up into three disjoint regions: you’ll have the y that satisfy φ(x), the y that satisfy φ*(x), and the y that satisfy neither φ(x) nor φ*(x). But this is just another untoward consequence against which philosophical intuition screams, and so your account fails to resolve all of the relevant problems, notably, “higher-order vagueness.” Reply. I am not sure what to make of “higher-order vagueness,” and so give at most a tentative reply here. The chief reply is that such “sharp borders” are to be expected if, as they are, they’re characterized in terms of negation. In other words, given the “border-drawing” nature of negation, such “borders” are inevitable and, in the end, to be expected—at least if, again, the given result essentially involves negation (e.g., that we have three different regions, etc., where “different” is understood via negation, or more to the point the “neither . . . nor” clause.) On the other hand, there’s no reason to think that we’d have “sharp borders” where this is characterized in terms of our “stars.” Negation has the job of cutting borders; the stars—our atomic contrary predicates—leave matters looser. So long as we do not have “sharp borders” expressed via stars (as opposed to negation), then it seems we have no reason to scream—at least on reflection. Objection. Your proposal is viable only if we have such “star mates” (i.e., atomic contraries) in our language. But why would we have such things? Reply. This is a very important question for which I currently have at best a tentative answer. Pending a full and plausible answer, the proposal remains only an option worthy of further exploration. My current (and
“Unsettledness” in a Bivalent Language
27
only sketchy) answer is twofold. First, we clearly have atomic contraries in our language. Predicates like “good” and “bad” are examples. (If these are not good examples, use examples that are good!) Assuming, as I think plausible, that “bad” is not defi nable in terms of “good” and negation, there must be some reason that we introduced the (atomic) contrary “bad.” (If there isn’t such a reason, then the current objection strikes me as less pressing. I assume that there is such a reason.) Whatever the reason might be, it is not implausible to think that there’d be motivation to generalize to all (say, “positive”) predicates. It is often very easy, though perhaps not always productive, to let the motivation in a particular case serve as motivation for all relevant cases. The conjecture is that we simply specified (only sufficient conditions) for star mates—in particular, “non-φ(x)” (or some such deviant usage of “not” or “non”)—for all of our positive, vague atomics. Of course, the details of this story need to be told, and the account ultimately turns on empirical linguistics. For now, notwithstanding details, I fi nd the basic proposal to be plausible, or at least a viable suggestion.
7
CLOSING REMARKS
I have suppressed a great deal of background assumptions about truth and logic, or at least about my own views on these matters (Beall forthcoming), which details partly motivate the current proposal. Still, there is an independently recognizable project that I’ve tried to briefly sketch and modestly answer. In short, if our language is bivalent, how can we accommodate apparent unsettledness in the language? One answer—or class of answers—is epistemic. Other answers might posit special features of truth. I have shied away from these only given my own background views on truth not because I think that they’re not worthy of serious consideration. Although such answers are interesting and prima facie viable, I have pursued a different one. Specifi cally, the modest suggestion is that, as Soames (1999) notes, we can recognize a sort of unsettledness arising from the fact that vague predicates have only suffi cient conditions of satisfaction. Unlike Soames, we do not also need to say that such predicates have (only) a suffi cient condition for “anti-satisfaction” (as it were), that is, satisfaction of the molecular predicate ¬φ(x). Given bivalence, the “anti-satisfaction” condition of such predicates is already ensured by features of negation (and other background logical features, including contraposition). Where unsettledness arises is in our “star mates,” atomic contraries of our other vague atomic predicates. Although philosophical intuition would rightly scream at SC*, there’s no need to scream at SC, as that’s just negation—in concert with other logical features—doing its sharp-cutting job. Given that we don’t get SC*, there’s no need to scream at all. This chapter, of course, gives only
28 JC Beall a very brief sketch of the idea. There is much work to be done by way of details. Still, I hope to have at least pointed in the direction of a viable, nonepistemic approach to enjoying both unsettledness and bivalence.10
10. Thanks to Colin Caret, Aaron Cotnoir, Hartry Field, Dominic Hyde, Vann McGee, Graham Priest, Greg Restall, Gillian Russell, and Lionel Shapiro for discussion. Thanks, too, to Heather Dyke for her understanding, patience, and the work involved in putting together this volume.
3
Vagueness and Truth Mark Colyvan
1 INTRODUCING THE PRINCIPLE OF UNIFORM SOLUTION It would be very odd to give different responses to two paradoxes depending on minor, seemingly irrelevant details of their presentation. For example, it would be unacceptable to deal with the paradox of the heap by invoking a multivalued logic, Ł ∞ , say, and yet, when faced with the paradox of the bald man, invoke a supervaluational logic. Clearly these two paradoxes are of a kind—they are both instances of the sorites paradox. And whether the sorites paradox is couched in terms of heaps and grains of sand, or in terms of baldness and the number of hairs on the head, it is essentially the same problem and therefore must be solved by the same means. More generally, we might suggest that similar paradoxes should be resolved by similar means. This advice is sometimes elevated to the status of a principle, which usually goes by the name of the principle of uniform solution. This principle and its motivation will occupy us for much of the discussion in this chapter. In particular, I defend a rather general form of this principle. I argue that two paradoxes can be thought to be of the same kind because (at a suitable level of abstraction) they share a similar internal structure, or because of external considerations such as the relationships of the paradoxes in question to other paradoxes in the vicinity, or because of the way they respond to proposed solutions. I then use this reading of the principle of uniform solution to make a case for the sorites and the liar paradox being of a kind. First, as I’ve already indicated, it is uncontroversial that we all subscribe to some version of the principle of uniform solution.1 There may well be disagreement about the application, scope, and details of formulation of
1 It is also clear that this principle has broad applications, across all areas of philosophy where paradoxes arise: philosophy of logic, decision theory, ethics, philosophy of space and time, and so on. See Colyvan (2008) for a recent implicit appeal to the principle to a family of paradoxes in decision theory.
30
Mark Colyvan
the principle, but there is no doubt that some version is agreed to by everyone. Such agreement is rare in philosophy, so it is worth pausing over why there should be such agreement here. I take it that it is not merely a matter of simplicity. That is, I take it that the reason philosophers accept some version of the principle of uniform solution is not motivated by a concern to deal with as many paradoxes as possible with a minimum of resources. There are two reasons why this cannot be the motivation. The fi rst is that this motivation makes no reference to, nor does it require that, the two paradoxes in question be of a kind. Simplicity considerations, if anything, would motivate a principle akin to treating all paradoxes with a minimum number of tools—use a hammer for every task, irrespective of whether all the tasks involve nails. The second reason that this cannot be the motivation for the principle of uniform solution is that simplicity considerations notoriously divide philosophers, and it would be extraordinary if the agreement we have over the principle of uniform solution were due to something so controversial as simplicity. I take the motivation for the principle of uniform solution to be a concern for treating the root of the problem—the disease—not just the symptoms. At least, this is what motivates my subscription to the principle, and I suspect that something like this is the motivation for others as well. The idea is that if you do not bring similar solutions to bear on similar paradoxes, you will be merely hiding the problem—masking the symptoms—rather than getting to the heart of the matter and treating the underlying pathology. An example will help. Let’s suppose that you are tempted to treat the paradox of the bald man by stipulating that some specific number, n, hairs on the head is the sharp cut-off between bald and not bald. And further suppose that you are tempted to treat the paradox of the heap by denying that there are any heaps. Both paradoxes are solved. But the concern is that you have not fully appreciated the fact that in both cases the vagueness of the predicates in question gives rise to borderline cases and that these are tolerant to incremental creep. Without addressing the underlying pathology—classical logic’s apparent inability to handle vague predicates—the problem will reemerge under a new guise. The principle of uniform solution, thus motivated, seems like a sound methodological principle to ensure against further outbreaks of trouble. There is still the question of what makes the different instances of the sorites paradox of a kind. Why are the paradox of the heap and the paradox of the bald man the same paradox? And what is the underlying pathology? Clearly both these paradoxes share some underlying form, and the fact that one is about sand and the other about hair is irrelevant. And it’s not too difficult to spell out what this shared underlying form is. But the problem gets more serious when we consider other, more difficult cases. Is the liar paradox of a kind with Russell’s paradox? Is Curry’s paradox of a kind with the liar? Is the totally ordered (numerical) sorites of a kind with the non-totally-ordered (multidimensional) sorites? And the case I turn to
Vagueness and Truth
31
in the last section of this chapter: is the sorites of a kind with the liar? In order to make a start on these questions, let’s revisit an old debate between Russell and Ramsey about the liar and Russell’s paradoxes.
2
THE LIAR AND RUSSELL’S PARADOXES
Russell and Ramsey disagreed about whether the liar and Russell’s paradox were of a kind. Ramsey (1925) stressed that the liar, along with Grelling’s paradox, Berry’s paradox, and other defi nability paradoxes, were linguistic, in that they relied upon thought or language. He argued that the “fault” was ambiguity in key terms such as “defines.” These paradoxes were to be contrasted with set theory paradoxes such as Russell’s paradox and the Burali-Forti paradox. Letting his logicism come to the fore, Ramsey saw the paradoxes of set theory as only relying on logical notions (such as set membership and number). 2 He thus saw two quite different kinds of paradox here: linguistic and logical/mathematical. Russell (1903), on the other hand, focused more on the underlying structure of the paradoxes and saw them all as paradoxes of impredicativity. And we might add as a further consideration in favor of Russell, both the set theory paradoxes and the liar paradoxes admit extended forms (cyclical liar and cyclical set theory paradoxes) that appear to be generated by the same formal feature: essentially the violation of the vicious circle principle (Giaquinto 2002).3 So on one way of looking at the liar and Russell’s paradox, we see a difference: one is about truth, the other about set membership. But on another way of looking at them, they are of a kind: they both depend on self-reference. What is the correct way to classify these paradoxes? Who won the day, Ramsey or Russell? It’s fair to say that although Ramsey’s classification was very influential for a considerable time, more recently it has fallen out of favor. For what it’s worth, I’m inclined to agree with Russell and not with Ramsey here, but the matter is neither trivial nor settled. Agreeing with Russell is one thing, but putting your fi nger on why the liar and the set theory paradoxes are of a kind is not possible without a substantial theory of similarity of paradoxes. Neither Russell nor Ramsey was attempting anything so ambitious as this, yet each, in their own way, was providing a partial theory about the similarity or dissimilarity of the two paradoxes under discussion. What was particularly interesting about Russell’s approach, at least, is that he can be seen to be looking for the underlying mechanism and was considering structural similarities rather
2 Moreover, Ramsey’s logicism prevented him from recognizing another possible distinction: between mathematics and logic. 3 I make more of this point in a later section.
32
Mark Colyvan
than the more superficial features of the paradoxes.4 And Russell did not need to invoke logicism or any other substantial philosophical theory to defend his view.5 In any case, let’s run with this (no doubt not entirely accurate) caricature of Russell and Ramsey on the liar and set theory paradoxes. My suggestion is that Russell was more interested in what made the paradoxes tick, whereas Ramsey was more concerned with the details of their construction (whether they were made from linguistic and belief terms or from terms from set theory). What is appealing about determining what makes the two paradoxes tick is that once this is known, a more systematic, unified treatment can be applied. We won’t be merely treating the symptoms, we will know what the real problem is. A brief look at another unifier, Graham Priest, will help to see the benefits of getting at the root of the problem.
3
THE INCLOSURE SCHEMA
Graham Priest (2002, 134), following a suggestion from Russell, has proposed a formal schema that can be used to help classify paradoxes: the Inclosure Schema. Inclosure Schema: There are two properties φ and ψ and a function δ such that (1) Ω = {y|φ(y)} exists, and ψ(Ω) (2) if x is a subset of Ω such that ψ(x), then (a) δ(x) ∉ x, and (b) δ(x) ∈ Ω. Priest sees this schema as a paradox litmus test: if the paradox in question can be made to conform to this schema, then it is of a kind with others that satisfy the schema. It should not be surprising that according to this
4 But in fairness to Ramsey, he probably did not see the fact that the liar involved truth, and Russell’s paradox involved set membership as superficial at all. Indeed, he may well have thought that this difference resulted in two quite different mechanisms. 5 Again, in fairness to Ramsey, it is easy to see reliance on logicism as a defect of Ramsey’s account, but at the time, logicism was a real contender for the philosophical account of mathematics. Invoking such an account to do some work in the classification of paradoxes is perfectly understandable.
Vagueness and Truth
33
way of carving up the territory, many of the usual suspects turn out to be of a kind: Burali-Forti paradox, Russell’s paradox, the liar paradox, liar cycle paradoxes, Berry’s paradox, and Grelling’s paradox, to name a few. But interestingly, one is left out in the cold (Curry’s paradox) and one other has its status in dispute (Yablo’s paradox). Let’s look at Yablo’s paradox in a little more detail, because the debate over it is instructive. Yablo’s paradox also comes in both set- and truth-theoretic versions. Here is the truth-theoretic version: (s1) For all k > 1, sk is not true. (s2) For all k > 2, sk is not true. (s3) For all k > 3, sk is not true. .
.
.
.
.
.
.
.
.
(sn) For all k > n, sk is not true. .
.
.
.
.
.
.
.
.
The above infi nite list is paradoxical and, on the face of it at least, there is no self-reference or cycles of reference. But the paradox is defi nitely liar-like. Indeed, it is a kind of limiting case of longer and longer liar cycles. Why this paradox is so important is that if, as Stephen Yablo (1993) originally suggested,6 it is non-self-referential yet liar-like, then self-reference does not seem to be the culprit after all. The source of the trouble in the so-called paradoxes of self-reference must be elsewhere. Moreover, if Yablo’s paradox is not circular, it fails to fit the inclosure schema and thus fails to be classified as of a kind with the liar. But everyone agrees that this paradox is liar-like. Some such as Priest (1997) and Beall (2001) suggest that, despite initial appearances, the paradox
6 And more recently supported by Roy Sorensen (1998) and Bueno and Colyvan (2003).
34
Mark Colyvan
is self-referential because a fi xed-point construction is required in the derivation of the paradoxical conclusion. There are many interesting points arising from this debate. First, those who argue that Yablo’s paradox is, in fact, circular do so by invoking a quite different notion of circularity than the more intuitive notion, appealed to in liar cycles. After all, “requiring a fi xed-point construction in the derivation of the paradoxical conclusion” is quite different from “refers to another sentence that refers to . . . that refers to the fi rst sentence.” I’m not claiming that thinking of circularity in terms of fi xed-point constructions is illegitimate. Far from it. This seems a perfectly reasonable sense of circularity (or self-reference). I am just noting that the inclosure schema is not doing all the work here; we can do a little massaging of terms like “circularity.” But perhaps most interesting of all are the reasons for thinking that even when the structural features of the paradox (whether it is circular and whether it satisfies the inclosure schema) are in dispute, there can be agreement about whether Yablo’s paradox is liar-like. It is worth spelling out some of the reasons for this agreement. 1. Yablo’s paradox is very naturally seen as the limiting case of longer and longer liar cycles. That is, it bears a certain close relationship with paradoxes that are liar-like. 2. Yablo’s paradox, like the liar paradox, has a set-theoretic analogue.7 3. Yablo’s paradox, like the liar, comes in both unstrengthened and strengthened forms. That is, Yablo’s paradox can be constructed in terms of falsity or in terms of untruth (respectively). 4. And last, but not least, both the liar and Yablo’s paradox use the same basic machinery (i.e., truth, negation, T-schema, etc.). These points of contact between the liar and Yablo’s paradox suggest that features other than internal structural features (as displayed by the inclosure schema) are playing a role. We also saw this in relation to Russell’s paradox and the liar, although there it was less clear because the two paradoxes in question also shared a common structure (as revealed by the inclosure schema). In the case of Russell’s paradox and the liar, the extrastructural feature was that they both admitted extended forms. That is, the two paradoxes had similar, nearby neighbors. There is another point of similarity between all four of these paradoxes: they admit the same kind of solutions. I discuss this issue further in the next section. The important point to note for now is that although the inclosure schema (or other ways
7 Whatever you think about the Russell–Ramsey debate over the liar and Russell’s paradoxes, it is clear that Russell’s paradox is at the very least a set-theoretic analogue of the liar paradox.
Vagueness and Truth
35
of revealing internal structural similarities) is clearly relevant, it cannot be the whole story. There is more to paradox classification than recognizing similarity of structure.8
4 WHAT COMES FIRST, THE DIAGNOSIS OR THE TREATMENT? Let me return to the medical analogy that I used to motivate the principle of uniform solution: we want to treat the disease and not merely the symptoms. The idea is that we don’t want to employ different solutions for what are essentially the same problem, for otherwise we are missing the “mechanism.”9 But sometimes we cannot fi nd the mechanism before we consider treatments. Some psychiatric disorders, I’m told, are like this. The mechanisms are complex and often inaccessible, but classifying the diseases in terms of the treatments they respond to is still helpful. Indeed, it might be argued that insofar as a classification by treatments works, it works because there is a common mechanism, and the treatment is latching onto that mechanism. To give one rather dramatic example, before the days of quick, reliable, and readily available brain imaging, there were two quite different types of acute cerebral disorders that presented identically (or near enough) clinically. Consider a patient with an acute onset of weakness down one side of the body, slurred speech, and relaxed facial muscles on the same side as the weakness. These are classic stroke symptoms, but, as it turns out, there are two quite different mechanisms for this. One is a thrombotic event (a piece of clot that lodges in an intercerebral artery and blocks blood supply to all points distal to the block. The second mechanism is a burst blood vessel in the brain that results in an intercerebral bleed that puts pressure on the nearby areas of the brain (and cuts off blood supply to those regions). There is a very reliable test to tell these two apart: administer an anticoagulant such as heparin. If the stroke were due to the fi rst mechanism (the thrombotic event), the patient would often make a dramatic recovery (if caught early enough) after the heparin. On the other hand, if the stroke were due to the second mechanism (the bleed), then the patient would see a rapid decline after heparin and possibly die. There are, of course, less extreme
8 There is also the substantial issue of the level of abstraction at which the structural similarities are revealed. See N. J. J. Smith (2000) and Priest (1994, 2000) on this issue. 9 In the case of diseases “mechanism” talk is legitimate, but of course in the logical and set-theory paradoxes, such causal language is out of place. But still there is the thought that there is something that is at the root of the paradox—that which makes it tick (to invoke another analogy).
36
Mark Colyvan
cases of this type of diagnosis via cure (or failure to cure), but this one is useful to make the point because the underlying mechanisms are quite straightforward. That’s all well and good for medicine, but what’s that got to do with classifying paradoxes? Well, we might take a leaf out of medicine’s book and consider classifying paradoxes by the methods they respond to (and fail to respond to). For example, here is a reason why we might consider the liar and the set-theory paradoxes as alike: they both respond equally well to hierarchical treatments. They also both respond to three-valued approaches (partial set membership and three-valued logics) equally well. Notice that I’m not claiming that either the hierarchical approach or the three-valued approach is a genuine solution; it’s just that in each case these approaches do about as well and face much the same problems. We could then suggest that the reason for this is that the treatments are latching onto, partially latching onto, or completely missing (as the case may be) the underlying mechanisms. So here we have the beginnings of a classification of the underlying problem in terms of the cure, partial cure, or failed cure (as the case may be). But recall that we are trying to justify the principle of uniform solution. In the context of this justification, isn’t classifying paradoxes by their solutions question begging or vacuous? After all, the suggestion seems to amount to something like paradox A and paradox B are the same paradox because they admit the same solution, therefore we should use the same solution on them. This vacuousness objection would be a fair criticism of simply using successful solutions as a means of classifying paradoxes, but that is not what I’m advocating here. Rather, I am advocating that we should use treatments—successful, partially successful, and unsuccessful—as a tool in the classification of paradoxes. Nor am I suggesting that this should be the only tool. I’ve already suggested that Priest’s inclosure schema is a very useful way to reveal internal structural similarities, and I also mentioned that the relationship to nearby paradoxes can be revealing. Indeed, the latter is closely related to the present suggestion of classifying by treatment. In the next section I examine the connections between these two proposals.
5 THE LOCATION OF PARADOXES IN “DIALECTIC SPACE” So far I have argued that there are a number of methods we might use to show that two paradoxes are essentially the same. Two of these, classification via response to treatments and classification via near neighbors, are related. Let’s return to the (simple) liar and Yablo’s paradoxes, each couched in terms of falsity. I have already noted that one reason for thinking that they are essentially the same (or at least of a kind) is that they
Vagueness and Truth
37
both have nearby neighbors—strengthened liar and strengthened Yablo paradoxes, couched in terms of untruth. But now consider the usual line of reasoning that relates the simple versions and the strengthened versions. Consider a truth-value gap approach to both the simple liar and the simple Yablo paradox. According to this approach, the paradoxical sentences in question (the liar sentence and every sentence in the Yablo list) fail to take a truth-value, and we invoke a suitable logic to support such claims (such as K 3). But, as is well known, this gappy approach suffers serious revenge problems at the hands of the strengthened versions of the paradoxes in question. I want to make it clear that I am not claiming that gappy solutions do not work (at least I’m not claiming that here); I am only pointing out that in both the liar and Yablo’s paradoxes, there are nearby strengthened versions that hit the gappy solution very hard and in much the same way. Now consider glutty approaches to the two paradoxes in question. According to this approach, the sentences in question are both true and false, and we invoke a suitable logic to support such claims (such as LP). As is also well known, these glutty solutions perform better when confronted with strengthened versions of the paradoxes. What this suggests is that although the relationship of the paradoxes to their nearby neighbors is significant, the dialectic connecting the various paradoxes is also important. The way the nearby neighbors respond or fail to respond to proposed treatments tells us something about what makes the whole family tick. Indeed, the very reasons for introducing the neighbors is important. In the cases just discussed, the neighboring strengthened paradoxes were introduced to raise problems for gappy approaches. So I am suggesting that we do not just consider the internal structure of the paradoxes—although that is undoubtedly important—we also consider the external relationships—the relationships to other nearby paradoxes. Moreover, I’m suggesting that we consider these latter relationships in terms of the dialectic that connects them all—the successes, partial successes, and failures of various treatments and the moves and countermoves in the debate that ensues. This proposal is not so neat and tidy as Priest’s elegant inclosure schema. Be that as it may, the added complexity is needed. For a start, the inclosure schema is at best limited in its application; it is only designed to classify paradoxes into two categories: those that fit the schema (the liar-like ones) and those that don’t. What about further classification among those that do not fit the schema? Priest is certainly not suggesting that they should all yield to the same treatment; he is only suggesting that those that fit the schema should be treated similarly. But we can reasonably ask about the similarity of paradoxes that do not fit the schema. Is the prisoners dilemma a Newcomb problem (Lewis 1979)? Is the two envelope paradox a St. Petersburg paradox (Chalmers 2002)? Are the St. Petersburg and the Pasadena paradoxes of a kind (Colyvan 2006; Nover and Hájek 2004)? And, as we will consider in the next section, are the sorites and the liar of a kind (Beall
38 Mark Colyvan and Colyvan 2001; Field 2004; Hyde 2001 and unpublished; Tappenden 1993)? To answer these questions we need something more discriminating than the inclosure schema. This is not a failing of the inclosure schema, of course. It was never intended to answer such questions. To answer these questions we need a more general framework for tackling the questions about similarity and identity of paradox. The proposal I’ve sketched here is a fi rst step towards such a framework.
6
THE LIAR AND THE SORITES
It is not obvious that the liar and the sorites share structural similarity,10 but there are other considerations that push for their similarity. The fi rst point of similarity is that both paradoxes have strengthened forms that raise problems for gappy solutions. So, as is well known in the case of the liar, if you are tempted to advance a gappy solution, whereby the liar sentence is neither true nor false, you face revenge problems in the form of the strengthened liar. The trick here is to take the tripartite division the gap theorist advances—true, false, and neither—and collapse two of the categories, and put pressure on the new bipartite division. The usual way this is done is by defi ning a new predicate, untrue, whereby a sentence is untrue if and only if it is either false or gappy. The revenge problem arises in the form of the strengthened liar sentence: This sentence is untrue. Gaps may solve the original liar but they fall foul of the corresponding revenge problem.11 Moreover, it is agreed by everyone that the strengthened liar is of the same kind as the original liar, and so wheeling in a different solution for the strengthened form of the paradox is not a viable option (for reasons of uniform solution). Now notice how so-called higher-order vagueness and the related higher-order sorites can be seen as a kind of revenge problem for gap approaches to the sorites. Recall that the gap approach to the sorites posits a third category—neither true nor false—to accommodate all the borderline cases. The higher-order problem involves the same trick as sketched above: the tripartite division is collapsed to a bipartite one via a new predicate. In this case, the usual predicate is determinately a heap, where something is determinately a heap if and only if it is not determinately a nonheap and not a borderline case. Now the higher-order sorites is rolled out and we fi nd that the boundary between the categories “determinately a heap” and the (fi rst-order) “borderline case” also
10 Although, see Chase (unpublished) for some reason to suspect that a continuous version of the sorites is structurally similar to the liar. 11 Although I take such revenge problems to be very serious for gap accounts, I am not suggesting that it’s game over for gappy accounts.
Vagueness and Truth
39
allows incremental creep. The dialectic structure is the same as with the strengthened liar, where we think of the higher-order sorites as a revenge problem, analogous with the liar revenge problems. Next it is worth noting that, in both the case of the liar and the sorites, glutty solutions fare better with regard to these revenge problems. But in both cases there is an intuitive revenge problem in the vicinity that would do to gluts what the above revenge problems do to gaps. Take the liar fi rst. Intuitively the predicate in question needs to collapse the three categories of true (only), both true and false, and false (only) to two. The predicate in question would appear to be “true-and-true-only” with the corresponding strengthened liar sentence: This sentence is not true-and-true-only. But here’s the problem for this line of attack on the glutty approach: this sentence is also paradoxical so is both true (only), and not true-and-true-only. The revenge problem one is grasping for can be formulated, but the sentence in question doesn’t convey the meaning you would like. Some have argued that this shows that the glutty solution is somewhat unsatisfying, in that it blocks revenge by what appears to be a trick akin to not allowing the opponent the expressive resources to formulate the relevant revenge problem.12 I won’t enter into that issue here. It is sufficient to note that the intuitive revenge problem for the glut solution does not deliver the serious blow that the analogous revenge problem does for the gap solution and that this lack of revenge is for a very specific reason. Now return to the sorites. The situation here is analogous. We’ve already seen how higher-order sorites can be seen as a revenge problem for gappy approaches. But what about a revenge problem for glutty approaches? Here the new predicate in question would need to be “true-and-true-only a heap,” and the strengthened sorites series would be constructed in terms of this predicate. But as with the liar, there is no problem allowing that borderline sentences are both true-and-true-only and also not true-andtrue-only. These hypercontradictions block the revenge problem in the case of the liar, and they do the same for analogous strategies in the case of the sorites. This is not to say that the glut approach solves the sorites. There is a lot more work to be done on spelling out the details of the glutty solution to higher-order vagueness. All I am claiming for present purposes is that, just as with the liar, revenge problems hit the gappy approach, but the corresponding glutty revenge problems lack any bite. Moreover, they lack bite for the same reason—because of hypercontradictions. I have argued that both the liar and the sorites have nearby neighbors in the form of strengthened forms—the strengthened liar and the higherorder sorites—that play similar roles in the dialectic against gaps (and for gluts). That is, there is a case to be made for the liar and the sorites occupying similar positions in their respective “dialectical spaces.” Also, it
12
For instance, Daniel Nolan has made this point in conversation.
40 Mark Colyvan seems that both the liar and the sorites respond to gappy approaches about as well as each other and that gap solutions have serious difficulties with strengthened forms. And, more tentatively, a case can be made that glutty approaches do somewhat better in the face of the relevant revenge problems in both cases. This suggests that there may be a link between the sorites and the liar via extrastructural considerations. Is this enough to make the case that the liar and the sorites are of a kind and therefore require a uniform solution? So far we have similarity established by two of the three means I outlined in earlier sections. The fi nal one is structural similarity, and that has not yet been established. Is two out of three good enough, or is structural similarity the one that really matters? Perhaps the other two (position in dialectic space and diagnosis by cure) are mere signs of underlying structural similarity? These are interesting questions that are difficult to get much traction on at the present time. But whichever way the answers to these questions fall, I take it that there are some points of contact between the liar and the sorites; these points of contact are sufficient to warrant further investigation into possible structural similarities13 and to keeping an open mind about a uniform treatment of both the paradoxes of vagueness and truth.14
13
See Priest et al. (unpublished) for some preliminary work in this direction. I am grateful to James Chase, Dominic Hyde, Carrie Jenkins, Kristie Miller, Hugh Mellor, Daniel Nolan, Graham Priest, Nick Smith, Roy Sorensen, and Zach Weber for correspondence and conversations on the issues dealt with in this chapter. I am also indebted to audiences at the Truth and Reality Conference at the University of Otago in January 2007, as well as audiences at Macquarie University, the University of Sydney, the University of Miami, the Society for Exact Philosophy Meeting at the University of British Columbia in May 2007, and the Australasian Association of Philosophy Conference at the University of New England in July 2007. This research was funded by an Australian Research Council Discovery Grant (grant number DP0666020). 14
Part II
Reality
4
Biological Realisms Michael Devitt
The species category does not exist. (Ereshefsky 1998, 113) Legitimate species . . . change with the explanatory schemata and practical interests of the biologists. (Stanford 1995, 83) Taxa of higher rank than species do not exist in the same sense as do species. (Eldredge and Cracraft 1980, 327) To the cladist true believer, there is no such thing as a reptile. (Sterelny and Griffiths 1999, 197)
1
INTRODUCTION
Realism issues tend to be confusing because of the bewildering number of “defi nitions” of what realism is. A large part of the problem, I have argued in Realism and Truth (1997), is that doctrines that should be metaphysical have become entangled with epistemological and semantic doctrines. The various realism issues in biology do not seem to have that problem, but there is still some unclarity about the nature of those issues, as David Hull notes. The most active of the issues arises out of the debate between “species monists” who think that there is just one good “species concept”—one good account of what it is to be a species—and “species pluralists” who think that there are many. Hull fi nds the combinations of monism + realism and pluralism + antirealism “quite natural” but the combinations of monism + antirealism and pluralism + realism “somewhat strained.” Still, he notes that “real can be defi ned in such a way” as to remove these strains (1999, 25–26). So he sees the issue as partly a verbal one over defi nitions.1
1
Elliott Sober (1980, 203) struggles with what Ernst Mayr means by “real.”
44
Michael Devitt
My main aim in this chapter is to explore realism issues in biology with particular attention to those generated by the monism–pluralism debate and other taxonomic disputes. 2 To this end I devote a considerable amount of energy to attempting to get clear about what the issues are. In §2, I summarize the various species concepts. In §3, I summarize the motivation for pluralism. In §4, I discuss realism in general, particularly “realism about the external world,” from a perspective elaborated previously (Devitt 1997). In §5, against this background, I define some doctrines of biological realism. In §6, I look sympathetically at Marc Ereshefsky’s moderate pluralistic antirealism (1998), and in §7, unsympathetically at Kyle Stanford’s radical pluralistic antirealism (1995). Both of these positions consist of one of the combinations that Hull finds “quite natural.” They will be compared with Philip Kitcher’s pluralistic realism (1984), which consists of one of the combinations that Hull finds “somewhat strained.” §8 deals with antirealism about genera and higher categories in the Linnaean hierarchy.
2
SPECIES CONCEPTS
Species pluralism arises out of the controversy over which species concept is correct. “The species problem is one of the oldest controversies in natural history” (O’Hara 1993, 231); it is “one of the thorniest issues in theoretical biology” (Kitcher 2003, xii). 3 There are around two dozen species concepts and “at least seven well-accepted ones” (Ereshefsky 1998, 103). Samir Okasha (2002) places them in “four broad categories”: 1. Phenetic concepts. On this sort of view, organisms are grouped into species on the basis of overall similarity of phenotypic traits. This is thought by its proponents to have the advantage of being fully “operational.” Okasha says that phenetic concepts are “the least popular” (2002, 199) and this is hardly surprising because they arise from the “philosophical attitude . . . of empiricism” (Sokal and Crovello 1970, 29). “Phenetic taxonomists have often wanted to segregate taxonomy from theory” (Sterelny and Griffiths 1999, 196). 2. Biological Species concepts (“BSC”). The most famous example of BSC is due to Ernst Mayr. He defined species as “groups of interbreeding
2 I shall not consider any threats to realism that might be thought to come from antiessentialism or from the indeterminacy of boundaries. I argue for biological essentialism elsewhere (2007). Part of this argument is that essentialism is not threatened by the indeterminacy. I think realism is not either, for similar reasons. 3 Although, interestingly enough, an issue that Darwin himself was skeptical about: he talks of “the vain search for the undiscovered and undiscoverable essence of the term species” (1859, 381).
Biological Realisms 45 natural populations that are reproductively isolated from other such groups” (Mayr 1969, 26). Kim Sterelny and Paul Griffiths remark that “If the received view has a received species concept” it is BSC (1999, 188). 3. Ecological Niche concepts. According to these concepts, a species occupies a certain ecological niche. “A species is a lineage . . . which occupies an adaptive zone minimally different from that of any other lineage in its range and evolves separately from all lineages outside its range” (van Valen 1976, 70). Okasha puts the view succinctly: species “exploit the same set of environmental resources and habitats” (2002, 200). 4. Phylogenetic–Cladistic concepts. In this view we “identify species in terms of evolutionary history . . . [with] particular chunks of the geneological nexus. . . . Species come into existence when an existing lineage splits into two . . . and go extinct when the lineage divides, or when all members of the species die” (Okasha 2002, 200). Sterelny and Griffiths claim that “something like a consensus has emerged in favor of a cladistic conception of systematics” (1999, 194). An important feature of the Phylogenetic–Cladistic concept is that it is, as everyone agrees, incomplete. It needs to be supplemented by a theory of speciation, a theory that explains when a lineage has split in two. For this, as Okasha says, it “will have to rely on a concept of one of the other types” (2002, 201). The various species concepts are answers to what Mayr (1982, 253–254) calls the “species category” problem: they are telling us what it is for a kind to be a species rather than a subspecies (variety), genus, or whatever. In claiming this I am not being controversial. However, my claim about the following question is controversial: do these species concepts tell us anything about what Mayr calls the “species taxon” problem? Do they tell us about what it is for an organism to be a member of a kind that happens to be a species? The answer to the “fi rst level” taxon problem tells us why Fido is a dog. An answer to the “second level” category problem tells us why dogs are a species, but poodles aren’t. Now it is common to think that the species concepts not only answer the category problem but also the taxon problem (Dupré 1981; Sterelny and Griffiths 1999; R. A. Wilson 1999a; Okasha 2002). I have argued elsewhere (2007) that this common thought is wrong. But the main point to make here is that theories about the species category are one thing, theories about species taxa, another. In particular, realism about the category is one thing, realism about taxa, another.
3
SPECIES PLURALISM
Now, as I have noted, controversy has raged over which of the many species concepts is right. In the face of this controversy some have argued that we should abandon the monist idea that just one concept is right and hence
46
Michael Devitt
that there is just one species category. Rather we should adopt the pluralist idea that many of the concepts are right, and hence there are many species categories. According to Kitcher, many concepts “can be motivated by their utility for pursuing a particular type of biological inquiry” (1984, 118). Stanford puts the point thus: “certain explanatory demands are inextricably bound to certain species concepts” (1995, 72). And there are many different, but equally legitimate, types of biological inquiry and explanatory demands: “we have independent and legitimate explanatory interests in biology which require distinct concepts of species” (1995, 76).4 Consider BSC, for example. Its popularity is undoubtedly well based. Thus, it had a famous triumph identifying reproductively isolated sibling species of Anopheles mosquitos, so important for understanding the spread of malaria: The sibling species . . . are similar morphologically and inhabit overlapping ranges, so neither a morphological nor an ecological criterion of species division could have been of much use in attempting to unravel this particular mystery. (Stanford 1995, 73) As Kitcher says, “there is no doubting the importance of reproductive isolation as a criterion for demarcating certain groups of organisms” (1984, 118). Yet BSC has some obvious problems as the only criterion. First, it is not good for paleontology: There is a perfectly legitimate paleontological question which focuses on the rates and patterns of morphological diversification within evolving lineages, and paleontologists pursue this question by dividing lineages into species according to morphological changes. (Kitcher 1984, 119) Second, the criterion obviously fails to apply at all to the vast number of asexual species: organisms that don’t breed at all are obviously not members of an interbreeding population. Attempts by Mayr and others to get around this difficulty are very unconvincing, as Kitcher points out (1984, 119–120).
4 As Hull (1997) emphasizes, many biologists require that a species concept be not only theoretically explanatory but also easily applicable: it must be “operational.” Hull rightly points out that “the philosophical arguments against operationism are decisive” but then goes on to be surpisingly tolerant of “operational criteria for theoretical concepts” (Hull 1997, 371).
Biological Realisms 47 Third, hybridization is a big problem for BSC, nicely described by Stanford: Hybridization . . . occurs with significant evolutionary consequences in butterfl ies, in leaf hoppers, in fishes, and in many kinds of birds. . . . In the plant kingdom, however, hybridization achieves full generality: naturally-occurring hybrids are found in every major plant group, and are common among higher plants. . . . The production of fertile offspring often results in the formation of hybrid swarms, that is, highly variable, but stable populations consisting of members of the two parent species. An even more complex arrangement is the syngameon, in which natural hybridization links anywhere from a few to a very large number of species together in one inclusive interbreeding unit. (Stanford 1995, 73–74) Clearly BSC is not going to do the explanatory job for us here: “We should do what plant biologists have already done, and divide species within syngameons on morphological and ecological grounds” (Stanford 1995, 74). Fourth, there are “ring species . . . chains of populations in which each link can breed with its neighbors, but populations separated by a number of links cannot”; for example, populations of black-backed gulls (Sterelny and Griffiths 1999, 189). In other words, being able to interbreed with is not a transitive relation. So BSC will lead to the contradiction that two populations both are and are not members of the same species. 5 These sorts of cases force our attention to “the diversity of biological interests” (Kitcher 1984, 125), motivating different ways to classify organisms into species. And these ways will often lead to different classifications. Thus Ereshefsky claims that the BSC interbreeding and the PCC phylogenetic approaches to species “carve the tree of life in different ways. Many interbreeding species fail to be phylogenetic species, and many phylogenenetic species fail to be interbreeding species” (1998, 105). I have doubts about the extreme form of pluralism urged by Kitcher, doubts that I air in §6. Still, the case for some form of pluralism strikes me as strong. But what bearing does this monism–pluralism issue have on biological realism? As we noted in §1, Hull fi nds the combinations of monism + realism and pluralism + antirealism “quite natural” but the combinations of monism + antirealism and pluralism + realism “somewhat strained” (1999, 25). To assess this, we need to be clear about what biological realism is. And to get clear about that, it will help to start by considering realism in general.
5 For more on the troubles of BSC, see van Valen (1976), Cracraft (1983), and Dupré (1999).
48 Michael Devitt 4
REALISM IN GENERAL
The background issue that is most relevant is often known as “realism about the external world,” concerned initially with the observable entities of common sense but spreading to scientific entities, both observable and unobservable. Let us attend only to scientific entities. What is realism about these entities? Why should we believe it? I have addressed these questions at length in Realism and Truth (1997) and subsequent essays (1999, 2001, 2002). I summarize. Realism is a doctrine with two dimensions. “The existence dimension” is a commitment to the existence of the entities posited by science. There are observable entities like rivers and planets and including biological entities like cats and mollusks. And there are unobservable ones like atoms and photons and including biological entities like dust mites and bacteria. The commitment is qualified, of course, to allow for some errors. Idealists, the traditional opponents of realists, have typically not denied this dimension; or, at least, they have not straightforwardly denied it. What they have typically denied is “the independence dimension.” According to some idealists, the entities identified by the existence dimension are made up of mental items, “ideas” or “sense data,” and so are not external to the mind. In recent times, under the influence of Kant, another sort of idealist has been much more common. According to these idealists, the entities are not, in a certain respect, “objective”: they depend for their existence and nature on the cognitive activities and capacities of our minds; we partly “construct” them by imposing our concepts. Furthermore, because we often differ in our world views and hence differ in our concepts, we construct different worlds. This “constructivism” is the view of the very influential philosopher of science, Thomas Kuhn (1970). Realists reject all such mind dependencies. We can capture the realist doctrine well enough as follows: Realism: Entities of most scientific kinds exist mind-independently. The independence dimension has rightly been the focus of the debate, but we should not lose sight of the importance of the existence dimension. For this dimension identifies the entities that are the subject of the dispute over independence. In particular, the dimension distinguishes a realism worth fighting for from a commitment to there merely being something independent of us, a commitment merely to “fi g-leaf” realism (Devitt 1997, 23). This realism about the external world needs to be kept quite distinct from another doctrine traditionally called “realism.” This other realism is about “universals,” abstract entities such as kinds, properties, or sets. Insofar as this realism arises from the “one-over-many problem” I follow Quine (1948) in thinking it arises from a pseudoproblem (Devitt and Sterelny
Biological Realisms 49 1999, 277–279). Furthermore, I lean towards the view that there are no good reasons for realism about universals—I lean toward nominalism—but I do not try to argue the matter. Indeed, it seems to me best to discuss the issues of realism in biology without any commitment on this vexed, millennia-old, metaphysical issue.6 I shall continue to talk of kinds as if they existed but remain neutral on whether this is a mere manner of speaking that can be paraphrased away or whether it amounts to a real commitment. However, there is one thing that we cannot remain neutral on: if cats exist then so also does the mereological sum of cats, the “cat fusion.” But this cat fusion is no more abstract than a cat. Realism about the external world is a compelling doctrine almost universally held outside intellectual circles. What, then, has persuaded so many philosophers out of it? The tradition provides a clear answer: the problem of extreme skepticism. In the First Meditation Descartes famously doubted the evidence of his senses. Is he right to believe that he is sitting by the fi re? Perhaps he is suffering from an illusion, perhaps he is dreaming, perhaps he is being misled by an evil demon. In the face of such doubts, how can it be rational to believe realism? Idealists think that it is not rational. They see an unbridgeable “gap” between the knowing mind and the objective independent world that the realist believes in. They propose to close the gap between us and the world by abandoning the independence dimension: the world is made up of ideas or is partly constructed by the knowing mind. Only thus, it is thought, could the world be knowable. A semantic variant of this argument can be abstracted from contemporary antirealist discussions (Kuhn 1970; Putnam 1978, 1981). Just as traditional philosophers argued for epistemological doctrines that show that we could not know the realist world, we can see contemporary philosophers as arguing for semantic doctrines that show that we could not refer to the realist world. So the world we refer to cannot be that world but must be a world we make. Abandoning realism and adopting idealism is, however, very costly because idealism is truly bizarre. Thus, consider constructivism, according to which we partly make the familiar world by imposing our concepts. But how could we literally make dinosaurs and stars with our minds? The idea that we could deserves an incredulous stare if anything does. I argue for two other responses to these arguments against realism (2002). First, there is a Moorean response that those arguments proceed in the wrong direction. The arguments are based on speculations about what we could know and refer to. Yet surely realism is much more plausible than these epistemological and semantic speculations that are thought to
6 And even more from commitment on the equally vexed issue of the nature of properties (supposing there are any).
50 Michael Devitt undermine it. So we should “put metaphysics fi rst” and argue from realism against these speculations. The second response stems from naturalism. From a naturalistic perspective, these speculations cannot be supported a priori, and they do not come close to having the empirical support enjoyed by realism. The arguments against realism use the wrong methodology and proceed in the wrong direction. One further aspect of realism debates looms large in biology. Philosophers have been concerned not only with whether the posits of science exist mind-independently but also with whether these posits are “appropriately special” rather than somewhat arbitrary. In particular there has been a concern about whether our scientifi c posits “carve nature at its joints,” about whether there is something in the nature of the world that, in some sense, determines our categorization of it. I take this to be a concern about whether the kind of entity posited by a theory plays a causally significant role, and whether it is partly because an entity is of that kind that it has the characteristics and behavior that it has. Theories need to posit such so-called “natural” kinds if the theories are to be genuinely explanatory.7 And Realists are likely to take it for granted that their paradigm entities—for example, cats and planets—are indeed of explanatory kinds. It is important to see that this requirement that a kind be explanatorily signifi cant is distinct from the requirement that entities of that kind exist objectively and mind-independently. As we shall see in §6, it is easy to name kinds of entities that are more arbitrary than causal-explanatory and yet those entities still exist mind-independently. And if there were kinds of entities that were mind-dependent in, say, a Kantian way, those kinds could still be causal-explanatory. For this reason, it would be better to keep the issue of explanatory signifi cance distinct from issues labeled “realism.” But that has not happened in biology and so I shall not insist on it. It is worth noting also that explanatory signifi cance, unlike existence, comes in degrees: positing some kinds may be very explanatory, positing others, only a little bit explanatory, positing others still, not explanatory at all. Let us call any realism that requires not only mind-independent existence but also explanatory significance an “explanatory” realism. So much for the general issue of realism about the external world. Let us turn now to much more particular realism issues in biology.
7 “[N]atural kinds are the sets that one picks out in giving explanations” (Kitcher 1984, 132 n. 16); “the naturalness (and the ‘reality’) of natural kinds consists solely in the contribution that reference to them makes to [the accommodation between conceptual and classificatory practices and causal structures]” (Boyd 1999, 141).
Biological Realisms 51 5
BIOLOGICAL REALISMS
I have noted in §2 the importance of distinguishing realism issues at two levels to do with species: at the fi rst level of species taxa and at the second level of the species category. At the fi rst level we want to capture commitments of this sort: tigers exist mind-independently, and it is explanatorily significant that they are tigers. The commitment is to entities that are members of kinds that we call “species.” So: 1st Level Explanatory Realism about Species Taxa: Organisms of most kinds, S, that we call “species” exist mind-independently, and it is explanatorily significant that these organisms are members of S. (“Most” allows for the possibility of some errors.) At the second level we want to capture commitment to there being kinds that are mind-independently species and to it being explanatorily significant that they are species. So: 2nd Level Explanatory Realism about the Species Category: Species exist mind-independently, and it is explanatorily significant that they are species. We have in mind, of course, that the kinds we call “species,” such as tigers, are the species in question, but 2nd level category realism does not commit to this: there could be species even if we have been wrong about which kinds are species. Similar pairs of realism issues can, of course, come up for higher categories in the Linnaean hierarchy. We shall consider these issues in §8. Finally, we need to take passing note of the popular view that species are individuals rather than kinds (Ghiselin 1974; Hull 1978). On that view, the species of tigers is an individual constituted by the fusion of all tigers, and any particular tiger is a part of that individual rather than a member of the tiger kind. This would require no change to the defi nition of 2nd level category realism and an obvious change to the defi nition of 1st level taxa realism that would have no significant effect on the realism debate. So I continue to write as if species are kinds. I shall start with Ereshefsky who urges an acceptable sort of biological antirealism. I shall then consider Stanford who urges a highly unacceptable one.
6
ERESHEFSKY’S ANTIREALISM
Ereshefsky’s antirealism concerns the species category not the species taxa: he claims that “the species category does not exist” (1998, 113), but he “does not call into question the reality of those lineages we call ‘species’” (1998, 104): he is a 2nd level antirealist not a 1st level one.
52 Michael Devitt Ereshefsky’s argument, the “Heterogeneity Argument,” starts from species pluralism: “a number of species concepts should be accepted as legitimate” (1998, 103). He takes this pluralism to imply antirealism. So his position is one of those that Hull finds “quite natural.” And Ereshefsky sees himself as being at odds with pluralistic realists like Kitcher (1984) and Dupré (1993) whose positions are among those that Hull finds “somewhat strained” (1999, 25). Ereshefsky rightly points out that “species pluralism implies that the world contains different types of species” (1998, 111). He goes on: What do these different types of species have in common that renders them species? If species taxa lack a common unifying feature, then we have reason to doubt the existence of the species category. (Ereshefsky 1998, 111) Ereshefsky fi nds a suggestion in the literature for what the common feature might be, a suggestion with two components: (i) a similarity in the “process that renders species taxa cohesive entities”; (ii) a similarity of “structure” (1998, 111). He argues that neither component can do the job (1998, 111– 113). He concludes that “what is left as the common feature of species taxa is the term ‘species’” (1998, 113).8 Hence he draws his conclusion that the species category does not exist.9 What could he mean by this conclusion? He certainly does not mean that tigers and the like do not exist: we have noted already that he does not reject 1st level taxa realism. And he seems to accept that there are “different types of species,” each one captured by a different species concept: for example, “interbreeding species,” “phylogenetic species,” and “ecological species” (1998, 115), or, more briefly, “biospecies,” “phylospecies,” and “ecospecies” (1998, 117). So he accepts that there exist organisms that are members of a biospecies, organisms that are members of a phylospecies, organisms that are members of an ecospecies, and perhaps organisms that are members of other types of species. But, then, since biospecies, phylospecies, ecospecies, and perhaps others are all types of species, all of these organisms are members of a species. Indeed to be a member of a species simply is to be a member of a biospecies, phylospecies, ecospecies, or perhaps other types of species. So what entities could Ereshefsky be denying the existence of?
8 Richard Boyd claims that the types of species are unified by “an especially close homeostatic relation between the classificatory practices” (1999, 171); see also R. A. Wilson (1999a, 203–204). 9 In another paper, Ereshefsky gives a related reason for doubting the existence of the species category: the failure of attempts to distinguish species from the higher taxa, in particular the failure to do so in terms of the processes of speciation and interbreeding (1999, 269).
Biological Realisms 53 His response to a “realist rejoinder” points to the answer. In this response, he denies that there is any “distinctive commonality” among biospecies, phylospecies, and ecospecies. He is not satisfied with a disjunctive species category of the sort just illustrated. “A disjunctive defi nition of the species category would not tell us why various taxa are species . . . disjunctive definitions lack ontological import” (1998, 115). What he fi nds lacking in the species category is explanatory significance or anything close to that; the category is too close to being arbitrary. But how could this lack be a lack of existence? I take it that Ereshefsky must be thinking of the species category as an abstract entity, a “universal,” and must be presupposing a sort of selective realism about such entities. Whereas an unselective realist is committed, roughly, to there being a universal for every predicate, a selective realist is committed to there being one for some but not all predicates.10 And Ereshefsky is committed to there being one for a predicate where the objects it applies to share a distinctive commonality, but not for any other predicates. According to his pluralism, species lack that commonality. So the species category, a universal, does not exist. Now, it would be better if we could capture Ereshefsky’s position on biology without commitment to a heavy-duty, highly controversial, metaphysical thesis of selective realism about universals. And we can. We can capture it as the rejection of 2nd level explanatory realism about the species category. And the rejection is not based on denying the existence of anything but on denying the explanatory significance of kinds being species. To see that this antirealism does not arise from nonexistence, it is helpful to note that entities of many kinds exist, even exist mind-independently, where their being members of those kinds is of little or no explanatory significance. Consider cousins, for example. These include first cousins, second cousins, fifth cousins three times removed, and so on. The explanatory significance of being a cousin is surely close to zero. And we get even closer to zero if we consider step-cousins. And there is nothing to stop us naming a totally arbitrary kind: thus we could call anything that is either an acid, a river, or a bachelor a “grugru.” Yet despite arbitrariness and explanatory insignificance, cousins, step-cousins, and grugrus really exist, and do so mind-independently. So, the best metaphysical message to take from biological pluralism is not that the species category does not exist or that it is not “real.” Nor is the best message that species do not correspond “to something in the objective structure of nature” (Kitcher 1984, 128) and are not “an objective feature of the living world” (Sterelny and Griffiths 1999, 180). For
10 Thus, David Armstrong (1978) is a selective realist, holding that empty predicates, negative predicates, and, most pertinently, disjunctive predicates, have no corresponding universal. He thinks that some predicates apply to the world in virtue of many universals. Most importantly, he looks to science to tell us which properties there are.
54
Michael Devitt
according to pluralism to be a species is to be a biospecies, phylospecies, ecospecies, or perhaps other species, and there is nothing subjective or otherwise mind-dependent about that. To think clearly about these realism issues, it is vital to distinguish sharply two sorts of freedom, a freedom we have and a freedom we don’t.11 The freedom we do have is to choose to name any kind we like, whether for explanatory reasons, for frivolous reasons, or for no reason at all; naming kinds is a subjective matter. The freedom we do not have is to choose whether something is a member of a kind, whatever our reason for naming that kind in the first place; kind membership is an objective matter. We have chosen to name cats for very good explanatory reasons and grugrus for no good reason at all. But grugrus exist as objectively and mind-independently as cats. And they existed long before I named them “grugrus.” My naming them “grugrus” didn’t make them grugrus any more than people naming cats “cats” made them cats. In sum, Ereshefsky is best seen as rejecting 2nd level explanatory realism about the species category because that category is not explanatorily, or even otherwise, significant. He is best seen as rejecting this because that is what his argument from pluralism supports without the baggage of selective realism about universals. I noted that Ereshefsky sees himself in disagreement with Kitcher and Dupré who, he claims, “suggest a realistic interpretation of species pluralism: various species concepts provide equally real classifications of the organic world” (1998, 103). Is there an actual disagreement and if so what is it? I only consider Kitcher and set aside any possible disagreement arising solely from Ereshefsky’s selective realism about universals. Kitcher does, of course, call his position “pluralistic realism” but what exactly is “realist” about his position? He sets aside as “trivially true” a realism about the species category (and taxa) that is like the doctrine I have called simply “realism,” requiring only mind-independent existence. He is interested in a stronger realism: Pluralistic realism rests on the idea that our objective interests may be diverse, that we may be objectively correct in pursuing biological inquiries which demand different forms of explanation, so that the patterning of nature generated in different areas of biology may cross-classify the constituents of nature. (Kitcher 1984, 128) Kitcher is clearly a 2nd level explanatory realist about various species categories, for example, about the biospecies, phylospecies, and ecospecies
11 I have suggested elsewhere that the constructivist idea that we make worlds with our theories may arise out of failing to distinguish these freedoms (1997, 245).
Biological Realisms 55 categories. He believes that there are kinds that are mind-independently biospecies and that it is explanatorily significant that they are biospecies; similarly, phylospecies, ecospecies, and some others. Setting aside for a moment a disagreement about the variety of species categories, I assume that Ereshefsky would agree about this (1998, 117). So, pluralism can be combined without any strain at all with a sort of realism. But, of course, explanatory realism about various species categories—biospecies, phylospecies, ecospecies, and the like—is not explanatory realism about the species category itself. For the latter realism to be true it would have to be explanatorily significant that some kinds are species. Yet, according to Ereshefsky’s argument, pluralism has the consequence that the species category is disjunctive and not explanatory. So if Kitcher’s pluralistic realism is about the species category itself and not merely about various species categories then Ereshefsky is indeed in disagreement with him. Is Kitcher’s realism about the species category itself? He certainly does not argue for such a realism. And the message I take from “the moral for philosophy of science” that he draws at the end of his paper is that he rejects this realism: “species” refers not to a natural kind but to a “heterogeneous collection” of natural kinds (1984, 129). If I am right about this, then Kitcher’s pluralistic “realism” is not, in this respect, different from Ereshefsky’s pluralistic “antirealism,” and so they seem not to be in disagreement after all. However, they do differ in their pluralisms and hence in the variety of species categories that they are prepared to be realist about. Kitcher’s pluralism is radical. He follows Mayr in drawing a distinction—a very important one in my view (2007)—between two types of explanation in biology, ones that Kitcher calls “structural” and “historical.” The structural seek to “explain the properties of organisms by means of underlying structures and mechanisms.” The historical seek to “identify the evolutionary forces that have shaped the morphology, behavior, ecology, and distribution of past and present organisms” (1984, 121). Kitcher claims that these two types “generate different schemes for classifying organisms” (1984, 122). He fi nds variations within the two types leading him to posit nine different taxonomies and hence nine distinct species categories (1984, 124). Ereshefsky’s pluralism is more conservative. He is prepared to accept only species categories that are justified by historically motivated taxonomies, for example, the categories of biospecies, phylospecies, and ecospecies. His form of pluralism assumes that all species taxa are genealogical entities. To assume otherwise places species outside of the domain of evolutionary biology. The explanatory backbone of evolutionary theory is the assumption that organisms are connected by genealogy. (Ereshefsky 1998, 107)
56
Michael Devitt
However, this is not really an argument against Kitcher’s radicalism. For Ereshefsky is attending only to the explanatory needs of evolutionary biology, whereas Kitcher is emphasizing that there are other explanatory concerns in biology: there are the concerns of structural explanations. And Kitcher thinks that these explanations motivate taxonomies that are different from those motivated by historical explanations and hence different species categories. Is there an argument against Kitcher’s radicalism? I think so. Suppose that we go along with him about the nine different taxonomies. What about the inference to nine different categories? If the taxa picked out by a certain taxonomy are to justify a certain species category, S, it has to be explanatorily significant that the taxa are S. Now given the role of species in theories of evolution it is plausible to think that a historically motivated taxonomy will justify a species category: as Ereshefsky puts it, “species taxa are the paradigmatic units in which descent with modification occurs” (1998, 107). But why suppose that a structurally motivated taxonomy will justify a category? That is, even if structural explanations demand certain taxonomies, what significance is there for such explanations that some taxon is a species? Consider Kitcher’s example: A biologist may be concerned to understand how, in a particular group of bivalve molluscs, the hinge always comes to a particular form. The explanation that is sought will describe the developmental process of hinge formation, tracing the fi nal morphology to a sequence of tissue or cellular interactions, perhaps even identifying the stages in ontogeny at which different genes are expressed. (Kitcher 1984, 121) It is hard to see how it makes any difference to the structural explanation we seek whether that group of mollusks is a species, subspecies, genus, or whatever. Our interest in structural explanations may demand that we group those mollusks together in our taxonomy, but it does not seem to demand that we assign them to any particular category. So, I am inclined to think that only historically motivated taxonomies can justify species categories and that Ereshefsky is right to be conservative. I return to this disconnect between taxonomies and categories in discussing the higher categories (§8). I have been writing in this section as if, in the pluralist view, “species” is an unambiguous term applying to any kind that is in the various species categories. This seems plausible to me, but perhaps the view should rather be that “species” is ambiguous, applying sometimes to any kind that is a biospecies, sometimes to any kind that is a phylospecies, and so on. If so, “realism about the species category” would also be ambiguous, referring to several different doctrines not to just one. So Ereshefsky would be mistaken
Biological Realisms 57 in responding to the realism issue here as if there were just one doctrine to deny. The only pluralist way to be realist about “the species category” would be to be a realist about biospecies, phylospecies, and so on, a realism that I am supposing Ereshefsky and Kitcher should agree on. There would be no one doctrine, “realism about the species category.” Finally, what should be the fate of the term “species” if Ereshefsky’s argument is right (and “species” is not ambiguous)? When the theoretical chips are down, it should really have no place in biology; it should be replaced by “biospecies,” “phylospecies,” and the like. Still, the chips are often not down and, as Ereshefsky points out, “practical considerations” count in the term’s favor (1998, 118).12 And continuing to use the term may be relatively harmless given that to be a species simply is to be a biospecies, phylospecies, ecospecies, or whatever, and each of these kinds are explanatorily significant. We turn now to Stanford’s antirealism (1995).
7
STANFORD’S ANTIREALISM
Stanford’s main stalking horse is Kitcher. Kitcher has this to say in describing his pluralism: Species are sets of organisms related to one another by complicated biologically interesting relations. There are many such relations which could be used to delimit species taxa. However, there is no unique relation which is privileged in that the species taxa it generates will answer to the needs of all biologists and will be applicable to all groups of organisms. (Kitcher 1984, 113) Stanford interprets remarks of this sort as the view that “judgments of what is biologically interesting can only be made relative to a particular time and theoretical context” (1995, 80–81). In brief, these biological judgments are theory-laden. But then, nowadays, it is common, and in my view correct, to think that all our judgments are theory-laden (Duhem– Quine Thesis). So, we have nothing very controversial so far. The controversy comes in the next step. Kitcher’s view is taken to imply that “the biologically interesting divisions of organismic diversity constitute legitimate species” (1984, 82; emphasis added). So not only are our judgments about species theory-laden, but the species themselves are. So species are
12 Despite the theoretical considerations that support pluralism, Dupré thinks that pragmatic considerations favor a “feeble monism” that leaves the traditional Linnaean taxonomy pretty much in place. We should settle for that taxonomy as a “general purpose reference system” (1999, 18).
58
Michael Devitt
mind-dependent. Furthermore, “what constituted a biologically interesting relation in the past has changed dramatically . . . legitimate species must then change with the explanatory schemata and practical interests of the biologists.” Thus Kitcher’s pluralism is incompatible with his realism about species (1984, 83); we should be antirealist. Set aside whether the controversial Kantian inference from the theoryladenness of judgments to the theory-ladenness of the world is justly attributed to Kitcher—and I don’t think it is13 —it is surely popular.14 Yet, despite this popularity, the inference is startlingly bad. Or so I have argued many times (1997, 1999, 2001, 2002). The inference is a paradigm example of the folly of inferring a metaphysics from an epistemology—of failing to “put metaphysics fi rst” (§4). A realistically inclined biological pluralist will think that, despite many errors in the past and probably a few now, what biologists fi nd interesting really is interesting: that it is an explanatorily significant fact about nature that some kind is, say, a biospecies. The theoretical interests of biologists lead them to posit causal structures that are really there. Now there is, of course, a long and painful history of skeptical challenges to this sort of realism. Stanford does not mount such a challenge but presumably one must underlie his argument. But if it does, its target is all of science not just biology, as Ereshefsky points out (1998, 108). Until new life has been breathed into this general skeptical challenge, realists about anything should be unmoved by arguments like Stanford’s. Finally, is Stanford’s antirealism a 1st level one about species taxa or a 2nd level one about the species category? The answer is not as clear as it should be. Unlike Ereshefsky, Stanford does not give the important category–taxa distinction the attention it deserves. In Ereshefsky’s view “Stanford’s arguments are aimed at the existence of species taxa. He makes no explicit reference to the existence of the species category” (1998, 113–114). I think that Ereshefsky is clearly right about the latter claim but not clearly so about the former. In presenting his argument, Stanford does not distinguish explicitly between antirealism about the species category and antirealism about species taxa.15 Suppose that biologists think that a kind, say tigers, is
13 Although it has to be admitted that Kitcher does evince small, but worrying, leanings toward Kantianism; for example, (1993, 169–173). 14 For evidence of its popularity in philosophy generally and philosophy of science in particular, and in sociology, literary theory, sociology of knowledge and even in feminism, see Devitt (1997, 235–237). 15 Mayr’s distinction is established, but it is often overlooked; for example, in Dupré (1981), Griffiths (1999), and Sterelny (1999). In his brief discussion of Stanford and Ereshefsky on realism, Kevin De Queiroz (1999, 74–75) does distinguish the taxa issue from the category issue, but he does so against a background of a
Biological Realisms 59 a species and that a particular organism, say Benji, is a tiger. The conclusion of Stanford’s argument might be (i) that the fact that biologists think tigers a species constitutes their being so: that is antirealism about the species category. Or the conclusion might be (ii) that the fact that biologists think Benji a tiger constitutes his being so: that is antirealism about species taxa. Or the conclusion might be both. Now it is charitable to take the argument to be for (i) because species pluralism, the key premise of the argument, is a doctrine about the species category not about species taxa. If Stanford’s argument from species pluralism were sound, which I have just argued it is not, it would yield (i) but have absolutely no bearing on (ii). We can see this by considering an even stronger antirealism about the species category than Stanford’s. Suppose, as Stanford does not (1995, 86), that there was nothing in nature that dictated our picking out tigers as a species and so our doing so was completely arbitrary. It would still not follow that Benji’s being a tiger was in any way mind-dependent or constituted by us; think of grugrus (§6). Indeed it would not even follow that being a tiger was not explanatorily significant. Even if our thinking of tigers as a species were guided entirely by our interests and not by nature, it still might be the case that being a tiger was explanatorily significant. And it surely is significant: it features in historical explanations because being a tiger is part of the evolutionary story; It features in structural explanations because a lot of the morphology, physiology, and behavior of Benji is explained by his being a tiger.16 In sum, 2nd level explanataory antirealism about the species category does not undermine 1st level explanatory realism about species taxa. And that 1st level realism is surely right. So it would be charitable to take Stanford to be arguing only for (i) antirealism about the species category. But it is hard not to see remarks like the following as also endorsing (ii) antirealism about species taxa: species are “entities whose existence and classifi cation are not independent of our interests and activities” (1995, 89); “species are not real or mind-independent entities” (1995, 90). So it rather looks as if Stanford has reached an antirealist conclusion about both the species category and species taxa from an argument that should be seen as aimed only at the category.
defi nition of realism that applies only to taxa. The defi nition also takes realism to be a commitment only to mind-independent existence, ignoring the importance of explanatory significance in this debate. 16 It is perhaps worth mentioning that being a member of a certain subspecies or variety is also explanatorily significant: Fido’s being a pitbull explains a lot about his morphology, physiology, and behavior; cf. Cracraft’s description of concern about “the ontological status of subspecies” (1983, 100).
60 8
Michael Devitt ANTIREALISM ABOUT THE HIGHER CATEGORIES
The same pair of realism issues that come up for species can come up for any of the higher categories (“ranks”) in the Linnaean hierarchy, in “the tree of life”: genera, families, orders, classes, phyla, and kingdoms. Thus, consider the kinds we call “genera.” At the fi rst level we want to capture commitments of this sort: Canis exist mind-independently and it is explanatorily significant that they are Canis. So: 1st Level Explanatory Realism about Genera Taxa: Organisms of most kinds, G, that we call “genera” exist mind-independently, and it is explanatorily significant that these organisms are members of G. At the second level, we want to capture commitment to there being kinds that are mind-independently genera and to it being explanatorily significant that they are genera. So: 2nd Level Explanatory Realism about the Genus Category: Genera exist mind-independently, and it is explanatorily signifi cant that they are genera. We have in mind, of course, that the kinds we call “genera,” such as Canis, are the genera in question, but 2nd level category realism does not commit to this: there could be genera even if we have been wrong about which kinds are genera. Let us start with the 1st level. Whatever the troubles for realism about the species category, we found nothing that threatened explanatory realism about species taxa. And we have just pointed out why that doctrine is so plausible. Surely explanatory realisms about the higher taxa are also plausible. Thus, whether or not there is any explanatory significance to the claim that Canis is a genus, surely being a Canis features in historical evolutionary explanations and structural explanations, just as being a Canis familiaris does. And so too, say, being a mammal. This is not to say that membership in a higher taxon is as explanatorily significant as species membership: explanatory significance comes in degrees, as I noted (§4). Still, surely some of the morphology, physiology and behavior of Fido are explained by his being a Canis. In light of this, one wonders what to make of the view that “taxa of higher rank than species do not exist in the same sense as do species” (Eldredge and Cracraft 1980, 327). According to Ereshefsky, this sort of view is part of “The Modern Synthesis.” He describes that Synthesis as holding: “Higher taxa . . . are merely artifacts of evolution at the species level. So while species are real and the ‘units of evolution,’ higher taxa are merely aggregates and ‘historical entities’” (2001, 229). Ereshefsky himself rejects this view (1991, 381), as does James Mallet: “Whether
Biological Realisms 61 species do have a greater ‘objective reality’ than lower or higher taxa is either wrong or at least debatable” (1995, 296). Clearly I think that Ereshefsky and Mallet are right in their rejection, but I do not argue the matter further. Taxonomic disagreements have led to some other surprising claims that bear on 1st level realism. Sterelny and Griffiths point out that “systematics has gone through a long period of controversy, some of it extraordinarily bitter” (1999, 194). They see this as having arisen from “an attempt to have biological classification respect three goals at once.” With 1, the goal is “a maximally efficient information store”; “the full name of the organism encodes the greatest possible amount of information about the organism.” This relates to a “phenetics” taxonomy. With 2, the goal is to “reflect the disparity of organisms and the extent of their evolutionary change.” This relates to an “evolutionary” taxonomy. With 3, the goal is to “describe the branching pattern . . . through which the tree of life has grown.” This relates to a “phylogenetic” or “cladistic” taxonomy. And the problem is that “no classification system can fully satisfy all three criteria” (1999, 195). So can any of these goals be ruled out? Sterelny and Griffiths fi nd problems with the phenetic 1 and the evolutionary 2 (1999, 196–197) and favor the cladistic 3, which they take to be the emerging consensus (1999, 194). Yet 3 generates a surprising claim. Aside from its focus on evolutionary history, cladistics involves “a metaphysical claim” that is very relevant to 1st level realism: real groups in nature are all, and only, monophyletic groups . . . groups that consist of a species and all, and only, its descendents. To the cladist true believer, there is no such thing as a reptile. “Reptile” does not name a real group, for there is no species that is ancestral to all the reptiles that is not also ancestor of the birds. (Sterelny and Griffiths 1999, 197) Now, we should note fi rst of all that the claim that “there is no such thing as a reptile” is a most unhappy way of putting the cladistic point. Of course there are reptiles because there are crocodiles, snakes, and lizards, and to be a reptile is simply to be one of those. Reptiles exist, and do so mind-independently, just as much as cousins, step-cousins, and grugrus. And it is not much more apt to say that reptiles are not a “real group” in nature: there is nothing interestingly unreal about crocodiles, snakes, and lizards. I take it that the best way to put the cladistic point is to say that being a reptile is not an explanatorily significant kind. In general, the cladist is against 1st level explanatory realism for any group that is not monophyletic. Is this restriction to monophyletic groups appropriate? Although Sterelny and Griffiths favor cladism, they are inclined to think that the restriction is too extreme:
62 Michael Devitt there may well be sensible evolutionary hypotheses about all the nonmarine mammals. The group is not a monophyletic clade, because there is no species ancestral to all the land-breeding mammals that is not also ancestral to the whales. (Sterelny and Griffiths 1999, 198)17 This is an evolutionary reason for abandoning strict monophyly. There are other explanatory concerns in biology that seem to demand this too. There are the sort of structural concerns that partly motivate the phenetic 1 and the evolutionary 2. There seems no reason to think that only monophyletic classifications can serve those other concerns. Perhaps being a reptile is explanatorily significant in many cases.18 Doubtless we should not be 1st level realists about some nonmonophyletic groups that biologists have posited, but the cladist’s case that we should not be realist about any seems inadequate at best, reflecting attention only to historical explanations.19 Turn now to 2nd level realism. We might expect taxonomic disputes to play a role in two ways: (I) the pluralism about the species category that we have been discussing may have repercussions for higher categories (ranks); (II) there is the possibility that the just-mentioned dispute between phenetic, evolutionary, and cladistic classifications is relevant. Consider (I). Species are the base taxa in the hierarchy. A consequence of species pluralism is that some kinds are biospecies, some, phylospecies, some, ecospecies, and so on. Then which type of species constitute the base taxa? Could we just say that they all do? In thinking about this, we need to note a range of possible relations that might exist among these types of species. (i) One possibility is that the members of a kind that is of one type of species are always coextensive with the members of a kind that is of another type. This possibility would have no repercussions for the hierarchy, but it is surely not a real possibility. (ii) A possibility that is real
17
Boyd is also skeptical of monophyly (1999, 182). It is also surely explanatorily significant that something is a predator or a parasite. Consider this, for example: “The Lotke-Volterra equations . . . describe the interactions of predator and prey populations” (Sober 1980, 202). But these are not the sort of classifications that concern the cladists. (Thanks to Marc Ereshefsky.) 19 Mishler concludes his summary of the argument for monophyly with this remarkably inadequate claim: “Because the most effective and natural classification systems are those that ‘capture’ the entities resulting from processes that generate the things being classified, the general biological classification system should be used to reflect the tree of life” (1999, 309–310). It is probably the case that the classification systems in all sciences “capture” entities resulting from processes— entities don’t come from nothing!—but it doesn’t follow that they should be classified to reflect those processes. 18
Biological Realisms 63 is that the members of a kind that is of one type might always exclude the members of a kind of another type; for example, sexual and asexual types. So each organism will belong to only one type of species. On this view, the base taxa could include all types of species (Mishler and Brandon 1987). (iii) It seems quite likely that members of a kind of one species type will occasionally include all the members of a kind of another type. In such a case, the tree of life could presumably accommodate both kinds. Still, it seems very unlikely that this arrangement would always be the case. (iv) A real possibility is that the members of a kind of one type will overlap with the members of a kind of another type: some organisms will be members of the one kind only, some members of the other kind only, and some members of both (Ereshefsky 1998, 106). This poses a severe problem for the hierarchy: it seems that we would have to have a distinct tree of life for each type of species; the species of each type are the base taxa for a distinct tree. If this should be the case, the strongest 2nd level realism about a higher category that we could hope for would be one for each type of that category, a position analogous to the Kitcher–Ereshefsky realism about types of species (§6). Now consider (II). In discussing Kitcher’s pluralistic realism, I noted that if the taxa picked out by a certain taxonomy are to justify a certain species category, S, it has to be explanatorily significant that the taxa are S. And, using Kitcher’s mollusk example, it is hard to see how it makes any difference to structural explanations whether a group is a species, subspecies, genus, or whatever. This carries over to the higher categories: there is a general disconnect between taxonomies and categories. So even if we could establish that a certain taxonomy—phenetic, evolutionary, cladistic, or whatever—was theoretically sound, so that classifying organisms in that way served our explanatory purposes in biology, that alone would not justify realism about any category posited by the taxonomy. Just because a taxonomy is right to classify a group of organisms as Canis and a subgroup as Canis familiaris does not show that there is any explanatory significance in treating the former as a genus and the latter as a species. Second level realism about a category requires more: the category itself must do explanatory work. Do the higher categories do any explanatory work? Sterelny and Griffiths think that within the phenetic and evolutionary “taxonomic pictures, the idea of genus, family, order, and so on makes quite good sense” (1999, 201). They do not cite any explanatory work that these categories would do, and so I suspect that they are being too generous. In any case, as noted, they are dubious of these pictures, favoring the cladistic picture instead. Cladists have a very negative view of the higher categories: taxonomic ranks make little sense . . . they do not think there will be any robust answer to the questions, when should we call a monophyletic group of species a genus? a family? an order? Only monophyletic groups should
64 Michael Devitt be called anything, for only they are well-defined chunks of the tree. But only silence greets the question, are the chimps plus humans a genus? (Sterelny and Griffiths 1999, 201) So, on the cladistic picture, all categories (ranks) above species must be abandoned. Ereshefsky agrees although, as we have noted, he abandons the species category as well. He rightly points out that if a certain category is to be acceptable, the taxa of that category must be “comparable” and draws attention to reasons for thinking that this condition is not met (1999, 299). Brent Mishler claims that “practising systematists know that groups given the same rank across biology are not comparable in any way” (1999, 310– 311). In a lengthier critique of the Linnaean hierarchy, Ereshefsky mentions the drive to introduce more ranks, leading to a hierarchy in flux (2001, 215), and to disagreements about the rank of certain taxa (2001, 226). The signs are that, although the higher categories may have some pragmatic value, they are doing no explanatory work. Second level explanatory realism for the higher categories seems to be false. In sum, I have found no good reason to abandon 1st level explanatory realism for the higher taxa, even for those that are not monophyletic. Second level explanatory realism for the higher categories is a different matter. If we go along with Ereshefsky’s species pluralism, as I am inclined to, the strongest 2nd level realism we could hope for with higher categories would be one for each type of genus, family, and so on. But the signs are that these higher categories do no explanatory work, and so 2nd level realism, hence the Linnaean hierarchy, must be abandoned. Finally, we should note that abandoning the Linnaean hierarchy is not abandoning a hierarchy altogether; it is not abandoning a tree of life. It is abandoning the labeling of categorical ranks in that tree (Ereshefsky 1999, 299; Mishler 1999, 311). But, in light of (I) and (II), we may have to accept that there is more than one correct uncategorized tree of life, each reflecting legitimate explanatory concerns.
9
CONCLUSIONS
My main aim in this chapter has been to clarify and assess realism issues in biology with particular attention to those generated by the monism–pluralism debate and other taxonomic disputes. I have approached these issues from a perspective of “realism about the external world,” a commitment to the mind-independent existence of the entities of science and commonsense. However, biological realisms typically have a further commitment: they are committed to what I have called explanatory significance, to the kinds of entities posited by biological theories playing causally significant roles. Throughout my discussion I have emphasized the importance of
Biological Realisms 65 distinguishing between 1st level explanatory realisms about taxa and 2nd level explanatory realisms about categories. Species pluralism is the view that there are several equally good accounts of what it is to be a species. Ereshefsky presents his argument from species pluralism to antirealism as an argument against the existence of the species category. I have argued that it is better seen as an argument against the explanatory significance of that category, hence an argument against 2nd level explanatory realism about that category. Not surprisingly, Ereshefsky sees his “pluralistic antirealism” as opposed to Kitcher’s “pluralistic realism.” Yet, on close inspection, the two positions are similar: they agree that the species category itself is not explanatory but that various types of that category are explanatory. However, they differ in that Kitcher’s pluralism is more radical: Kitcher thinks that structural explanations in biology justify some species categories whereas Ereshefsky thinks that only historical evolutionary explanations can do so. I presented an argument that Ereshefsky is right: even if structural explanations motivate taxonomies they do not seem to show that a species category plays an explanatory role. Stanford’s antirealism is also derived from species pluralism but is much more extreme than Ereshefsky’s: it denies the mind-independence of species. This antirealism rests on a spurious inference from the theory-ladenness of judgments to the theory-ladenness of the world. Since species pluralism is about the species category, it would be charitable to take Stanford’s mistaken argument to be only against 2nd level realism but it looks as if it is also supposed to count against 1st level realism. Finally, I considered antirealism about the higher categories. Despite the urging of cladists, there seems to be no good reason to restrict 1st level explanatory realism to monophyletic higher taxa. However, the signs are that 2nd level explanatory realism about the higher categories is false: the higher categories seem to do no explanatory work. If they don’t, then the Linnaean hierarchy must be abandoned. Furthermore, the case for various taxonomies suggests that we may have to accept more than one uncategorized tree of life. 20
20 Earlier versions of this chapter were delivered at the Truth and Reality conference in Dunedin in January 2007 and at Johns Hopkins University in February 2007. I am indebted to the discussions at those meetings for several improvements. I am also indebted to Marc Ereshefsky for helpful comments.
5
Pleonastic Platonism Alan Musgrave
1
INTRODUCTION
Platonists believe in abstract entities. Abstract entities do not exist in space and time (or in spacetime), are unchanging, and do not interact causally with spatiotemporal things. Why should anybody be a platonist and believe in such entities? My answer to this question is uncharitable and depressing and Wittgensteinian. My answer is that people are tricked by language into believing in abstract entities. I begin with a silly example of the way in which language might trap the unwary into believing in things that do not exist. The example concerns the creeps. The example is meant to be uncontroversial—nobody actually believes in the creeps. I then generalize the lessons of this example to a platonist doctrine called “ontological minimalism.” I prefer to call it “ontological maximalism” or “pleonastic platonism.”
2
CREEPS REALISM
Osama Bin Laden gives me the creeps. I dare say that I am not the only person to whom Osama Bin Laden gives the creeps. So it seems that there is a strange entity, the creeps, which Osama Bin Laden gives to me and, no doubt, to other folk as well. After all, the sentence “Osama Bin Laden gives me the creeps” is true. And it asserts that the three-place relation of giving holds between Bin Laden, me, and the creeps. How can the sentence “Osama Bin Laden gives me the creeps” be true, if there is no such thing as the creeps? The expression “the creeps” is a referring expression, a singular term. So there must be something for it to refer to, if the sentence “Osama Bin Laden gives me the creeps” is to be true. Having gone this far down the royal road to realism about the creeps, creeps metaphysicians might then set to work to tell us about the creeps. They might reflect that the creeps might still exist even if Bin Laden had never given the creeps to me or anybody else—indeed, might still exist even if nobody or nothing had ever given it to anybody or anything. They might wonder whether
Pleonastic Platonism
67
the creeps is an eternal or timeless entity. Or whether it can be located in space, like Osama Bin Laden or I can. Or about the action-at-a-distance that is involved when Osama gives the creeps to me. And so forth . . . Enough! Creeps realism is silly. Nobody is a creeps realist. But if there is no such thing as the creeps, must we conclude that the sentence “Osama Bin Laden gives me the creeps” is false? That seems odd, as well—not quite so odd as believing in the creeps but still odd enough. For Osama Bin Laden does make me feel nervous, a fact that I colorfully express by saying that he gives me the creeps. Are we to say that “Osama Bin Laden makes me feel nervous” is true, whereas “Osama Bin Laden gives me the creeps” is false? Must antirealists about the creeps “distort ordinary linguistic usage by treating claims that seem to be equivalent as not even having the same truth-value” (Thomasson 2001, 327)? We all know how to avoid this oddity. We all know that “Osama Bin Laden gives me the creeps” is an idiomatic way of speaking, which is not to be taken at face value for metaphysical or ontological purposes. It is true enough, to be sure, but what makes it true is the fact that Osama Bin Laden makes me feel nervous or something like that. Given this, what is the relationship between the true sentences “Osama Bin Laden gives me the creeps” and “Osama Bin Laden makes me nervous”? In what sense are they equivalent? They are not logically equivalent. They are not semantically equivalent, one cannot be obtained from the other by the substitution of synonymous expressions (as with “Osama is a bachelor” and “Osama is an unmarried man”). You distort ordinary English usage if you say that the two mean the same thing, that “Osama Bin Laden gives me the creeps” can be translated without loss as “Osama Bin Laden makes me feel nervous.” You distort it precisely by failing to see an idiom for what it is. This is no trivial distortion—idiom, metaphor, and other nonliteral uses of language help make the treasures of literature possible. The most we can say is that the two sentences are materially equivalent, they have the same truth-value. And what guarantees that they have the same truth-value is that they have the same truthmaker: what makes one true makes the other true, too. Notice that where we have idiomatic or nonliteral ways of speaking, meanings and truthmakers come apart. “It is raining cats and dogs” does not mean “It is raining heavily”—they just have the same truthmaker.1 Notice, too, that when A is an idiomatic way of saying B, the relation between them is not an equivalence relation because it is not symmetrical—if A is an idiom for B, then B is not an idiom for A. (I used to
1 Heather Dyke taught me that the idea that meanings and truthmakers can come apart is the foundation of the new B-theory of time. Contrary to the old B-theory, tensed sentences do not mean the same as tenseless sentences, they just have the same truthmakers. So contrary to the A-theory, there are no “tensed facts,” no tenses or tensed properties in reality. See Dyke (2003).
68
Alan Musgrave
speak of idiomatic equivalence, but I now think this was a mistake.) After all, it is this asymmetry that lies behind our treating them differently when we come to do metaphysics. Antirealists about the creeps who say that “Osama Bin Laden gives me the creeps” is false are guilty of the same oversight as creeps realists are. They are guilty of plonking literalism and insensitivity to idiom. They are right in thinking that “the creeps” is a referring term that does not refer to anything. They are wrong in thinking that every time we use an empty referring term, we say something false. Idiomatic uses of singular terms are not like literal uses of them. The terms “dog” and “cat” are nonempty referring terms. But nobody should think that they are being used to refer to dogs and cats—ones that are falling from the sky—whenever it is truly said “It is raining cats and dogs.” That empty terms can be used to utter truths is really quite familiar. “Santa Claus” is an empty term. But “Santa Claus does not exist” is true. And so are “I believed in Santa Claus when I was four,” “I was once delighted that Santa Claus had not forgotten me,” and so forth. Idiomatic uses of empty singular terms are similar. Creeps antirealists are right that they are not literally true, of course—just wrong that they are not true. Creeps realists are the real culprits, however. They base an absurdly bloated ontology on their plonking literalism. Talk of the creeps is but one example of idiomatic or nonliteral language use. “Jane arrived in a Porsche” requires for its truth the existence of Jane and the Porsche. Does “Jane arrived in the nick of time” similarly require for its truth the existence of Jane and the nick of time? Are there sakes, and does goodness have one? For goodness’ sake, no. Examples like these can be multiplied and, of course, nobody sensitive to the nuances, idioms, metaphors, and other riches of the English language is fooled by them. We all, somehow, refuse to take these ways of speaking seriously, from an ontological point of view. We all, somehow, fi nd a way of refusing their ludicrous ontological implications. Can it really be that platonism results from the same kind of confusion as belief in the creeps or the nick of time might result from? My uncharitable and depressing Wittgensteinian answer is “YES.”
3
PLEONASTIC TRANSFORMATIONS
Are there properties? It seems crazy to deny it. Suppose it true that Fido is a dog. Now consider “Fido has the property of being a dog.” Obviously,
I prefer to speak of “truthmakers” rather than the more familiar “truth conditions.” The latter tend to hover somewhere in between things in the world (truthmakers) and meanings—as in the idea that “Meanings are truth conditions.”
Pleonastic Platonism
69
this is true, too. The two are equivalent in some sense or other, merely different ways of saying the same thing. But the latter says that the two-place relation of having holds between Fido and a new entity, a property. So the property of being a dog exists, as well as Fido. Instead of referring to this property by means of the defi nite description “the property of being a dog,” we can invent a proper name for it if we like—hence “doghood” or “dogness.” Then metaphysicians set to work to tell us about the property of being a dog, or doghood. Some platonize. They say that doghood does not depend for its existence on its exemplification by Fido. Doghood does not depend for its existence on its exemplification by any critter at all. The property of being a dog does not exist in time, as dogs do. The property of being a dog is eternal or, rather, timeless (outside time). When dogs evolved, the property of being a dog came to be exemplified. But the property would exist even if no dogs had ever evolved to exemplify it. Nor does doghood exist in space, as dogs do. Doghood exists outside space and time. Other metaphysicians (beginning with Aristotle?) try to avoid these platonic excesses. They say that doghood does depend for its existence on dogs, so that if there were no dogs there would be no doghood either. Doghood came into existence when dogs did. (There are no uninstantiated universals.) Moreover, doghood is spatial as well as temporal—it exists wherever dogs do. But doghood is an extraordinary thing that can wholly exist in more than one place at once, whereas ordinary things like dogs cannot be in two different places at the same time. Both parties agree, however, that the property of being a dog does not depend for its existence on us, in particular, on language, on our invention of the definite description “the property of being a dog” let alone the proper name “doghood.” Yet other metaphysicians wonder about this. Perhaps the property of being a dog does, after all, depend for its existence on language, in particular, on our invention of the phrase “the property of being a dog.” Which brings me to what is called “ontological minimalism” (see Thomasson, 2001). This is the idea that the metaphysical excesses of platonism can be avoided by recognizing that the property of being a dog is somehow created by language or thought. The singular term “the property of being a dog” still refers to an entity, but that entity has a lesser metaphysical status than traditional platonic entities. In particular, the property of being a dog was created or invented when the words “the property of being a dog” were created or invented. As language evolves, it spins off an evolving platonic realm (which Popper called “World 3”). I allow that languages, and the words that they contain, are human creations, human artifacts. The idea that once we invent or create a word or phrase, we invent or create an entity for that word or phrase to stand for, is word magic. Ryle called it “Fido-Fido-ism.” It is a kind of linguistic idealism:
70 Alan Musgrave Words yield things, language creates the world (or at least, creates a world, Popper’s World 3). So-called “ontological minimalists” describe entities like the property of being a dog not just as “minimal” but also as “pleonastic.” I had to look this up. In my dictionary a pleonasm is defi ned as “the use of more words than are needed to give the sense,” and to speak pleonastically is to “be redundant.” So it is even worse than the usual word magic, which consists in assuming that whenever we invent a word there must be a thing for it to stand for. It is the idea that by using more words than are needed to give the sense, by using redundant words, you can create an entity of some kind (albeit not an eternal platonic entity). It is bad enough to assume that words make things—it is even worse to assume that redundant ones do. Bloated language yields a bloated ontology. There is nothing minimal, or even sparse, about it from a metaphysical point of view. Rather than call it “ontological minimalism,” I prefer to call it pleonastic platonism. Why should we countenance pleonastic entities? Schiffer (1996, 151– 152) argues that to deny that any singular term refers has unpalatable consequences. What exactly are these? Reference failure views are said to “overthrow . . . our ordinary usage” (Thomasson 2001, 330 n. 8). They “distort ordinary linguistic usage by treating claims that seem to be equivalent as not even having the same truth-value” (Thomasson 2001, 327). In fact, what distorts ordinary English usage is insisting that every use of a singular term is a referring use. Remember the creeps—or rather, remember the singular term “the creeps.” Since “the creeps” is a singular term, if we eschew creeps realism, don’t we have to say that “Osama Bin Laden gives me the creeps” is false? No. We can say that it is just an idiomatic way of saying that Osama Bin Laden makes me feel nervous. It is not to be taken at face value; in particular, its ostensibly referring term “the creeps” is not to be taken as a referring term. This is not to distort ordinary English usage, rather it is to be sensitive to it. What about the property of being a dog, where this is viewed as a “pleonastic entity”? Why not view “Fido has the property of being a dog” as a long-winded, idiomatic, or at least nonliteral way of saying “Fido is a dog”? The two are not logically equivalent, nor are they equivalent in meaning. They are materially equivalent, because they have the same truthmaker. Or perhaps better, they are factually equivalent because they are both made true by the same fact. The fact that Fido is a dog makes “Fido is a dog” true—and it also makes “Fido has the property of being a dog” true (not to mention “Fido exemplifies doghood” as well). So far the only pleonastic entities I have considered are properties. There are others.2 There are propositions, for example the proposition obtained
2 Such is the richness of the English language that there are heaps of others. Here are a few alternative pleonastic equivalents of “Fido is a dog”:
Pleonastic Platonism
71
by transforming “Fido is a dog” pleonastically into “[The proposition] That Fido is a dog is true.” There are events, for example, the event got from transforming “Fido bit Fi-Fi” pleonastically into “[The event of] Fido’s biting Fi-Fi occurred.” There are states, for example, the state gotten from transforming “Fido was angry” pleonastically into “[The state of] Being angry was occupied by Fido.” And of course, there are facts as well, for example, the fact got from transforming “Fido is a dog” pleonastically into “It is a fact that Fido is a dog.” I reject pleonastic platonism. I do not believe in pleonastic entities. Such entities are said to have the following features: 1. They are in some sense language-created. 2. The terms referring to them are guaranteed of reference. 3. To learn of their existence and all there is to know about them, one need only adopt the relevant language game. (Thomasson 2001, 320) Against this, I suggest that no entities other than linguistic ones (for instance, words or sentences) are language-created, that no term is guaranteed of reference, and that adopting a language game does not tell us anything about anything, let alone tell us all there is to know about anything. In support of the idea that pleonastic entities are language-created, Schiffer writes: A trivial transformation takes one from a sentence in which no reference is made to a property to a sentence that evidently contains a singular term whose referent is a property. Thus, from “Fido is a dog,” whose only singular term is “Fido,” we can infer its pleonastic equivalent “Fido has the property of being a dog” wherein the ostensible singular term “the property of being a dog” evidently refers to the property of being a dog. (Schiffer 1996, 149) In support of guaranteed reference, Schiffer writes:
Fido is a dog if and only if Fido has the property of being a dog Fido is a dog if and only if Fido displays the feature of being a dog Fido is a dog if and only if Fido manifests the characteristic of being a dog Fido is a dog if and only if Fido possesses the trait of being a dog Etc. Are properties, features, characteristics, and traits all the same things? Or will analytic metaphysicians squabble about whether the metaphysics of having properties is to be preferred to that of manifesting characteristics?
72 Alan Musgrave The sentence “Fido is a dog,” whether or not it is true, also yields the singular term “the property of being a dog,” which we are assured of referring to the property of being a dog. (Schiffer 1994, 304) In support of complete and a priori knowledge, we are told that whereas to learn about things like dogs we must undertake substantive investigations into the things themselves, to learn all there is to know about things like the property of being a dog “one must only study the language games by means of which they are deposited in our ontology” (Thomasson 2001, 321; see also Schiffer 1996, 159). Such pleonastic entities have “no hidden and substantial nature for a theory to uncover” (Johnston 1988, 38). Do any entities have these three features—language-dependence, guaranteed reference for their singular terms, and complete knowability? In an interesting discussion, Amie Thomasson distinguishes three kinds of “ontological minimalism” (Thomasson 2001, 323–325): 1. An absolutely minimal entity has its existence guaranteed by pleonastic transformation of a sentence whether or not that sentence is true. Examples are properties and propositions. 2. A relatively minimal entity has its existence guaranteed by pleonastic transformation of a sentence provided that sentence is true. Examples are events and states. 3. A linguistically minimal entity has its existence guaranteed by the fact that certain sentences are used in special ways. Examples are fictional characters introduced by fictional uses of language (however these are to be analysed). I fi rst discuss fictional characters, Thomasson’s paradigm case of “linguistically minimal entities.” I argue that fictional characters are not pleonastic entities at all and that pleonastic platonism cannot acquire innocence by association with them. I next argue that Thomasson’s distinction between absolutely and relatively minimal entities is quite unmotivated: pleonastic entities, if there are any, are all absolutely minimal in Thomasson’s sense. Finally, I ask whether we should accept that any pleonastic entity exists. 4
FICTIONAL CHARACTERS
I assume, for the purposes of this discussion, that fictional characters exist. They are social constructs, like money or the rule of the road. They are brought into existence by human activity, specifically by the activity of story telling, if you like, by the language game of fiction. So I accept that the fictional character Sherlock Holmes (to use the favorite example) is, in some sense or other, a linguistic creation.
Pleonastic Platonism
73
But we must carefully distinguish the expressions “Sherlock Holmes” and “the fictional character Sherlock Holmes.” The name “Sherlock Holmes” is an empty name, Sherlock Holmes does not exist, there is no person of that name. But the description “the fictional character Sherlock Holmes” is not empty; it refers (we are assuming) to an entity created by Conan Doyle’s stories, to some sort of social construct. This description has an empty name as a syntactic part of it, but the description is not itself empty. So far we have a defi nite description for the fictional character Sherlock Holmes but no name for it. But, of course, we might coin a name for it as we might for any entity. The standard practice, so as to avoid wasting words, is to call it by the name “Sherlock Holmes.” So we standardly use the same name for two different things, a nonexistent person and an existent fictional character, much as we name ships and buildings after people. As with ships and buildings that have the same names as people, indeed, as with different people with the same name, we must rely on context to resolve ambiguities. Schiffer writes (using a different example): What is remarkable is that this [James Joyce’s] pretending use of the name “Buck Mulligan” should create the existence of something whose name is “Buck Mulligan,” thereby making it possible to use the name in a genuinely referential way in true statements about that referent. The thing brought into existence is a certain abstract entity, the fictional character Buck Mulligan. (Schiffer 2003, 50–51) This remarkable word-magical feat is a chimera, brought about by using the same name for two different things. The description “the fictional character Sherlock Holmes” is not brought about by a pleonastic transformation, and the fi ctional character Sherlock Holmes to which it refers is not a pleonastic entity. “The fictional character Sherlock Holmes smoked a pipe” is not a pleonasm for “Sherlock Holmes smoked a pipe,” as the former is true and the latter false. Of course, the true sentence “The fi ctional character Sherlock Holmes smoked a pipe” nowhere occurs in the Holmes stories. You betray a plonking misunderstanding of the entire language game of fiction if you think that anything said in the Holmes stories about Sherlock Holmes could just as well be pleonastically transformed into a statement about the fictional character Sherlock Holmes. Conan Doyle would insist that the Holmes stories are not about the fictional character Sherlock Holmes but about Sherlock Holmes! Of course, a different fictional character might have figured in a Holmes story. One can tell stories in which storytelling figures, just as one can paint pictures in which pictures figure. In Northanger Abbey fictional characters from The Mysteries of Udolpho are mentioned, not by descriptions of the form “the fi ctional character
74 Alan Musgrave X,” but by the same names that nonexistent people bear in The Mysteries of Udolpho! But nobody is confused. Fictional characters are pretty “abstract,” like other social or cultural constructs, but they have their own robust criteria of reality. We cannot “pleonastically transform” any name into a description (or name) of a fictional character as well—at least, not salva veritate. Whereas “Sherlock Holmes” is an empty name and “the fictional character Sherlock Holmes” is not an empty description, it is the other way round with the name “Alan Musgrave” and the description “the fictional character Alan Musgrave.” The name refers to me, but the description is empty, because there are no fictions about me! (Bill Bartley once told me that there is an awful fi lm, Lord Love a Duck, with an Alan Musgrave in it who lectures in California. But the Alan Musgrave in Lord Love a Duck is not me.) Schiffer is quite right that once fictional characters have been brought into existence, many true statements can be made about them. His examples are (2003, 51; the numbering is mine): I. Joyce created the fictional character Buck Mulligan. II. Buck Mulligan isn’t as well known as certain of Joyce’s other characters, such as Molly Bloom. III. Buck Mulligan isn’t a fictional detective. IV. Buck Mulligan was based on someone Joyce knew. In I the name “Buck Mulligan” appears only as a syntactic part of the definite description “the fictional character Buck Mulligan.” In that context, the name is not being used referentially at all—it refers neither to a person nor to a fictional character. In II–IV it is clear from the context that the name “Buck Mulligan” is being used as short for “the fictional character Buck Mulligan.” Fictional characters are not brought about by pleonastic transformation, as has been explained. Nor have they another alleged feature of pleonastic entities, that their “essential nature” can be uncovered a priori, by armchair perusal of our talk about them—that all there is to know about them is determined by our use of fictional names, so that the study of this “language game” is all that we need to undertake. This is quite false. Consider Schiffer’s own list I–IV of truths about the fictional character Buck Mulligan. Not one of these truths is accessible from a philosopher’s armchair. They all need the attentions of literary historians and/or sociologists of literature. And we have to read the stories to learn that Sherlock Holmes is (in the Holmes stories) a detective or that Mr. Bennett in Pride and Prejudice is not a very sympathetic character. Fictional characters are not timeless platonic entities. Once we accept that Conan Doyle created the fictional character Sherlock Holmes by writing the stories, we also accept that the fictional character Sherlock Holmes did not exist before Conan Doyle wrote the stories. Nor do fictional characters exist
Pleonastic Platonism
75
in all “possible worlds.” Conan Doyle might not have written the Holmes stories, in which case the fictional character Sherlock Holmes would not have existed at all. Finally, fictional characters lack another alleged feature of pleonastic entities, that they do no work in the world, make no causal differences, “come softly into existence, without disturbing the preexisting causal order in any way” (Schiffer 2003, 59). Goodness me, fictional characters can frighten people or make their creators a lot of money! Fictional discourse is an important cultural invention, which does a great deal of work in the world. Think how the lives of children would be changed, if they were never told stories. Schiffer might respond that it is the story telling, the linguistic practice, the “language game,” that does the work, not the fictional characters introduced by it. If so, we need not postulate fictional characters to explain how fiction works in the world, and they are idle posits that should be eliminated. I have been assuming that the best “theory of fiction” has fictional characters as its theoretical entities. That is why I have been granting that fictional characters exist—in much the same way, or for the same kind of reason, that scientific realists grant that electrons exist. If a better “theory of fiction” can be worked out that does not posit them, so be it. There are interesting and formidable problems in working out a theory or theories of fiction. It is work for scientists—historians, psychologists, literary theorists, sociologists. It is up to the literary scientists who posit fictional characters as their theoretical entities to tell us what kind of entity they are and to explain how they work. Fictional characters viewed in this way are neither pleonastic nor platonic entities. They are, then, in the present context, a red herring, and I shall say no more about them. I actually think that the fictions or the stories exist, but the fictional characters introduced by them do not. That is because I am attracted by a kind of if-thenist or postulationist or “disguised inference” view of fictional truth. This says, roughly, that S is true-in-the-fiction F if and only if the conditional “If F then S” is (logically) true. What is F? It consists of the things stated in the fiction F, together with anything else that readers of the fiction F take for granted. So it is true-in-the-fiction that Sherlock Holmes smoked a pipe, and it is also true-in-the-fiction that he had two legs. The former is explicitly stated in the Holmes stories, the latter implied by common knowledge that readers bring to the Holmes stories (that Holmes is a human being, that human beings have two legs).
5
PLEONASTIC ENTITIES
Let us get back to pleonastic entities proper and to Thomasson’s distinction between absolutely and relatively minimal pleonastic entities. I fi nd this distinction problematic. Absolutely minimal entities are supposed to have
76 Alan Musgrave their existence guaranteed by sentences whether or not those sentences are true, whereas relatively minimal entities require true sentences for their existence to be guaranteed. “Fido is a dog” is supposed to generate the property of being a dog whether or not it is true—if it is true the property is exemplified by Fido and if it is not true the property is not exemplified by Fido but exists nonetheless. Similarly with propositions. “Fido is a dog” generates the proposition that Fido is a dog by being transformed pleonastically into “[The proposition] That Fido is a dog is true.” If “Fido is a dog” happens to be false, the proposition that Fido is a dog exists nonetheless and happens to be false, too. Thomasson thinks it otherwise with events and states, which require true sentences for their existence to be guaranteed. But why does “Fido bit Fi-Fi” not generate the event of Fido’s biting Fi-Fi whether or not it is true? If Fido did not bite Fi-Fi, then the event of Fido’s biting Fi-Fi did not occur. Similarly, if pleonastic transformation of “Fido was angry” into “Fido occupied the state of being angry” generates a state, does this state not still exist—unoccupied by Fido—if the sentence happens to be false? If there are false propositions and uninstantiated properties, why are there not unoccupied states and events that do not occur? Thomasson’s distinction between absolutely and relatively minimal entities is quite unmotivated. All pleonastic entities are “absolutely minimal” in Thomasson’s sense: properties, propositions, events, states, and so forth. There is a more important point, which the so-called “ontological minimalists” themselves see. It concerns the alleged language-dependence of pleonastic entities. Such things as properties, propositions, events, states, and so forth, do not depend upon language for their existence. All such things, if they exist, existed long before people evolved and started playing language games. All such things, if they exist, are platonic entities that do not depend upon us and our linguistic doings for their existence. (In this they are to be contrasted with fictional characters—if there were no fictions, there would be no fictional characters—remember that “the fictional character Alan Musgrave” is an empty description!) You cannot soften the ontological excesses of platonism by talking of pleonastic transformations. The fi rst alleged characteristic of pleonastic entities, that they are “in some sense language created,” is a complete nonstarter. (The same goes for Popper’s evolving “World 3.” We cannot create platonic entities, and the world they inhabit does not evolve.) The third alleged characteristic of pleonastic entities was that “to learn of their existence and all there is to know about them one need only adopt the relevant language game.” This is an epistemic characteristic. It represents the chief virtue of pleonastic platonism. If pleonastic entities have this remarkable epistemic feature—that playing the relevant language game assures us that they exist and tells us all there is to know about them—then pleonastic entities avoid the chief worry about platonic entities. This is Paul Benaceraff’s epistemic worry. It is
Pleonastic Platonism
77
plausible to think that we can only fi nd out that something exists if we can somehow interact with it, however indirectly. But if platonic objects are not spatiotemporal and have no causal powers, then how on earth do we get to know that they exist or fi nd out what they are like? “No problem,” says the pleonastic platonist—“All you need to do to acquire this knowledge is to learn how to play the relevant language games.” Platonic knowledge, on this view, is a priori knowledge. It requires no causal interaction, however remote, with its objects. It is a priori in much the same way that knowledge of analytic truths is a priori. You do not have to inspect bachelors, or causally interact with them in any way at all, to know that they are all unmarried. All you have to do is learn that the word “bachelor” is short for “unmarried man.” What are we to make of this? On the one hand, the entities are language-independent—they do not depend on the evolution of language for their existence; they did not pop into existence when folk started playing language games. On the other hand, playing language games assures us that they exist and tells us everything there is to know about them. This really is word magic. Platonic knowledge appears as the rabbit appears from the word-magician’s hat. We play magical “something-from-nothing” language games (Schiffer)—“though not a true ontological free lunch, it is perhaps a case of ‘buy one, get one free’” (Thomasson 2001, 324). As for causal objections to Platonism, they are misguided. Alexander’s dictum that to be real is to have causal powers is misplaced when it comes to pleonastic entities: Schiffer recently . . . proposed that a necessary condition for a concept F securing the existence of Fs as pleonastic entities is (roughly) that introducing Fs brings in no new causal consequences over those that would be entailed without Fs. This fits well with the idea that, in the case of pleonastic entities, we should expect no new causal consequences; the lack of such causal additions is thus no argument against positing them. (Thomasson 2001, 331 n. 11) But it is not a matter of positing these entities, as scientists posit theoretical entities in the hope of explaining phenomena, while always admitting that their posits might turn out not to exist. Pleonastic entities do no explanatory work, because they do no causal work. Nor can it turn out that pleonastic posits do not exist or lack some feature attributed to them. We are supposed to know that they exist and to know all there is to know about them. I think this platonic knowledge is a sham, because its objects—pleonastic entities—are a sham. Any object that we posit or postulate might turn out not to exist. Any name or description that we invent might turn out to be empty. An object that cannot fail to exist is no genuine object at all, for
78
Alan Musgrave
there are no necessary existents. Platonic realism is to be contrasted with scientific realism. Scientific posits may be unobservable and “abstract” in various ways. But they are all explanatory and causal, and none of them is a “necessary existent.” In fact, all that we have here are long-winded ways of saying things about what does exist. We need not say that these long-winded pleonastic ways of saying things are false. If it is true that Fido is a dog, then it is also true (if you like) that Fido has the property of being a dog. But this longwinded, pleonastic way of putting it is like an idiomatic way of putting something—its truth does not require the existence of “pleonastic entities,” for what makes it true is just the fact that Fido is a dog. Have I not just committed myself to facts? No, for “It is a fact that Fido is a dog” is just another long-winded way of saying “Fido is a dog.” Nor am I saying that language should be purged of pleonasms, or of idioms, or of metaphors, or of any other nonliteral ways of saying things. Not only do nonliteral ways of speaking help make the treasures of literature possible, as I already said they also make life easier for all of us, in all sorts of ways. It is almost impossible to avoid platonist ways of speaking and apparent reference to platonic entities. All I advocate is that we do not “read off” platonist metaphysics from these conveniences, perhaps necessities, of language. Advocates of the tenseless theory of time do not say that we can, or should, eliminate tensed idioms from language. After all, tensed idioms reflect our temporal perspective on tenseless reality. But the ineliminability, without loss of meaning, of tensed idioms does not mean that tenses or tensed properties exist out there in the world.
6 PLATO’S HEAVEN—A GRAVEYARD FOR LOST IDIOMS AND DEAD METAPHORS? Everybody accepts that sentences like “Osama Bin Laden gives me the creeps” or “George Dubbya invaded Iraq in the nick of time” are idioms, not to be taken literally or at face value when we are doing metaphysics. Could it be that the belief in a platonic realm of eternal, nonspatial, acausal necessarily existing entities arises from lost idioms and dead metaphors? Or, if that seems implausible, could it be that platonism arises from longwinded or pleonastic ways of talking that get taken more seriously than they should by metaphysicians? Or, if “long-winded” is implausible as a general rule, from alternative ways of talking that get taken more seriously than they should by metaphysicians? My caution about “long-winded” being a general rule is prompted by cases where an “abbreviatory defi nition” is introduced, and many words are replaced by few. Suppose “bachelor” is short for “unmarried man.” Still, nobody will suppose that “Osama is a bachelor” says anything different from “Osama is an unmarried man,” or that the shorter form of words should be taken ontologically seriously, or
Pleonastic Platonism
79
that Osama has two properties, the property of being a bachelor and the property of being an unmarried man. Quine famously demanded “No entity without identity”—we should not postulate entities without giving identity-conditions for them. His example, as I recall it, was this. We start with “A tall person might be standing in the doorway” and “A clever person might be standing in the doorway.” We move to “It is possible that a tall person is standing in the doorway” and “It is possible that a clever person is standing in the doorway.” Then we move to “A possible tall person is standing in the doorway” and “A possible clever person is standing in the doorway.” Whereupon Quine asks “Is the possible tall person standing in the doorway identical with the possible clever person standing in the doorway?” We might also ask “How many possible people can stand in the doorway?”—which is worse, when you think about it, than “How many angels can dance on the head of a pin?”! Unfortunately, Quine’s “No entity without identity” is not enough to rule out abstract or platonic entities. There is a way of arriving by pleonastic transformation at abstract entities that do automatically satisfy Quine’s demand—it is called, appropriately enough, abstraction. We start from “concrete” or nonplatonic entities and from an equivalence relation between them—that is, from sentences saying that these entities are in some respect identical with one another. Then we abstract on those equivalence claims to obtain abstract entities or, more precisely, sentences that refer to such entities. Some examples will make the process clear: The direction of line A is identical to the direction of line B if and only if line A is parallel to line B. The shape of A is identical to the shape of B if and only if A is geometrically similar to B. The speed of A is identical to the speed of B if and only if A moves as fast as B. The weight of A is identical to the weight of B if and only if A weighs as much as B. The age of A is identical to the age of B if and only if A has lived as long as B. The meaning of A is identical to the meaning of B if and only if A is synonymous with B. The set of As is identical to the set of Bs if and only if As are exactly the same things as Bs. The number of As is identical to the number of Bs if and only if there are exactly as many As as Bs. And so on, and so forth. In these abstraction schemes A and B stand for the “concrete” entities. On the right-hand sides we have sentences asserting that an equivalence relation holds between concretes—that is, that they are identical in some
80 Alan Musgrave respect. On the left-hand sides defi nite descriptions of new entities (directions, shapes, speeds, weights, ages, meanings, sets, numbers) are introduced. Notice that abstraction schemes state identity-conditions for the new entities that they introduce. Quine was suspicious of meanings, accepted the abstraction scheme for them, and became suspicious of synonymy as well. That was, I think, a mistake. Synonymous expressions exist and so do analytic truths generated by their existence. We are not committed to platonic meanings in saying this—provided we view the abstraction scheme as just a pleonasm and do not take it seriously from an ontological point of view. Are the entities introduced by abstraction schemes (directions, shapes, speeds, weights, ages, meanings, sets, numbers, and so forth) really abstract or platonic entities? It seems so. After all, are there not directions that no line has, speeds that nothing has moved at, weights that nothing has, ages that nobody has lived to, meanings possessed by no word, and so forth? Indeed, wouldn’t there be directions if there were no lines, shapes if nothing had one, speeds if nothing moved, ages if nobody had ever been born, meanings if there were no words, and so forth? The way of the world is not independent of our sayings and other doings because our sayings and doings make a difference to the world. But the way of the platonic world is surely independent of our sayings and doings. On the other hand, there are all these “concrete” things (lines with directions, objects with shapes, etc). And because they exist, we can acquire a priori necessary knowledge of platonic entities introduced by abstraction schemes. There is always an equivalence relation on the right-hand side. Equivalence relations are always reflexive. We know a priori that every line is parallel to itself, that every object is geometrically similar to itself, that everything moves as fast as or weighs as much as or has lived as long as itself, that words are synonymous with themselves, that exactly the same things are As as are As, and that there are as many As as there are As. So if there are lines, physical objects, people, words, we know a priori that there are also directions, shapes, speeds, weights, ages, meanings, sets, and numbers. We do not always need an abstraction scheme, with an equivalence relation on the right-hand side, to get abstract entities. Who can deny that there are colors? After all, red is a color, as we know from the undeniable truth “Red is a color if and only if everything red is colored.” The right-hand side is true if nothing is red. From which it follows that red would still be a color, even if nothing were red. Colors are platonic objects, too. What are we to make of abstraction schemes? Crispin Wright thinks they are conceptual truths and that it is conceptually incoherent to deny any of them: Now, how exactly do these reflections illuminate the sense of incoherence which the ideas that a person might lack an age, a geometric figure a shape, a fi nitely instantiated concept a number of instances, etc.
Pleonastic Platonism
81
are apt to inspire? Like this, I think. When the concepts in question are explained as they are explained: when that is, the obtaining of an equivalence relation between a pair of things of some sort is stipulated as necessary and sufficient for the truth of identity-statements under those concepts, then it is a straightforward consequence of the explanation, coupled with the reflexivity of the relation of the definiens, that, for example, any particular fi nitely instantiated concept has a number . . . The suggestion that a fi nitely instantiated concept might lack a number . . . is thus in confl ict with the form of explanation which abstract sortal concepts of this family receive. . . . Hence the sense of incoherence which such suggestions inspire in us: for the suggestions sound like expressions of scepticism; and no such scepticism requiring recourse to the relevant concepts for its very formulation can be anything but incoherent. (Wright 1983, 151–152) But any expression of scepticism about Xs, for any X, requires recourse to the X-concept. I cannot say that there are no flying saucers without employing the flying saucer concept. Wright must be aware of this, despite what he says. Perhaps he thinks there is something special about concepts introduced by way of abstraction schemes? After all, his rehabilitation of Frege’s Conception of Numbers as Objects rests entirely on the above abstraction scheme for natural numbers, now christened “Hume’s Principle.” Perhaps it is only when we have an abstraction scheme that we can be word magicians and pull metaphysical rabbits out of linguistic hats? How might a skeptic about directions, say, proceed? A skeptic about directions is a person who denies that there are such things, while accepting that there are lines and that some of them are parallel to others. There are two ways for the skeptic to proceed. One is to bite the bullet and say that since there are no directions, definite descriptions of the form “the direction of line X” are always empty. That being so, the left-hand side of the abstraction scheme for directions is always false. But the right-hand side of the abstraction scheme might be true (the skeptic about directions allows this, remember). Thus the abstraction scheme itself is false. Because it is false, it is not a conceptual truth, and it is not something we know a priori. I do not think, pace Wright, that there is anything incoherent about this position. It is the bullet-biting position which Hartry Field, a skeptic about numbers, adopts in the face of the abstraction scheme for numbers (see Musgrave 1986). It is the bullet-biting position that we all might adopt when it comes to the creeps scheme “A gives B the creeps if and only if A makes B feel nervous.” But as the example of the creeps makes clear, there is another, softer way for the skeptic to go. An antirealist or skeptic about the creeps can accept the creeps scheme but only as a codification of an idiomatic or nonliteral way of speaking that is not to be taken at face value for metaphysical purposes. Similarly, the
82 Alan Musgrave skeptic about platonic directions, shapes, speeds, weights, ages, meanings, sets, numbers, and so forth can accept the appropriate abstraction scheme but only as codifying an idiomatic or nonliteral way of speaking that is not to be taken at face value for metaphysical purposes. Thus, for example, the skeptic about numbers can cheerfully accept that the number of fi ngers on my right hand is five, without thereby being committed to the existence of the number five. The skeptic about numbers will just regard “The number of fi ngers on my right hand is five” as an idiomatic or pleonastic or longwinded or nonliteral way of saying “There are five fi ngers on my right hand.” In this way the antiplatonist can “speak with the learned but think with the vulgar” (to invert Berkeley’s unhappy idealist phrase into a happy realist one). David Stove used to say that half of the bad philosophy stemmed from use/mention confusions. It is use/mention confusions that spawn wordmagical philosophy in general (see Musgrave 2001). And use/mention confusions need to be avoided here, too. How can abstraction schemes codify idioms and be accepted as such, as I just said they could? It is words and phrases that are idiomatic, nonliteral, long-winded, or pleonastic. No words or phrases are spoken of in abstraction schemes. An explicit codification of the creeps idiom is not the creeps scheme “A gives B the creeps if and only if A makes B feel nervous” but, rather, “A sentence of the form ‘A gives B the creeps’ is an idiom for a sentence of the form ‘A makes B feel nervous.’” Similarly, what records the pleonasm that gets us to numbers is not the numerical abstraction scheme “The number of As is identical to the number of Bs if and only if there are exactly as many As as Bs,” but rather “A sentence of the form ‘The number of As is identical to the number of Bs’ is a pleonasm for a sentence of the form ‘There are exactly as many As as Bs.’” Once we avoid use/mention confusions, the softer way with abstract platonic objects comes to this. We reject the abstraction scheme for such objects as false. We accept the corresponding metalinguistic scheme as true. (The metalinguistic scheme does not have the false ontological implications of the abstraction scheme because you cannot existentially generalize on referring terms that occur within quotation marks.) Accepting the latter as true explains why two sentences that do not mean the same have the same truthmaker. It also explains why we can speak with the learned while thinking with the vulgar. And it enables us to deplore the Platonist tendency to confuse use and mention, run the two together, and pull metaphysical rabbits out of linguistic hats. This is all old hat. Abe Lincoln, a great metaphysician, said it all. At the risk of boring you with this familiar story, let me remind you of it: “How many legs does this donkey have?” “Four, Mr. President.” “And how many tails does this donkey have?”
Pleonastic Platonism
83
“One, Mr. President.” “And if tails as well as legs were called ‘legs’, how many legs would the donkey have then?” “Five, Mr. President.” “No, Sir, you cannot make a tail into a leg by calling it one!” Every time I try this out on an unsuspecting audience, about half the people answer “Five” to Abe Lincoln’s last question. It is a trick, of course. They confuse Lincoln’s question “ . . . how many legs would the donkey have then?”, in which the word “leg” is used, with the question “ . . . how many [things called] ‘legs’ would the donkey have then?”, in which the word “leg” is mentioned. They agree with Derrida-Da that all we can ever talk about are words. But Derrida-Da recently died, thank goodness, and we must not speak ill of the dead.
7
DISCLAIMER
I have not established that there are no platonic entities. Indeed, given their (alleged) nature, how could one establish this? All I have tried to do is to dispose of a pervasive line of thinking that leads people to suppose that there are platonic entities. That line of thinking is to read (platonist) metaphysics off language or off our language games. I have not established that there are no properties, propositions, events, states, or whatever, when these things are NOT viewed as platonic entities. I have not, in other words, objected to “realist” theories of properties, propositions, or whatever, where the realism is scientific rather than platonic. How does scientific realism differ from platonic realism? Scientists postulate theoretical entities of various kinds to help them explain the world. To do that, they invent descriptions and even names of their theoretical posits. But scientific realists admit that any theoretical description or name they invent might turn out to be empty. Scientists invented the description “the intra-Mercurial planet responsible for the anomalous perturbations in the orbit of Mercury” and even the proper name “Vulcan.” But this was no mere “language game,” and it turned out that both the description and the name were empty. They stood for nothing in the world. They might have stood for something—the theories containing them were not self-contradictory. At which point platonistic philosophers will say they did stand for something in another possible world than the actual one—and off we go again! After all, who can deny the truth of “There is a possible world in which P if and only if possibly P”? But scientists are interested in the world, not in “possible worlds.” So I have not objected to a realist theory of properties, say. Such a theory would have to allow that expressions like “the property of being a dog” or “doghood” might stand for nothing and that it depends on the way the
84
Alan Musgrave
world is whether any such expression stands for anything. The difficulty for scientific realists about properties is to explain how fi nding out more about doghood differs from fi nding out more about dogs. In particular, having established that there are dogs, how do we go about establishing the further point that doghood exists as well? (If scientific realism is another name for what is called in philosophy “nominalism,” then so be it.)
6
Anomalous Monism John Heil
But let the wise be warned against too great readiness at explanation: it multiplies the sources of mistake, lengthening the sum for reckoners sure to go wrong. (George Eliot, Middlemarch, ch. XLV)
1
LINGUISTICIZED METAPHYSICS
For much of the twentieth century, metaphysics was held in disdain by a majority of analytic philosophers. Old questions about reality gave way to questions about talk about reality: philosophy took the linguistic turn. The linguistic turn was the product of a philosophical Weltanschauung that encompassed Wittgenstein (early and late), Carnap, Quine, Ryle, Austin, and their many disciples. The movement waned, but its influence has lingered in the practices of analytic philosophers, even those who would regard themselves as opponents of old-style linguistic philosophy. One of these influences can be observed in a persisting tendency to conflate the linguistic and the nonlinguistic. Philosophers move unselfconsciously from observations about descriptions of the world to claims about features of the world itself. Discussion of predicates, for instance, or at least predicates concerning which we are avowed realists, goes proxy for talk of properties. Indeed the thought that predicates might fail to correspond oneone to properties yet nevertheless be used to express truths about genuine, causally operative features of the world is widely taken to be incoherent. When confronted with a predicate, “is in pain,” for instance, that apparently fails to designate a unique physical property shared by everything to which it truly applies (and in virtue of which it applies), we seem compelled to choose from among three equally unpalatable alternatives: I Dualism; pains are states of immaterial substances. II Eliminativism; nothing corresponds to the pain predicate; nothing is in pain. III Reduction; talk of pain can be reduced to, hence replaced by, talk of various physical goings-on. Philosophers are nothing if not resourceful, however. Thanks to the efforts of Hilary Putnam (1967) and Jerry Fodor (1968), among others, we
86
John Heil
have learned that dualism, eliminativism, and reductionism do not exhaust the possibilities. The pain property is a genuine property (so eliminativism is false). Pains are not physical properties, however (reductionism is false). Rather, the pain property is a “higher-level” property “realized by” physical properties (a dualism of substances is false). One and the same higher-level property could have indefi nitely many “lower-level” realizers. The maneuver enables us to account for the failure of reductionist programs while leaving intact the idea that significant predicates stand in a one-one relation to properties.1 Higher-level properties are somehow dependent on and constrained by, but nevertheless distinct from, their lower-level realizers. It is one thing to countenance descriptive, or explanatory, or compositional levels; however, quite another matter to parlay this into a serious ontology encompassing irreducible levels of being. Many philosophers will go with Quine here. We find ourselves deploying certain predicates in “quantifying over” individuals. This apparently commits us to properties corresponding to those predicates. If you want to know what there is, just look at predicates figuring in favored descriptions of what there is. Ontology is a piece of cake.2 This is an exaggeration, of course. Most of the time property talk is not undertaken with full ontological seriousness. Properties are just whatever answers to favored predicates: reference to properties is ontologically innocent. We can speak, for instance, as the Episcopal Prayer Book speaks, of God, “whose property is always to have mercy.” Apparent reference to a property here might be read as a mildly pretentious way of asserting that God is always merciful. In these contexts, property ascriptions could be interpreted as affirmations of commitments to the truth of particular kinds of predication. This is all well and good, until we move from unguarded talk of properties directly to more substantive theses. We can find ourselves encumbered with unfortunate conceptions of what properties there are and how these properties are related. The tendency is abetted by the implicit idea that realism about a given domain requires something like a one-one correspondence between predicates in that domain and properties. If we are to be ontologically serious about properties, let us be ontologically serious from start to finish. It is a bad idea to begin with a relaxed, linguisticized conception of what properties there are, then, from this dodgy foundation, extract substantive ontological conclusions about relations these
1 Antipathy towards reductionism stems, I believe, from a prior commitment to the idea that properties line up one-one with significant predicates. This leads to a hierarchical conception of being, which can then be recruited to buttress antireductionist sentiments. 2 Or is it? Quantum theory affords a striking instance of the gap between accepting a theory as true and uncovering the nature of the theory’s truthmakers. Even among physicists who accept quantum theory as true, there is little agreement over what the world must be like if it is true.
Anomalous Monism
87
properties bear to one another and their roles in causal interactions. As Wittgenstein notes, “the first step is the one that altogether escapes notice. . . . The decisive movement in the conjuring trick has been made, and it was the very one that we thought quite innocent” (Wittgenstein 1953, §308).
2
PREDICATES TO PROPERTIES
Putnam’s original defense of functionalism was titled “Psychological Predicates.” Subsequently Putnam rechristened the piece “The Nature of Mental States,” the title under which it has been most frequently anthologized and cited. The move nicely illustrates the unselfconscious ease with which analytic philosophers can slide from consideration of predicates to ontologically loaded talk of properties. Putnam’s arguments provided an important insight into our deployment of psychological terms. Such terms evidently apply truly to many different kinds of actual and possible creature but not in virtue of those creatures’ shared, first-order properties. Instead of concluding that psychological terms are multiply satisfiable, however, we hold fast to the idea that significant predicates must designate unique properties. We are moved to postulate tailor-made higher-level properties that are shared by diverse creatures. These properties are distinct from, but “realized by,” sundry first-order properties. Higher-level, multiply realized properties have posed numerous challenges for their proponents. The nature of the realizing relation has, for instance, remained a mystery. If the realization relation is a genuine relation, must it be multiple? Might there be properties that could have only a single realizer? The idea seems crazy, but why? What of the “causal relevance” of higher-level properties? Available options—epiphenomenalism, overdetermination, “topdown” causation—are all, in one way or another, unattractive. Could it be that we have been barking up the wrong tree? Perhaps we have erred in our initial assumption: our conviction that the only way to take psychological (or biological, sociological, meteorological) predicates seriously is by assuming that such predicates are aligned with distinct families of higherlevel properties. What are the alternatives? Substance dualism and eliminativism are unpromising. Are we left with some form of property dualism? Perhaps not. Suppose we go part of the way with the property dualists. There are properties, all right, properties that figure as truthmakers for what might reasonably be called higher-level predications. Properties answer to such predicates, however, not by lining up with the predicates one-one, but by constituting families of similar properties. The pain predicate, for instance, might be satisfied, not by a single property, being in pain, but by any of a family of similar properties. Think of these properties as including all the different ways of being in pain. If functionalism is correct, the properties will be causally similar; if functionalism tells only part of the story, qualitative similarity will be important as well. In this
88
John Heil
conception, you and an octopus are both in pain, not because you share a single higher-level property, the pain property, but because you and the octopus possess causally, or qualitatively, or perhaps causally and qualitatively, similar properties. My aim here is not to defend this thesis, however (see Heil 2003). My focus, rather, will be historical. In particular, I am interested in how an argument that might naturally have been seen as advancing a thesis starkly at odds with the prevailing conception of mental predicates as names for higher-level, multiply realized properties came to be regarded as a pillar of “non-reductive physicalism,” a conception of mental properties as dependent on, but distinct from physical properties. The culprit, I suspect, is a presumed, but rarely articulated, commitment to the thesis that realism about the mental requires that mental predicates uniquely designate properties possessed by everything to which they apply (and in virtue of which they apply). As a warm-up, it might be useful to consider the origins of the idea that mental and physical predicates designate members of distinct categories of property.
3
THE MENTAL AND THE PHYSICAL
The mind–body problem strikes us as imponderably deep. According to David Chalmers, for instance, this is the last remaining “hard problem” (Chalmers 1996, xi). If the mental and the physical are obviously and dramatically different, however, why has there been so little agreement over the nature of this difference? Why have we found it so difficult to say with any precision what the mental–physical distinction boils down to (Crane and Mellor 1990)? Let me mention one possibility. Nowadays, even dualists are skeptical of Cartesian dualism. The idea that a mind is a wholly immaterial substance has lost whatever appeal it might once have enjoyed. Despite our rejection of Cartesianism, however, we have stuck with Descartes’s classification of properties as either mental or physical. Under the circumstances, it might be worth considering how Descartes was looking at this fundamental ontological distinction. The universe, Descartes held, comprises two kinds of substance: mental and material. Kinds of substance are distinguished by distinctive attributes. Mental substances are characterized by the attribute of thought; a material substance is extended. You could think of attributes as generic characteristics, determinables. A substance possesses an attribute by possessing determinate modes of that attribute. A material body is extended, but to be extended is to be extended in some determinate way; to have shape or size, is to have some determinate shape or size. In common with many of his contemporaries, Descartes distinguishes primary from secondary properties. Primary properties are properties possessed
Anomalous Monism
89
by every material thing, including the fundamental things.3 Secondary properties are arrangements of the primaries. The idea, very roughly, is that an object is spherical in virtue of being spherical; an object is red in virtue of its surface’s possessing a particular geometrical structure. For Descartes, extension and its modes are unproblematic. Extension can be exhaustively characterized geometrically. Once we turn our attention to the range of properties that Descartes counts as mental, however, it is more difficult to discern a unifying thread. Mental states and properties include sensations, attitudes, emotions, judgments, and images. It is hard to see this motley collection as constituting anything resembling a fundamental natural kind. To be sure, if you believed, as Descartes believed, that thought and extension were exhaustive and exclusive, the task of distinguishing the mental and the physical is simplified. You need only give a principled characterization of the one and consign anything that failed to satisfy this characterization to the other.4 Physical modes are modes of extension. Their nature is wholly geometrical. A mental mode, then, is just a mode that is not a mode of extension. Once we reject the fundamental Cartesian picture, however, this easy move is no longer open to us. Once we allow that extended bodies might think, the line between mental and physical modes begins to blur. Mental properties start to resemble material secondary properties. In reflecting on the situation, you might gradually come to suspect that the really pernicious aspect of our Cartesian legacy is not Descartes’s commitment to a dualism of substances but his commitment to a dualism of properties. While rejecting the idea that the world comprises mental and physical substances, we have clung to a Cartesian bifurcation of properties. The seeds of an alternative picture can be discerned in Locke and Priestly. There is but a single world characterizable in various ways. Some goings-on in the world answer to mental characterizations, characterizations couched in a vocabulary a Cartesian would classify as mental. These very same features of the world would answer to descriptions couched in a physical vocabulary. Differences among descriptions need not reflect differences among fundamental properties but only differences among families of predicates contrived to serve different purposes. Constraints on the application of mental predicates need not mirror constraints on applications of physical predicates. To the extent that this is so, it would be pointless or impossible to translate or analyze mental descriptions in physical terms, hence impossible to “reduce” the mental to the physical.
3 Or thing: Descartes is most naturally read as advocating a monistic conception of material substance as space itself. Ordinary objects are local “thickenings,” modes not substances. Motion of objects through space is a kind of wave motion. 4 In defending anomalous monism, Davidson (1970b, 211) makes a complementary point: descriptions of the physical are “recessive” relative to the mental.
90 John Heil You might think of this as a kind of neutral monism. It distinguishes the mental and the physical at the classificatory level, the level of predicates, but not in the reality answering to those predicates. And you might wonder who could possibly take such a monism seriously.
4
DAVIDSON AS A “NONREDUCTIVE PHYSICALIST”
In introducing the topic of higher-level, multiply realized properties, the heart of nonreductive physicalism, I invoked Putnam and Fodor. But nonreductive physicalism has another important source: Donald Davidson’s influential paper, “Mental Events” (1970b). “Mental Events” advances a thesis Davidson dubbed “anomalous monism.” According to Davidson, anomalous monism provides a way of coming to grips with a commitment to the following, apparently inconsistent, theses (1970b, 208): (1) “Some mental events interact causally with some physical events.” (“Causal Interaction”) (2) “Events related as cause and effect fall under strict deterministic laws.” (“Nomological Character of Causation”) (3) “There are no strict deterministic laws on the basis of which mental events can be predicted and explained.” (“Anomalism of the Mental”) Events, for Davidson, are physical by virtue of satisfying physical predicates, and mental by virtue of satisfying mental predicates: “an event is physical if it is describable in a purely physical vocabulary, mental if describable in mental terms” (Davidson 1970b, 210). One way to put this, a way that might surprise some of you, is that, by Davidson’s lights, the mental–physical distinction is not a deep one: any mental event or any event satisfying a mental predicate could be given a physical description. This, I think, is the core of the influential doctrine of the supervenience of the mental on the physical. Similarly, some physical events, events answering to physical predicates, could, at least in principle, be given mental descriptions (descriptions incorporating mental predicates). I see Davidson’s anomalous monism as consistent with the ontologically reductive position sketched earlier. If this sounds shocking, that might be because a common interpretation of anomalous monism reads into the doctrine a robust metaphysical thesis with exciting ontological implications. The interpretation I have in mind is largely associated with the work of Jaegwon Kim, particularly Kim’s earliest work on supervenience (see Kim 1973, 1984a, 1984b, 1985, 1989, 1993b). Kim’s reading of Davidson clearly expresses what has subsequently been labeled “non-reductive physicalism.” Kim construes Davidson’s talk of mental and physical predicates as talk of mental and physical properties. Mental–physical supervenience in this way becomes a meaty doctrine about necessitation relations holding among
Anomalous Monism
91
families of properties. If, as Davidson contends, mental descriptions cannot be translated into physical descriptions, and if you take description talk to be oblique for property talk, you will read Davidson as holding that mental properties are not reducible to, not identifiable with, physical properties, yet mental properties are somehow dependent on physical properties: mental properties have physical realizers. Now trouble looms. Mental events, Davidson notes, evidently enter into causal relations with physical events. Interpreting Davidson’s talk of predicates as property talk, this amounts to the thesis that events with mental properties can interact causally with events possessing physical properties. Kim puts it this way: The claim “Mental events cause physical events” only comes to the assertion . . . that events with some mental property or other are causes of events with some physical property or other. (Kim 1993a, 20) But according to Davidson, events stand in causal relations by virtue of falling under strict causal laws, and strict causal laws are expressible only in a physical vocabulary. Kim’s translation: one event causes another event in virtue of these events instantiating certain physical properties. It appears to follow, given the “anomalousness of the mental,” that causes and effects can include mental properties, but their being causes and effects must be due only to physical properties they include as well. The upshot: although every mental token might be identical with some physical token—every event possessing a mental property is (is identical with) an event possessing a physical property—mental and physical types (properties) are distinct. Davidson is read as embracing token identity but rejecting type identity, where types are presumed to be properties. The upshot is “non-reductive physicalism”: token identity, coupled with the rejection of type identity, plus the thesis that the mental supervenes on the physical.
5
DAVIDSON AS AN ANOMALOUS MONIST
For most philosophers, this is familiar territory. Some readers will recall the halcyon days of the early 1980s when it looked as though nonreductive physicalism provided a cost-free solution to the venerable mind–body problem. Largely thanks to Kim, however, the initial euphoria subsequently gave way to sober reflection on the mind–body problem in its contemporary guise: the problem of causal relevance (see McLaughlin 1989; Kim 1989, 1993a, 1993c). In the course of sorting out these issues, some philosophers have thought it important to distinguish Davidsonian and Kim-style events. The idea is that, for Davidson, events are, so to speak, dynamic substances, particulars possessing a variety of properties. Some of these properties will figure
92
John Heil
unproblematically in causal transactions; some will not. For Kim, in contrast, events are fi ner-grained: dynamic states of affairs. A Kim-style event is the instantiation, by a particular, of a property at a time (Kim 1973). Suppose a particular, c, at t, instantiates M, a mental property; and suppose, as well, that, at t, c instantiates a physical property, P. Davidson might regard this as a single event; for Kim there are two events: c’s instantiating M, and c’s instantiating P. Kim’s events are distinct but, given supervenience, the mental event depends on the physical event. I could go on about this, addressing, for instance, the nature of supervenience, but I lack the heart. I believe that much of the discussion of anomalous monism and its implications for nonreductive materialism is based on an initial misconstrual of Davidson.5 Anomalous monism is manifestly not a thesis concerning mental properties and relations these bear to physical properties. The mistake is to read Davidson’s unguarded talk of descriptions and predicates—yes, and occasional talk of properties—as ontologically serious talk of properties.6 Davidson is intent on keeping ontology at arm’s length. His interpreters, in contrast, have heard Davidson’s guarded comments about mental and physical descriptions as affording a tidy metaphysical blueprint. My suggestion: Davidson has an ontological side, but this is not it. With these thoughts in mind, we might reformulate Davidson’s three principles: (1´) Some events satisfying mental predicates interact causally with events answering to physical predicates. (2´) When events are related causally, they are describable in a way that invokes a strict law. (3´) Formulations of strict laws exclude mental predicates. Davidson puts it this way: It should now be evident how anomalous monism reconciles the three original principles. Causality and identity are relations between individual events no matter how described. But laws are linguistic; so events can instantiate laws, and hence be explained or predicted in light of laws, only as those events are described in one or another way. The principle of causal interaction deals with events in extension and is therefore blind to
5 A misconstrual that Davidson himself sometimes failed to appreciate, thereby giving rise to the impression that he was missing the thrust of his critics’ arguments; see Davidson (1993). 6 The term “property” occurs twice in “Mental Events” in a single passage (1970b, 214) in which Davidson characterizes supervenience. Subsequent passages, however, make it clear that Davidson is using “property” in the Episcopalian sense.
Anomalous Monism
93
the mental–physical dichotomy. The principle of anomalism of the mental concerns events described as mental, for events are mental only as described. The principle of the nomological character of causality must be read carefully: it says that when events are related as cause and effect, they have descriptions that instantiate a law. It does not say that every true singular statement of causality instantiates a law. (Davidson, 1970b, 215) One way to put the thesis is in terms of truthmakers. (Yes, I know: Davidson would not approve; but bear with me.) Truthmakers for mental descriptions will always answer to some physical description. The supervenience of the mental amounts to this thesis together with a consistency constraint on the application of mental predicates: events satisfying the same physical descriptions must satisfy the same mental descriptions. There is, however, no prospect of translating or analyzing mental terms into a purely physical vocabulary. Suppose a mental predicate, “M,” truly applies to an agent. Given supervenience, we know that whatever it is about the agent in virtue of which “M” applies truly would also make true the application of some physical predicate, “P”: two truths, one truthmaker. Supervenience allows distinct applications of the very same mental predicate to be made true by different kinds of physical state—states answering to different physical descriptions. If you read Davidson this way, you might begin to see why a familiar criticism of anomalous monism misses the mark. Suppose you—wrongly, by my lights—took Davidson to hold that mental predicates designate mental properties, and physical predicates pick out physical properties. Suppose, further, you thought that Davidson regarded supervenience as a relation among families of property: mental properties are in some way “grounded in” or “realized by” distinct physical properties. Then, as Kim and others have made clear, Davidson’s account of mental causation would be susceptible to the “qua problem.”7 Consider a particular event, c, and imagine that c is a mental event: event c possesses some mental property; M. Supervenience tells us that M’s presence depends on c’s possessing some physical property; P. So c is both a mental event and a physical event. Now imagine that c causes some event, e, and e possesses some physical property, P´ (we could suppose, as well, that some mental property, M´, supervenes on P´; see Figure 6.1.) Query: does c cause e in virtue of—qua—being P, in virtue of being M, or in virtue of being both P and M? Davidson’s commitments to the nomological character of causation and the anomalousness of the mental imply that it is c’s being P that causes e’s being P´; c’s being P causes, as well—albeit only indirectly via supervenience—e’s being M´. More generally, it appears that, whenever a mental event has a physical or mental
7 Fred Stoutland (1976) was, as far as I know, the fi rst to raise this difficulty for Davidson. Many others followed.
94 John Heil
Figure 6.1 Higher-Level Causation
effect, it does so qua physical, not qua mental. Mental properties are apparently epiphenomenal. If this is a problem for mental properties, it is a problem for any higherlevel property. Particular “token” events could possess such properties, but they—the properties—would seem to make no causal difference to their possessors. An event’s “causal clout” is borne by the event’s fundamental physical properties: properties on which higher-level properties are taken to supervene. The qua problem is a monumental stumbling block for strains of “nonreductive physicalism.” It is not, however, a problem for Davidson. Mental predicates—indeed “higher-level” predicates generally—hold true of objects in virtue of those objects’ possessing assorted properties answering to physical predicates. Pretend, for a moment, that you are in pain in virtue of your C-fibers fi ring: your C-fibers fi ring serves as the truthmaker for “You are in pain” and for “Your C-fibers are fi ring.” Davidson is thinking that, if the fi ring of those C-fibers causes you to walk to the medicine cabinet, it would be no less true to say that your being in pain causes you to walk to the medicine cabinet. Davidson rejects reduction if this is taken to be analytical or explanatory reduction. Translation of pain talk into C-fiber talk, for instance, is hopeless. “Bridge principles” relating mental and physical predicates are not in the cards. Explanations couched in mental terms are not convertible into explanations couched in a physical vocabulary. A rejection of analytical reduction, however, is consistent with acceptance of ontological reduction. One way to put this would be to say that, in a given application, expressions can differ in their truth conditions while sharing truthmakers. Couple this idea with the idea that laws of nature are grounded in fundamental features of the fundamental thing or things. These features will be picked out in the vocabulary of basic physics. Formulations of laws of nature, then, inevitably involve an exclusively physical vocabulary. This is what the “anomalousness of the mental” amounts to. Applying these facts to knowledge of identities, we see that it is possible to know that a mental event is identical with some physical event without knowing which one (in the sense of being able to give it a unique physical description that brings it under a relevant law).
Anomalous Monism
95
Even if someone knew the entire physical history of the world, and every mental event were identical with a physical [event], it would not follow that he could predict or explain a single mental event (so described, of course). (Davidson 1970b, 224) You can see how this might work for many ordinary predicates. Consider the color predicate “is red.” It seems likely that this predicate applies to objects by virtue of those objects’ possession of any of an open-ended family of similar physical properties.8 It is unlikely that we could translate color talk into talk of these physical properties. Of course, if the physical properties answering to color predicates are finite in number, we might provide a list and associate items on this list with each color predicate. But this would no more constitute a translation than would the association of a list of all the penguins with the predicate “is a penguin.”
6
DAVIDSON, PHYSICALISM, AND REALISM
Some of you will balk at my reading of Davidson. Others, indifferent to the reading, will regard the position I am ascribing to Davidson as a nonstarter. In some cases this will be because you think that, thus interpreted, Davidson becomes an antirealist about the mental. Although there are undoubtedly difficulties for Davidson’s position—it embodies, after all, a range of contentious philosophical theses—I do not think antirealism is one of them. Davidson accepts that mental ascriptions are often true. It is true that we harbor beliefs and desires and, on the basis of these, form intentions to act. It is true that actions—bodily motions, according to Davidson—are caused by mental events. What is not true is that mental predicates designate nonphysical properties. Mental predicates are satisfied by complex properties, perhaps, but these very same properties also satisfy physical predicates. There is nothing more to the properties’ (or states’, or events’) being mental—or physical—than their satisfying mental and physical predicates, respectively. If it is difficult, or impossible, to distinguish mental and physical predicates in a crisp way, this is not the end of the world. It is not the end of the world because the distinction lacks deep ontological significance. In “Mental Events,” Davidson labels his position “anomalous monism” and repeatedly invokes identity as the relation holding between mental and physical states. Most proponents of identity theories of mind regard themselves as physicalists. So is Davidson, at heart, a physicalist?
8 This is intended merely as an illustration, not as an account of color; see Heil (2003, ch. 17).
96 John Heil To think so is to miss a central tenet of anomalous monism. Suppose I am right in thinking that Davidson regards the mental–physical distinction as ontologically insignificant. A physical state is a state answering to a physical predicate. Such states can also satisfy mental predicates. The only sense in which the physical takes precedence over the mental is that, whereas every state of the world is physical—each answers to a physical predicate—every state need not answer to a mental predicate. Indeed Davidson is at pains to show that there is nothing in principle to keep us from describing any physical state in mental terms. Davidson, in the passage in which he makes this point, mentions Spinoza. Spinoza held that there was but a single substance, God or Nature, possessing an infi nity of attributes. Of these attributes, we are privy to just two: the Cartesian attributes, thought and extension. Ordinary features of the world are modes of thought and modes of extension. But how are these modes related? Spinoza is usually characterized as a “dual aspect” theorist. You could understand this as the view that every property has (at least) two aspects: a mental aspect and a physical aspect. But what are aspects, and how are aspects related? The issues here are complicated, but I believe that a case could be made for the idea that the relation among Spinoza’s aspects is strict identity: every mental mode is strictly identical with some physical mode (and vice versa). Thus interpreted, Spinoza’s position is, as Davidson notes, a close relative of anomalous monism. Neither Spinoza nor Davidson can be seen as advancing reductionism, but neither is a dualist either. There is a dualism, all right, but it is a dualism of conception, not a dualism of substances or properties.
7
MONISM IS NOT PHYSICALISM
We are now in a position to see why Davidson insists that it is misleading to describe the mental as “nothing over and above” the physical. This is not because mental properties are distinct from physical properties. Rather, it is because describing the mental as nothing over and above the physical misleadingly privileges the physical. It suggests that only physical predicates literally hold true of objects. But the same region of the world that makes true the application of some complex physical predicate can make true—make true—the application of some mental predicate. Davidson, like Spinoza, is a monist, not a physicalist: not a reductive physicalist, not a nonreductive physicalist. All this is easy to miss so long as we continue to think that the legitimacy of a given predicate turns on its designating a unique property. This condition is met, if at all, only when we reach fundamental physics. There, predicates do indeed appear to correspond one-one with properties. But this by no means implies that truths about the world require formulation
Anomalous Monism
97
in the vocabulary of fundamental physics. Fundamental physics provides us with a way of getting at the fundamental substances and properties, arrangements of which serve as truthmakers for endless other, less fundamental, predicates. None of this makes the world mind- or theory- or language-dependent. The world is “out there” waiting to be considered and described by conscious agents. The significance of our predicates depends on us. It is a contingent fact about us that we use the predicate “is red” to pick out redness and not greenness, or saltiness, or triangularity. But this in no way makes redness, greenness, saltiness, or triangularity language- or mind-dependent. What we have in Davidson is a flat-out realism of outlook coupled with a reluctance to say more than absolutely necessary about fundamental ontology. This very reluctance prevented Davidson from appreciating arguments advanced by his critics and from replying to those arguments in a way that his critics would have found satisfying. All of this comes to a head in “Thinking Causes” (1993), one of Davidson’s last attempts to respond to difficulties Stoutland, Kim, and many others found in his view. In that paper, Davidson makes what he regards as an innocent move that turns out to have disastrous results. Davidson concedes Kim’s reading of Davidson’s “predicates” as referring to properties. If we understand properties in an ontologically serious way, this in effect legitimizes Kim’s longstanding unhappiness with Davidson. The charitable interpretation here, however, takes Davidson to be understanding “property” only in the pleonastic way mentioned earlier: only in the Episcopalian sense in which an object could be said to possess a property, F, insofar as “is F” is truly predicable of it. As I have tried to suggest, bad things happen when you move directly from talk of properties in this deflated sense to ontologically robust conclusions. Davidson’s early reservations about talk of properties undoubtedly stemmed from a stark nominalism inherited from Quine. A nominalist construes property talk as ontologically thin. If you understood properties as sets of objects, for instance, you would not see properties as causally operative features of the world. You might interpret Davidson’s silence on properties as an endorsement of nominalism. Kim reads Davidson as accepting property realism, and Davidson’s talk of predicates as shorthand for talk of properties. But it is more plausible to interpret Davidson’s silence on properties as nothing more than silence on properties. I have suggested that Davidson’s position is consistent with an ontology that includes properties. Davidson tells us, in effect, that the truthmaker for the application of a given mental predicate is going to be the truthmaker for the application of some physical predicate, but he tells us nothing of the truthmakers’ ontology. Other philosophers believe that we have excellent independent reasons to believe that the world comprises, in addition to objects, or substances, properties. As I see it, all this is consistent with Davidson’s considered views. The interpretive mistake is not that
98
John Heil
of reading a nominalist theory as realist, but in reading an ontologically neutral theory as endorsing a linguisticized realism.
8
DÉNOUEMENT
My aim here has been to set the record straight and provide a sympathetic reading of Davidson. If I am right, then many of the most popular lines of criticism of Davidson are off base. This leaves open, of course, the very real possibility that the kind of account I take Davidson to be offering is itself ultimately unsatisfactory. If we are to put Davidson behind us, however, we ought to do so for the right reasons.
7
On the Metaphysical Implications of Some Epistemological Commonplaces Frank Jackson
1
INTRODUCTION
Epistemology is a different subject than those of metaphysics and of the philosophy of language. All the same, there are important links between, on the one hand, what’s evidence for what and, on the other, the right account of certain properties. The reason for this is obvious. Whether or not something is evidence for the possession of property F is a function in part of what F is. Something looking square is a reason for holding that it is square; it is not a reason for holding that there will soon be peace in the Middle East, and the difference lies in part in the difference between the property of being square and the property of peace being imminent in the Middle East. Someone having red hair is not a (good) reason for increasing one’s expectation that he or she is hot tempered but is a reason for increasing one’s expectation that they have Nordic ancestry. Again, the reason for the difference in support lies in part in the nature of the two properties in question. These observations, and the many like them that we might have made to illustrate the point, are relative to one or another presumed background body of evidence. An expert on the Middle East might have agreed with me to post on her Web site a figure that looks square if and only if her examination of all the available information leads her to believe that peace in the Middle East is close at hand. In this special case, something looking square (to me on her Web site) would be a reason for me to increase my expectation that there will soon be peace in the Middle East. But this doesn’t alter the fact that, against a normal background, something looking square is no evidence for peace in the Middle East but is evidence for that thing being square. My purpose in this chapter is to show how some simple observations, commonplaces really, about what’s good evidence for what can impact on some highly contested positions in metaphysics. I discuss color properties and mental properties. It will be obvious how the “method” might
100
Frank Jackson
be used more widely.1 This chapter is an attempt to publicize a way of addressing difficult metaphysical issues by drawing on relatively noncontroversial theses about what’s evidence for what. It is a way that has a long history but one that is currently under a cloud (a judgement confi rmed by the discussion at the conference). I start by giving the line of argument in the two cases listed in an unadorned form. I then discuss some objections that might be brought against it. A number of these objections came up, in one way or another, in the discussion at the conference that spawned this volume. My hope is that I’ve addressed a representative sample of the objections that are responsible for current doubts about using epistemic commonplaces to reach metaphysical conclusions. How is this relevant to a volume on realism and truth? I think Michael Devitt is right when he insists that a doctrine that limits itself to affi rming that “[s]omething objectively exists independently of the mental” is a realism that “is so weak as to be uninteresting.” He calls it “fig-leaf realism” (Devitt 1984, 22). A realism worth affi rming must include inter alia respect for a good part of our ordinary convictions about what’s good evidence for what. A realism about stars and tables which holds that, although stars and tables exist and do so independently of our observations, our opinions about where they are and what they are like are worthless is not an honest-to-goodness realism. This is because it has no substantive answer to the question, What should we be realists about? Tables and stars won’t be a substantive answer on the view in question because the view precisely denies that we have any well-founded idea of what they are like.
2
COLOR
Many philosophers hold that (surface) color properties are reflectance profi les (or some such). 2 The epistemology of color tells us that this cannot be right. Something looking red is a good reason for holding that it is red but isn’t a good reason for holding that it has such and such surface reflectance profi le (or anything along similar lines). What we have, that is to say, is an application of Leibniz’ law, as the following makes explicit. Premise 1. Something looking red is good evidence for its being red.
1 For another application of the method, see Jackson (2006). For how the method can be used to support a controversial thesis about the semantics of names, see Jackson (2007b). 2 Including my former self, see, e.g., Jackson and Pargetter (1987). I did not see the (really very obvious) problem until I wrote “Colour for Representationalists” (2007a). The view that colors are reflectances (or some such) is often known as Australian realism about color. An early defense of it is in Armstrong (1968a, ch. 12).
On the Metaphysical Implications
101
Premise 2. Something looking red is not good evidence for its having such and such a reflectance profi le. Intermediate conclusion: being red has a property that having such and such a reflectance profile lacks—looking red is good evidence for the fi rst but not the second. Therefore, being red ≠ having such and such a reflectance profi le. Of course something looking red is defeasible evidence for its being red. We may know that we are under an illusion as happens when white objects look red under red illumination. Also, in certain circumstances, something looking red may be good evidence that it has such and such a reflectance profile. Taken in conjunction with certain experimental results linking looking red with the reflectance in question, a thing looking red is good evidence that it has such and such reflectance profile. However, it remains true that, in normal circumstances and absent special knowledge from optical science, looking red is good evidence for being red but is not good evidence for having such and such a reflectance profile, and that’s all we need to use Leibniz’ law to infer that being red ≠ having such and such reflectance profile. Nonidentity follows from there being at least one property possessed by being red that is not possessed by having such and such reflectance profile.
3
PHENOMENAL PROPERTIES
Many philosophers nowadays hold that phenomenal properties are neurological properties. Until recently, most philosophers of mind held that the possibility of multiply realizing the functional role of any given phenomenal property shows that phenomenal state types could not be identified with brain types. This was the standard reason for replacing the “Australian” mindbrain identity theory with functionalism (see, e.g., Churchland 1988, ch. 2, and the references therein). However, there has recently been a significant change of heart on the question (see, e.g., Papineau 2002, ch. 1; Hill 1991). Sometimes the impetus is the special character of the phenomenal. The idea is that although belief and desire are multiply realizable functional states which should not be identified with neurological states, those mental states with a phenomenology—pain, seeing red, and all the rest—are different. They should be identified with neurological state types. That way we capture their felt character. Sometimes the impetus is what we are said to learn from scientific identities like that of water with H 2O. We wrongly thought, runs this second line of thought, that mental states are multiply realizable because we ran together contingency and aposteriority. What realizes a mental state role is a posteriori but not contingent. This means that multiple realizability is false. In this view, a possible creature, a “Martian,” who writhes and screams
102
Frank Jackson
in exactly the way we do when a leg is broken, and whose internal functional organization and information-processing parallels ours, is not in pain if his functional roles and information processing are realized in very different stuff from ours; he has “fool’s pain” or “twpain,” not pain proper. Sometimes both lines of thought are at work in tandem. Philosophers who urge that phenomenal mental types (or all mental types) are neurological types often describe themselves as scientific or empirical functionalists. The contrast is with commonsense or analytical functionalism: functional role is seen as a kind of reference fi xer for mental types rather than any kind of analysis of the mental. The epistemology of the phenomenal tells us that scientific functionalism cannot be right, and we can see this without needing to enter the debate over the possibility of multiple realizability of the mental. The argument is very like the argument above for color and reflectances. The crucial point is that the kinds of circumstances which normally warrant our believing that someone is in pain, seeing red, itching, or whatever, do not warrant our believing that someone’s brain is in such and such a neurological state. Here is how it looks, highlighting the role of Leibniz’ law: Premise. The kind of evidence that typically warrants believing that the property of being in pain is possessed by x differs from the kind of evidence that typically warrants believing that the property of having a brain with C-fibers (or whatever) fi ring is possessed by x. Intermediate conclusion: being in pain and having a brain with C-fibers fi ring have different evidential properties. Therefore, being in pain ≠ having a brain with C-fibers fi ring.3 I said that our two arguments would be given in unadorned form. I have kept the “to-ing and fro-ing” for the objections and my replies.
4
FOLK OPINION
Objection i You are taking for granted common opinion about what’s evidence for what. Reply No doubt the folk have some wrong opinions on what’s evidence for what. But it is hard to accept that they are wrong about such facts as: looking so
3 Equally, the same line of thought with minor reformulations leads to the conclusion that being pain ≠ being C-fibers fi ring.
On the Metaphysical Implications
103
and so is evidence for being so and so, other things equal; and that we often have good evidence for how we feel or how others behave, including what they say that gives us good evidence for someone—sometimes ourselves, sometimes someone else—being in pain, seeing red, or believing in God. Equally, it is hard to deny that this kind of evidence is, except in very special circumstances, no evidence concerning the reflectances of red objects or the neurology of pain and itching. We need to carry out detailed scientific investigations to get evidence about reflectances and neuroscience. In discussion it has sometimes been put to me that the folk do in fact know, for example, the reflectance profi le of red objects. This is because looking red is a reliable indicator of the reflectance profi le, and a reliable indicator theory is the right theory of knowledge. A similar claim is sometimes made about folk knowledge of neurology. But of course our argument was framed in terms of good evidence, not in terms of knowledge. (I would also object to the presumed version of a reliable indicator account of knowledge but there’s no need for me to do that to make my case.)
5
PROPERTIES AND CONCEPTS
Objection ii When David Armstrong argued that surface colors were physical properties of surfaces, reflectances say, he emphasized that he was not offering this as an account of the concept of color. This was of a piece with his treatment of mental state types, where he argued that they are neurological state types, while insisting that the concept of a mental state was to be understood in causal-functional terms (Armstrong 1968a). All you’ve shown, runs the objection, is something about concepts. You may have added to the case for insisting that our concept of color and our concept of the mental are quite distinct from our concept of reflectance and our concept of the neurological, respectively, but you’ve shown nothing about properties. Reply Concepts and properties don’t float free of each other in the way presumed by this objection. When I believe that something is red, I believe that the object falls under the concept RED. But this is to believe that the object is a certain way. Items fall under concepts, and concepts apply to items because of the way the items are. Concepts wouldn’t be any use to us in categorizing our world if that were not the case. What way does something need to be to fall under the concept RED? The red way—of course. To believe that something is red isn’t to believe that it is some way other than red. But, as we’ve already argued, the way I believe the object to be when I believe it to be red is the way supported by the object looking red—defeasibly, other things equal—and that way
104
Frank Jackson
≠ having so and so reflectance, for the reflectance way isn’t supported by the object looking red. I am not saying that concepts and properties line up one-one. The individuation of concepts is a contentious issue but most agree that there are cases where there is more than one concept for a single property.4 Being the smallest prime and being the smallest even number are plausibly different concepts but there’s one property, namely being the number two, that corresponds indifferently to falling under either concept.5 What I am saying is that to believe that something falls under a given concept is to believe that it is a certain way, and this means that concepts don’t float free of properties. For each concept, there is the question of what way something has to be if it is to fall under that concept. This means that drawing the concept–property distinction doesn’t avoid our argument. When we ask the question, What does it take for an object to fall under the concept RED, the answer is not that the object has so and so reflectance for the reason already given. What it takes is something supported by looking red, whereas having so and so reflectance is not supported by looking red. Mutatis mutandis for the case of mental properties.6 What is true is that it would have been a mistake to argue as follows: the concept RED ≠ the concept REFLECTANCE, therefore being red ≠ having so and so reflectance. That this would be a mistake follows from the fact that concepts outrun properties. But our argument did not proceed from the difference in concepts. It proceeded from the difference in epistemic properties as between being red and having so and so reflectance.
6
OPACITY
Objection iii Quine urged many years ago that it is possible to believe that Cicero denounced Catiline without believing that Tully denounced Catiline. All the same, Cicero = Tully. Likewise, he observed that one might believe that the tallest spy is a spy without believing that Orcutt is a spy even in the
4 Everyone agrees about this on the narrow notion of property that relates to fundamental joints in nature, but even on the much more inclusive notion of property at work in this chapter, it is plausible that concepts outrun properties, as we’ll see. 5 Although there’s a use of “property” that we might tag “property concept” as opposed to property in re, according to which if there are two concepts, there are two properties. 6 And mutatis mutandis for the reply to the knowledge argument that contents itself with saying that the argument overlooks the concept–property distinction, see Jackson (2005).
On the Metaphysical Implications
105
case where Orcutt = the tallest spy.7 In the same way, runs the objection, there might be evidence for Cicero’s denouncing Catiline that did not apply to Tully’s denouncing Catiline; or, even more obviously, for the tallest spy being a spy that did not support Orcutt being a spy (evidence for the fi rst we get for free but not for the second). But it would be an obvious error to infer from the differences in epistemic properties that Cicero ≠ Tully or that the tallest spy ≠ Orcutt. In short, you are forgetting that epistemic properties induce referential opacity.8 Reply In thinking about these examples, it is important to be careful about what bears the various properties and, in these cases especially, about whether it is a person or a proposition that bears the properties. Those who hold that we can believe that Cicero denounced Catiline, while not believing that Tully did, should not infer that Cicero has a property that Tully lacks: namely that of being believed to have denounced Catiline, observing that Leibniz’ law fails for belief properties. Leibniz’ law applies universally. Rather, they should insist that what has the property of being believed is the proposition that Cicero denounced Catiline, and what lacks this property is the proposition that Tully denounced Catiline. We have, that is, a valid application of Leibniz’ law that shows that the two propositions are distinct despite the fact that Cicero = Tully. In the same way, cases where there is evidence that Cicero denounced Catiline, which is not evidence that Tully did, are cases where there is evidence in favor of the proposition that Cicero denounced Catiline, which does not support the proposition that Tully denounced Catiline. The difference in evidential properties shows a difference in propositions, but that’s consistent with the different propositions concerning one and the same person. Mutatis mutandis for the Orcutt example. Of course, nowadays many philosophers—those who hold certain direct-reference theories about proper names—reject the two proposition position on the Cicero–Tully example. They insist that the proposition that Cicero denounced Catiline is the same proposition as the proposition that Tully denounced Catiline. But they rightly grant in the same breath that this implies that one cannot believe the one proposition without believing the “other.” Similarly, one cannot, according to the stance on proper names of these philosophers, have evidence in favor of the proposition that Cicero denounced Catiline without having evidence in favor of Tully denouncing
7
In a number of places, see, e.g., Quine 1960, § 30. Thus, “it is fallacious to combine Leibniz’s law with sentences containing verbs of propositional attitude—that is, with sentences that are concerned with psychological states like certainty, knowledge, and belief” (Hill 1991, 85). 8
106
Frank Jackson
Catiline. For “they” are one and the same proposition. (I might also add that these consequences are widely seen as a problem for the direct-reference position on names; we return to this issue below.) The situation is as follows. If Quine is right that it is possible to believe that Cicero denounced Catiline without believing that Tully did, the objection wrongly takes certain properties of propositions to be properties of persons. If the direct-reference theorists are right, the alleged datum is not a datum in the fi rst place. I should emphasize that I am not suggesting that one cannot have beliefs about objects and properties, or that one cannot have evidence that bears on whether or not some property is or is not present. Indeed, the master argument of this chapter turns precisely on the fact that among the properties of properties are such properties as being or not being justifiably believed to be present on the basis of so and so evidence. I argued, for example, that being red and having so and so reflectance profi le cannot be the same property precisely because what was good evidence for the presence of one property differed from what was good evidence for the other. The crux of my reply to the objection is only that some cases where we appear to be ascribing a property to an object are, from the Quinean perspective, cases where we are in fact ascribing a property to a proposition. Indeed, it would be very implausible to deny that properties have epistemic properties. It is a commonplace that we marshal evidence that bears on the distribution of properties in our world and that we have beliefs about the distribution of properties.
7
A FALLACY OF EQUIVOCATION
Objection iv Whatever should be said about the Tully–Cicero example, we can construct a similar example except that it applies to properties instead of objects. Here is the case. Mary knows that Fred wants to curry favor with the King by wearing the King’s favorite color. The King’s favorite color is red, but Mary doesn’t know this. Perhaps she thinks it is blue. Now consider the following argument: Premise 1. Mary’s body of evidence supports Fred’s wearing something that has the King’s favorite color. Premise 2. Mary’s body of evidence does not support Fred’s wearing something that is red. Intermediate conclusion. The King’s favorite color has a property red lacks, namely, its being possessed by something worn by Fred is supported by Mary’s body of evidence.
On the Metaphysical Implications
107
Conclusion. The King’s favorite color ≠ red. It is clear that the argument commits a fallacy—its premises are true and its conclusion false. This tells us, runs the objection, that we cannot use epistemic arguments to draw conclusions about the identity or nonidentity of properties. Reply The displayed argument commits a fallacy of equivocation. The reading on which premise 1 is true is a reading which takes Mary’s body of evidence to support an item of Fred’s clothing having the property of being the color the King most likes, whatever that property may be. This property is not being red, even in the case where the King’s favorite color is red. It is instead the property x has if and only if there is a property which is the King’s mostliked-color, and x has that color, whatever it may be. It is the property of being a color that stands in a certain relationship to the King’s likings. We can think of it in infinite disjunctive terms: x has the property iff (x is blue and the King’s favorite color = blue) or (x is red and the King’s favorite color = red) or (x is yellow and the King’s favorite color = yellow) or . . . . This means that the intermediate conclusion that follows from the two premises is Intermediate conclusion*. Being the King’s favorite color has a property red lacks, namely, its being possessed by something worn by Fred is supported by Mary’s body of evidence. But now the conclusion that follows is: Conclusion*. Being the King’s favorite color ≠ red. This conclusion is true. What we learn from the example parallels what we learn from the Cicero–Tully example: there’s no problem about using Leibniz’s law with epistemic properties, but one needs to be careful about what has what epistemic property.
8
PROPERTIES AND PRESENTATIONS
Objection v You mentioned that direct-reference theorists about proper names hold that one cannot believe that Cicero denounced Catiline without believing that Tully denounced Catiline. They deny Quine’s claim. But, proceeds the objection, they allow that there is something that needs attention in the example. Although Quine is wrong, the mistake is understandable. What direct-reference theorists say at this point varies, but there is a common
108
Frank Jackson
general shape to their discussion. They distinguish objects from their modes of presentation. Although the proposition that Cicero denounced Catiline is the very same proposition as the proposition that Tully denounced Catiline (and so, by Leibniz’s law “they” have the very same properties including epistemic ones), there is a distinction to be drawn between “Cicero” presentations of Cicero–Tully, and “Tully” presentations of Cicero–Tully. When Fred expresses what he believes using “Cicero denounced Catiline” while insisting that “Tully denounced Catiline” does not express what he believes, there is, unknown to Fred, one proposition believed that is graspable in two ways, and in some sense Fred knows about the “Cicero” way but not about the “Tully” way. We should say something similar, runs the objection, for the example of red and reflectance. There is one property that presents itself in two ways: one way is in its guise as a property of optical physics; the other is in the way it affects those with normal color vision. (And of course something like this is sometimes said in discussions of the knowledge argument.) Similarly, runs the objection, pain may present itself in its neurological clothing or in its sensory clothing. But there is just the one property differently presented. Reply I think that there is a grain of truth in this objection. It tells us that we can distinguish two properties when we discuss color and the sensory and that what I’ve been saying is true for one member of the pair but not the other. But before I go concessionary, let me start with why there is only a grain of truth. Of course properties can and do have properties—it is a property of heat that it is an energy property, of bright yellow that it attracts attention, of a perfect circle that it is hard to draw, and so on. In consequence, there is nothing problematic about a property having a property through which it draws itself in one way or another to our attention: a presentational property if you like. But how does this allow us to identify redness with reflectance so and so? When something looks red to us in normal circumstances, we typically believe that the thing is red, and it is this property we typically believe the thing has. This property is not reflectance so and so. For we do not believe that the thing has reflectance so and so. Maybe reflectance so and so is the property that is presenting itself to us but that’s not the property we believe the thing to have. That is to say, maybe when we believe that x is red, we have a belief of the form: property x has the property which presents in the red-looking way, where the property that presents in the red-looking way = reflectance so and so. But having the property which presents itself in the red-looking way ≠ the property which presents itself in the red-looking way (whatever it is). The reason parallels the one we noted in my discussion of the case where red is the King’s favorite color. The fi rst
On the Metaphysical Implications
109
property is the property x has if and only if there is a property which red presents, and x has it. To put it in disjunctive form: x has the property iff (x has property1, and property1 presents itself in the red-looking way), or (x has property2 , and property2 presents itself in the red-looking way), or (x has property3, and property3 presents itself in the red-looking way), or . . . , where one but only one value of propertyi is reflectance so and so. In sum, redness is the property we believe something to have when we believe it to be red, and that property is not reflectance so and so although it may well be the case that (i) it is the property of having the property that presents in a certain way familiar to those with color vision, and (ii) the property that is presenting in that way is reflectance so and so. Mutatis mutandis for the case of phenomenal experience. Now for the concession. If to be red is to have the property that is thus and so, then one might argue that the philosophy of color needs to acknowledge two properties and not one: being red—that’s the property of having the property that is thus and so, and red—that’s the property that is thus and so (reflectance so and so, as we’ve been supposing). The situation would be very like that which some functionalists of the identity theory school hold obtains for phenomenal properties and mental properties more generally. We have the property of having the property that plays a certain functional role and the property that plays the role. There are, that is, two different properties, both of which can be thought of as mental properties. Here is how it looks for the case of pain on the usual pretence that C-fibers fi ring is the state type that, as it happens, plays the pain role:9 Being in pain = being in the state that plays the pain role; The state that plays the pain role = C-fibers fi ring. But it doesn’t follow from this pair that being in pain = C-fibers fi ring. If we call the state that plays the pain role pain, we get, by transitivity, in the way made famous by the Australian materialists: Pain = C-fibers fi ring. In the past I have defended both the Australian materialists’ type-type identity of mental properties with neurological properties (Jackson et al. 1982) and the Australian position that colors are reflectances by appeal to a twoproperty story of the kind sketched above (Jackson 2007a). I now think it is important to distinguish the plausible claim that there are two properties
9 I suppress the complications in formulation needed to cover the fact that roles played may vary from time to time and creature to creature. They don’t matter for our discussion.
110 Frank Jackson which we need to keep track of in our philosophizing about color and mental properties,10 from the much less plausible claim that the reflectance properties and the neurological properties are properly called color and mental properties, respectively. Isn’t red the property we believe things to have when they look the way blood looks, and that’s not reflectance? Isn’t pain the property we believe we have when a pin is stuck into us, and that property is not C-fibers fi ring or anything of that ilk? The identity theorists were wrong about the featured event; what they got right concerned a supporting act. Similar remarks apply, it seems to me, to the case of water and H 2O. I gave earlier, as an example where we have two concepts and one property, the distinct concepts of being the smallest prime and being the smallest even number, each of which corresponds to the single property of being the number two. Many would give as another example the concept WATER and the concept H 2O, holding that they are distinct concepts corresponding to the one property. But the claim that there is just the one property here faces a major objection. Think of the time before it was known that water is H 2O and those who believed at that time that some stuff before them was water. They believed the stuff was a certain way, and what way could that be other than its being water? But that way wasn’t its being H 2O, for what they believed about the nature of the stuff before them was something that they were typically entitled to believe on the basis of what they knew at that time, but they weren’t entitled to believe that the stuff before them was H 2O. We can set the objection out in Leibniz law style as follows: Premise 1. Being water was a property people in the early 1700s were often entitled to believe was possessed by the stuff before them. Premise 2. Being H 2O was not a property people in the early 1700s were often entitled to believe was possessed by the stuff before them. Conclusion. Being water ≠ being H 2O. Of course there is something right about the idea that “water = H 2O” is a true type-type identity sentence (be it necessary or not). What’s right, in my view, is the fact that to be water is to be the kind which is thus and so, and that kind = H 2O, for any at all plausible spelling out of the “thus and so.” As with color and mental properties, when we philosophize about water, we have two properties to keep track of—being water and water, as we might tag them following the lead of the discussion of color and mental
10 The view that there are two properties to keep track of when discussing mental properties is implicit in some early presentations of the mind-brain identity theory. As far as I know, it fi rst became fully explicit in Lewis (1966).
On the Metaphysical Implications
111
properties—and one of them, water, is being H 2O, but the other, being water, is not. Or consider the examples that early defenders of the mind-brain identity theory appealed to. They pointed out that type-identity claims like “temperature in gases = mean kinetic energy,” and “lightning = an electrical discharge” are true (give or take points of scientific detail that are by the way in this context). It follows, as the early identity theorists emphasized, that there is something seriously mistaken in concluding that these scientific identities are false on the ground that once upon a time we believed that there was some lightning in the sky, while failing to believe that there was an electrical discharge in the sky, and that once upon a time we believed that a gas had such and such a temperature, while failing to believe that it had so and so kinetic energy. But it is important to appreciate precisely what the mistake would be. There is nothing mistaken about arguing that the property we believe a happening to have when we believe that it is lightning is not the property of being an electrical discharge on the basis of their difference in epistemic properties. There is nothing mistaken about arguing that the property we believe a gas to have when we believe it has such and such a temperature is not the property of having so and so kinetic energy on the basis of their difference in epistemic properties. To say otherwise would require denying Leibniz’ law or else that properties have epistemic properties, and neither option has any plausibility. What would be a mistake is to infer that the type of phenomenon in the world which has the property we believe something to have when we believe it is lightning can’t be an electrical discharge. It can be and is. We should distinguish lightning from being lightning. Being lightning is the property we believe something to have when we believe it to be lightning. Lightning is the type that has the property. The second but not the fi rst is a kind of electrical discharge. The discovery that lightning is an electrical discharge is the discovery that the kind that has the property of being lightning is an electrical discharge. Mutatis mutandis for temperature in gases. I said near the beginning that it would be obvious how the “method” advertised in this chapter might be used more widely. I’ll stop on that prospective note.
8
Mathematical Platonism J. J. C. Smart
1
QUINE’S PLATONISM
Of late I have been of the opinion that there is no really satisfactory philosophy of mathematics at present but that a broadly Quinean one is the most satisfactory one on offer. In particular it solves the epistemological problem that bedevils traditional Platonism. We need no mysterious faculty of intuition of the Platonic entities: we know of numbers and sets through the hypothetico– deductive method of science, in particular physics. This is because mention of real and complex numbers is part and parcel of physics, which is tested holistically. Ockhamist tendencies might lead one to restrict the amount of Platonistic mathematics that one should believe, perhaps to the constructivist mathematics of Errett Bishop. This might do for physicists and engineers, but it seems to lack the elegance of the classical approach due to Dedekind, Cauchy, and the like. Still, the indispensability argument works well for this less elegant mathematics. The classical approach could be defended by appealing to considerations of a sort of simplicity as a guide to truth and which may be regarded as part of the hypothetico–deductive method. Those who reject such considerations can still defend a more limited Platonistic ontology on the basis of indispensability conditions. Physical theories indispensably quantify over numbers, sets of numbers, and so on, no less than over electrons and protons, the theories are tested holistically, and so we believe in the mathematical entities just as we believe in electrons and protons. Let us concede, then, as I certainly want to, that Platonism can be defended in a Quinean way and that mathematical objects exist and are neither spatiotemporal nor causal. Unlike traditional Platonism it requires no epistemological mysteries and no intuitions of reality that cannot be explained in principle in a naturalist, indeed physicalist manner. We can imagine even an artificial machine that operated the hypothetico–deductive method in physics and became a Quinean Platonist. I believe that some of us, though not artificial, are such machines. Nevertheless I have a desire that pure mathematics should not be indebted to physics. Connected with this is my wish that a fully believable philosophy of mathematics should be one that G. H. Hardy would have liked.
Mathematical Platonism 2
113
QUINE AND HARDY
Are there any really believable philosophies of mathematics other than (possibly) Quine’s or perhaps as modified in ways suggested by Colyvan (2001) and in some of his more recent work? Well there are some halfbelievable philosophies of mathematics but these strike me either as refutable or as profoundly unattractive as well as implausible. Logicism, which in a sense identified mathematics with set theory but also identified logic with set theory, got exploded by the realization that, as Quine has shown us, there is a clear and natural break between logic proper and set theory. Three considerations make the break in the same place. Taking logic to be fi rst order predicate logic (quantification theory, as Quine called it), we have: (i) logic has no constant predicate, other than the identity predicate which is itself eliminable in application of the logic to a discourse with a fi nite vocabulary; but set theory has the set membership predicate. (ii) logic is complete, that is, all valid schemata are provably so from a fi nite set of axioms; whereas set theory is, as Gödel proved, incompletable; (iii) logic is not Platonistic in Quine’s sense, that is, it has variables and predicates but does not have expressions for, or quantification over, abstract entities such as sets. I regard so-called higher order logic as in Quine’s witty phrase “set theory in sheep’s clothing.” (If we accept George Boolos’s plural quantification, logic loses features (i) and (ii) but not (iii).) Thus logic and set theory have a clear break between them. Wittgenstein and Ramsey had persuaded Russell that logic was tautologous, and this profoundly depressed Russell. Early in life he had thought of mathematics as the study of a shining realm of Platonic entities. It was not nice for him to come to think that mathematics was a matter of saying in more and more complicated ways the same thing, namely nothing. Similarly it now amazes me how at Oxford in my younger days we were complaisant about thinking that mathematics was analytic. To be sure it seemed to solve the epistemological problem but at what a cost! Of course that a view is emotionally repellent is no argument against it, but Quine’s pointing out the clear break gives us a rational argument against the idea that mathematics is analytic. Indeed we can now argue that what was called “logicism” was really a form of Platonism. There is still the epistemological problem, and the indispensability argument plausibly solves that for all the mathematics needed for physics, which is most if not all of mathematics except for the remoter parts of set theory. What about nominalism? There seems to be considerable doubt whether Hartry Field’s nominalistic program can be achieved. He wants to show that all of physics could be stated without the mathematics, which he sees as a mere convenience. Even if it can, it still leaves the usefulness of mathematics for physics as a puzzling phenomenon. Moreover Field’s hoped-for achievement would seem to reduce to instrumentalism or fictionalism about mathematics: neither, even if true, a happy thought for the pure mathematician.
114 J. J. C. Smart 3
HARDY’S PLATONISM
I indicated earlier that what I would like would be a viable philosophy of mathematics that would have been liked by G. H. Hardy. Even Quine’s Platonism would no doubt have made mathematics, except as mathematical recreation, too much beholden to physics. Of course we cannot always have what we would like to have. It is no argument for God, freedom, and immortality that it would be horrible if they did not exist. Nor is it an argument against mathematical nominalism, instrumentalism, and fictionalism that we or G. H. Hardy might not like them. Nevertheless, if one does not feel emotionally satisfied by any of the extant philosophies of mathematics, this may be useful in motivating us to try harder to think of a philosophy of mathematics that is philosophically viable and which happens also to be one that we would like. And I would like one that I think that Hardy would have liked. (His own attempts to express such a philosophy are not in my opinion such as to fi ll the bill.) I think that Hardy believed in the autonomy of pure mathematics. At any rate he hoped that none of his discoveries would be of any use. Partly, but only partly, I think, this was because he worried that any useful discovery might be used in war. He was not much interested in applied mathematics. He said in his A Mathematician’s Apology (1940)1 that quantum mechanics was all right because it would never be of any use, just when unknown to him it was being used to plan the atomic bomb. He respected Eddington (who wouldn’t?), 2 and indeed astrophysics does not seem to be a potentially warlike pursuit. Eddington himself was a Quaker. I think that on the whole Hardy comes out as a Platonist about mathematics3 and, of course, not a Quinean one. He would not have liked to think of mathematics as beholden to physics as Quine implied that it was. Hardy at one time showed much interest in the Principia Mathematica of Whitehead and Russell, the fi rst volume of the fi rst edition of which he reviewed in the Times Literary Supplement (1911). He was interested in Russell’s theory of types, which he did not accept. He published a lecture on Hilbert’s logic in the Bulletin of the American Mathematical Society (1929a). 4 He had discussions with Wittgenstein and F. P. Ramsey. Despite this interest I at least wonder if he liked logicism as suggesting that pure mathematics is a way of devising more and more complex ways of saying the same thing, namely
1 It was reprinted by Cambridge University Press in paperback with a fi ne biographical introduction by C. P. Snow in 2000. 2 Perhaps with hindsight the “E=mc2” in special relativity could have caused some foreboding. 3 See for example Hardy (1929b, 29–30). 4 I am indebted for useful biographical information about Hardy to I. GrattanGuinness who kindly sent me a preprint of his (2001).
Mathematical Platonism
115
nothing. However it did not seem to depress Ramsey who despite being a leading mathematician was philosophically a pragmatist. In correspondence with me Alan Musgrave described Hardy as a bad philosopher. Of course Musgrave does not like Platonism, but more important, it is easy to forget how confused even good philosophers were in Hardy’s day. Think for example of the horrendous use–mention confusions that vitiate much of Principia Mathematica, great and historic contribution to logic and set theory though it is. Hardy indeed respected philosophers and did not think of himself as a professional expert in their subject. Also he had certain qualms about writing about mathematics rather than doing it. He admired poets and painters but despised critics. Sometimes he advocated for its beauty what he called “real mathematics.” Useful mathematics, he thought, is typically not beautiful. Even Littlewood, he said, could not make ballistics beautiful. (Littlewood worked on ballistics, as a junior officer unambitious for military promotion, during the Second World War. 5) I do not think that Hardy would have liked fictionalism about mathematics. However he sometimes praised mathematics as a wonderful construction of the human mind. Well, even though it is fiction George Eliot’s Middlemarch is a great construction of the human mind.6 On the other hand we could point to the great theories of physics as wonderful constructions of the human mind, and yet they are about reality. I therefore do not see this aesthetic praise of mathematics as incompatible with a Platonistic realism. This is unlike the case of music which is also a great construction of the human mind but which has aesthetic appeal only. Would Hardy have liked the idea that he was writing fiction? (Moreover unless, or even if, Hartry Field’s program succeeds there is a problem of accounting for the mysterious fact of the applicability of mathematics in physics.) Hardy’s most defi nite statement of mathematical realism is when he compares a mathematician with an observer who makes out different peaks on a distant mountain range and proofs as like tracing a route from one mountain peak to the next (1929b, 18). Even his saying that a mathematician makes constructions out of ideas need not be taken antirealistically: the ideas may still refer to reality. In any case, for familiar reasons “ideas” here cannot be understood psychologistically. As abstract ideas they themselves may have to be thought of platonistically. Russell at one time prized mathematics because of the shining mystical attraction
5 Grattan-Guinness judged this remark to be snobbish and insensitive. I do not read it in this way. I do not see Hardy as denying that it is praiseworthy to undertake less pleasant work when it is one’s patriotic duty to do so, though perhaps his pacifism might possibly have led him to do so in this instance. 6 Of course the novel is not unrelated to social and political reality in the nineteenth century, into which it gives one insight, and so my example might not be felt quite persuasive.
116
J. J. C. Smart
of Platonic objects. When he was (wrongly) convinced by Wittgenstein and Ramsey (and Ramsey was a fi ne mathematician) that mathematics is tautologous he lost interest in mathematics. So I do wonder that sixty years ago I was so unworried by the thought (whether true or not) that mathematics consists of analytic propositions.
4
PURE MATHEMATICS AND RELIABLE KNOWLEDGE
From time to time mathematicians also introduce defi nitions which function much like axioms. Consider the defi nitions in the main part of mathematics which is called “analysis,” and so without reference to geometry, of the real variable functions sine and cosine, and also the hyperbolic functions sinh and cosh. Similar but more complicated definitions can be made in the theory of functions of a complex variable. It seems right to take plausibility in the light of total science as an important indicator of metaphysical truth. The hypothetico–deductive method of science seems to be the most reliable method of getting knowledge. But doesn’t the method of pure mathematics give us reliable knowledge? Why not take it at face value, as Penelope Maddy at one time did, as telling us about Platonic entities? I think that the diffi culty is that of giving a plausible naturalized epistemology for it, as Maddy has tried to do. The epistemological diffi culty points the way to a Quinean version of Platonism. Maddy’s position was slightly different, giving more autonomy to mathematics but I think at a cost to the plausibility of the epistemology (Maddy 1990).7 Nevertheless, as I said, I feel that mathematics ought not to be beholden to physics. Mathematics must be pure, we feel. But then we may be left with a philosophy of “If . . . then . . .”, and all mathematical truths reduce to logical ones and hence to more and more complex ways of saying nothing at all. Somehow this is hard to believe and certainly not pleasant to believe. Quinean Platonism still beckons intellectually, even though like Hardy (as I suppose) we might not like the idea of mathematics being essentially beholden to physics. However let us explore the possibility of going one better and defending a Platonism that deals with the epistemological problem and which is not indebted to physics. The natural sciences have surely established themselves as having the most reliable methodology for attaining knowledge about our wonderful universe, and they should be reliable pointers to metaphysical truth. In the same way should we not take the ontological claims of pure mathematics at face value? We might not worry about how we get mathematical knowledge but might worry about how we get the
7
She retreats from realism a bit in her (1997).
Mathematical Platonism
117
necessary concepts by which we can make the mathematical claims. In taking mathematics ontologically at face value, I am in part echoing David Lewis’s sarcastic remarks about the sad history of philosophical putative discovery compared with the secure edifice of mathematics. Thus it was important for him to claim in his Parts of Classes (1991, 59) that with his use of plural quantification plus mereology he recovers the whole of classical set theory and hence enough (I would want to say) of it to recapture algebra and analysis. Scientific method seems to be the most reliable method of getting knowledge of our universe, and so plausibility in the light of total science and especially physics is the best pointer to metaphysical truth. I can hear a philosopher asking “What science? Don’t theories get overturned?” Well we cannot of course predict the fundamental physics of the future. If string theory works out (though I gather that it probably does not) then it will greatly change our way of looking at fundamental particles. Still most of workaday physics will remain intact. It will hardly affect chemistry, still less biology and neuroscience. Galen Strawson has recently argued with brilliant rhetoric and argument that what I would call nonphysical qualia will come to be seen as emergent but physical properties (2006a, 2006b). I regard this as most implausible and of course have argued philosophically that our belief in such properties is due to confusion. See also D. M. Armstrong’s great and commendably brief paper “The Headless Woman Illusion and the Defence of Materialism” (1968b). Now why should we not, in the spirit of our acceptance of plausibility and in the light of the empirical sciences, accept plausibility in the light of the great nonempirical science of pure mathematics and take its referring expressions at face value, namely as referring to eternal (not sempiternal) entities? The suggestion is that we should believe in the Platonic entities because the well-established edifice of pure mathematics through its use of objectual (as opposed to substitutional) quantification purports to refer to them. No spooky gazing at Platonic forms as supposed (most of the time) by Plato. Of course there should be a naturalistic account of how earthly creatures such as us could have attained such an edifice. Here I shall pretty well identitfy mathematics with analysis. Thus Frege’s Grundlagen (1968) deals only with elementary number theory. He has a notion of “concept” whereby concepts are nonpsychological, and here he does not seem to me to mean anything different from “set.” He does not seem, like Wittgenstein, to identify a concept with a use of a word. I like to think of J. J. Thomson and Dirac talking about electrons. There might be sentences that one but not the other believes, but if there is a sufficient proportion of sentences with respect to which they each assent or dissent, we could say that they both have the concept of an electron. This is clearly not Frege’s sense of “concept,” which I suspect is not importantly different from that of “set.” Frege thought that a concept is not an entity, but how could anything not be an entity? Again, Bob Hale and Crispin Wright in the collection of their essays
118
J. J. C. Smart
defend a neo-Fregean antirealism about elementary arithmetic, but when it comes to analysis Hale seems to revert to classical set theory (2001).
5
NATURALISTIC ORIGINS OF MATHEMATICAL KNOWLEDGE
If one is a naturalist in metaphysics generally but Platonist about mathematics one will need a naturalist account of how the edifice of mathematics could have come about. We might begin with the most primitive beginnings in which protomathematical abilities could have arisen by natural selection. I think that it has been claimed that certain birds can distinguish between five pebbles and four pebbles or (more likely) between five eggs in the nest and four eggs. Something similar must surely have been the case with our mammalian ancestors. Perhaps they do not have any perception that if you take away one egg from a nest with five there are only four or some such behavioral equivalent of the idea that five is the successor of four. It is likely that natural selection could have led to such an ability, and supposing that intellect arose by natural selection too (as indeed it surely must have), the notion of “successor” could have arisen in a similar way. Then we were on the way to the arithmetic of natural numbers, Peano’s axioms, and further developments. Of course there are many sequences and successor relations that satisfy Peano’s axioms. This is a bit of an aside, but I shall here put forward a rejection of structuralism. The numbers 12, 14, 16, . . . satisfy Peano’s axioms as well as do 0, 1, 2, 3, . . . . However, if we are concerned with finite sections of the sequences all we need is first order logic with identity. Thus “There are five eggs” can be translated into a sentence in first order logic with identity. It would be a rather long sentence, and the translation of “There are five million eggs” would be beyond ordinary human powers. With very small numbers we do have the required ability. Extrapolation to infinity would then be quite natural and unmysterious. Another consideration is a perhaps Wittgensteinian one. We should consider the use of number words in ordinary language. This obviously goes with 0, 1, 2, . . . and not with 12, 14, 16, . . . . In any case this is perhaps a diversion from my main theme because I see structures and patterns as set theoretic and so Platonic entities. If a bird, say, is able to react differently and appropriately to the sight of four and five eggs in the nest, it is well on the way to having a protoconcept of “set.” Indeed, since earthly creatures like us do attain to the beautiful concepts of mathematics, or for that matter to the amazing flights of fancy (and yet testable fancy) of theoretical physics, there must be a naturalist explanation of how we do it. I am not defending a traditional or supernatural Platonic epistemology. This does not of course mean that a developmental psychologist or neuroscientist could have predicted Einstein’s discovery of the general theory of relativity. To do this they would have to have had the genius and background knowledge of Einstein. Still, we expect that in the abstract a purely natural explanation must exist.
Mathematical Platonism 6
119
FICTIONALISM
Given the possibility of a naturalistic epistemology let us take mathematics as a going concern and a secure source of knowledge. It would emerge in many plausible steps from eggs in the nest to actual counting and to the defi nition of zero. Then would come the rationals and infi nite sequences and series of them and so to the real and the complex numbers and sets of these. At every stage rigorous defi nitions following plausible thinking and all in a way compatible with a naturalist epistemology. If the ensuing mathematics purports to refer (via objectual quantification) to Platonic entities, as it surely does, why should we not take it at face value? The idea is to ascend by an empiricist epistemology to a nonempiricist ontology of mathematical objects. I am aware that I have given only hints of the possibility of this, and there would need to be a lot of hard work. Even so, I know that many hardheaded nominalists, fictionalists, and extreme empiricists will resist this quasi-epistemological approach to ontology. The most plausible of these seems to be fictionalism. Now consider (for example) the conclusion that e and π are transcendental. See the short and elegant proof of this by Alan Baker (1975, 3–5). This proof uses no unfamiliar mathematical tools but is extraordinary in its ingenuity. It is hard to feel fictionalist about it. Prima facie it not only justifies but tells us something exciting and true about these numbers. In passing I would like to make the following point. Fictionalists often use a sentential operator “in the story” when stating their position. I think that this is wrong, just as treating “he says that,” “he believes that,” “he desires that,” and “it is necessary that” as sentential operators is wrong. The extensional “it is not the case that” is of course all right. I am of course following Donald Davidson (1968) in treating “that” as a demonstrative referring to a sentence or proposition “in the offing.”) We should treat fiction as pretense history or whatever. Roughly, Sir Walter Scott and George Eliot were not really operating with a Meinongian semantics but were pretending to operate with a Tarskian semantics. (I oversimplify because historical novels contain many sentences that are true or, occasionally, erroneous history not presented as pretense.) So we may see the fictionalist philosophy of mathematics as the view that mathematics is pretense Platonism. I suggest that this is highly implausible.
7
CONCLUSION
Though G. H. Hardy was apparently a Platonist in his (1929b), he later characterized mathematics as a great and beautiful construction of the human mind (1940). So perhaps he wavered. However the two characterizations are not incompatible. Gibbon’s Decline and Fall of the Roman Empire is full of mostly true historical statements and is also a great construction of the human mind, no less than George Eliot’s Middlemarch.
120
J. J. C. Smart
I am therefore inclined to suppose that Hardy was explicitly Platonist and that he saw pure mathematics as building a true description of the Platonic realm of numbers, etc. Still, we must be careful. I do not wish to commit the fallacy of appeal to authority. (Not that appeal to authority should always be irrational. Quite the opposite. We do have to rely on what experts tell us. Thus politicians who express skepticism about the reality of climate change despite the scientific consensus may be accused of ignorance or else of vote-getting spin doctoring.) However in philosophy we should not accept the word of experts except insofar as we are concerned with scientific plausibility. In the present context I admit that greater mathematicians than Hardy have come out with extraordinarily antirealist views. Thus David Hilbert has said that “Mathematics is a game played according to certain rules with meaningless marks on paper.”8 It is therefore the case that nominalists, fictionalists, and the like may feel that I have been selective in applauding G. H. Hardy’s Platonism. Nominalism (if it can be got to work) and fictionalism have an advantage in their austere ontologies. Platonism scores more easily on semantics which is the account of truth and reference in mathematics. G. Kreisel has been reported as saying that in philosophy of mathematics truth of sentences but not reference is what is important. However, following Donald Davidson, and in effect Tarski, I want to say that it is reference, not picturing of facts, that hooks language on to the world. In the Tractatus, Wittgenstein said that the world consists of facts not of things. This was because he had an inadequate, merely substitutional, not objectual, notion of the quantifiers (see my 1986). Facts are not in my ontology. The word “fact” has its place in epistemology, as when we say “Damn it, that’s a fact.” Colin Cheyne has a causal theory of reference (Cheyne 2001). I do not buy that, though that is another story. I think that Tarski satisfaction does all we need. “Cause” is highly contextual and indeed useful for the natural history of how children learn language, and useful for plumbers and brain surgeons, but with its contextuality it is out of place in theoretical physics as well as in metaphysics. I fear that Colin and for that matter Alan and others will suspect that the Australian sun may have got to my head. If so, let the gentle climate of Dunedin restore sanity!
8 Quoted in Rose (1988). Not a bad statement of the formalism to which Hilbert aspired, at least before Gödel and Tarski.
Part III
Truth and Reality
9
On Creeping Minimalism and the Nature of Minimal Entities 1
Luca Moretti
1
INTRODUCTION
The general tendency or attitude that Dreier (2004) calls creeping minimalism is ramping up in contemporary analytic philosophy. Those who entertain this attitude take for granted a framework of deflationary or minimal notions—principally semantical2 and ontological—by means of which to analyze problems in different philosophical fields—for example, the theory of truth, metaethics, philosophy of language, the debate about realism and antirealism, etc. Let us call a sweeping minimalist the philosopher affected by creeping minimalism. The framework of minimal notions that the sweeping minimalist takes for granted encompasses, for instance, the concepts of truth, reference, proposition, fact, individual, and property. Minimal notions are characterized in terms of general platitudinous principles expressed by schemata like the following (see Dreier 2004, 26): “S” is true if and only if S; “S” is true if and only if “S” corresponds to the facts; a has the property of being P if and only if a is P. In this schemata, “S” and “a is P” stand for sentences satisfying superficial constraints of truth-aptitude (i.e., sentences in declarative form subject to communally acknowledged standards of proper use), and “a” and “P”
1 I am very grateful to Berit Brogaard, Uriah Kriegel, and Tommaso Piazza for valuable discussions and important criticisms upon previous versions of this chapter. A special thanks to Amie Thomasson. 2 In semantics, minimalist theories are often opposed to contextualist theories (see, for instance, Borg 2007). In this chapter when I speak of minimal semantical notions I do not mean this specific and technical sense of “minimal,” though there might be correlations between my sense of minimalism and that sense.
124 Luca Moretti stand for expressions with the syntactical roles of, respectively, a singular term and a predicate.3 Dreier argues that since creeping minimalism has broken through in philosophy, the debate about realism has become more puzzling. For, by appealing to platitudes like the above, the sweeping minimalist seems to be entitled to assert everything the realist can assert (Dreier 2004, 23–31). For instance, in metaethics—the specific field analyzed by Dreier—many sentences satisfy superficial constraints of truth-aptitude; consequently they satisfy the above three platitudes. Given that some of these sentences do meet proper standards of assertibility (e.g., “Hitler is a bad man”), the sweeping minimalist can assert them. Hence, she can join the realist in asserting that these sentences are true (e.g., “Hitler is a bad man” is true), that they correspond to moral facts (e.g., “Hitler is a bad man” corresponds to the facts), that there exist moral properties (e.g., Hitler has the property of being a bad man), and so on. The bewildering consequence is that, when a nonrealist goes sweeping minimalist, she will apparently be entitled to assert everything the realist can assert. It is, however, dubious that this nonrealist will actually be turned into a realist. For, principally, it appears implausible that realism could be obtained in such a cheap way! When nonrealists go sweeping minimalist, it becomes puzzling to explain why they are not realists though they seem to be so. The problem of creeping minimalism is that it threatens to make nonrealism indistinguishable from realism.4 This chapter has two distinct but intimately related purposes: the fi rst is to illuminate the nature of creeping minimalism and its broad implications for the configuration of the debate between realists and nonrealists in metaphysics. I fi nd it natural to develop my interpretation of the phenomenon of creeping minimalism through the discussion of views put forward by three philosophers who—it seems to me—may have contributed more than any other to the development of this philosophical tendency. These philosophers are Paul Horwich, Crispin Wright and Stephen Schiffer. Papers by these authors are often mentioned by sweeping minimalists. The sweeping minimalist is often committed to maintaining that there are entities that fall under minimal notions (e.g., facts, properties, etc.).
3 Dreier (2004) is not explicit on these constraints for the applicability of the platitudes, which—I think—are typically assumed by sweeping minimalists. These constraints qualify the position called disciplined syntacticism that I sketch in §2.2. 4 To quote Dreier’s own words: “Minimalism sucks the substance out of heavyduty metaphysical concepts. If successful, it can help Expressivism recapture the ordinary realist language of ethics. But in so doing it also threatens to make irrealism indistinguishable from realism. That is the problem of Creeping Minimalism” (2004, 26). Although Dreier mentions here only the nonrealist position called irrealism, in other passages he seems to assume that creeping minimalism threatens to also make antirealism indistinguishable from realism. I come back to this issue in §3.
On Creeping Minimalism and the Nature of Minimal Entities 125 Let us call the latter minimal entities. The second purpose of my chapter is to clarify the ontological status of these objects (are they real entities or mere linguistic posits?) and to investigate the bearing of their existence on the fate of the debate between realists and nonrealists in metaphysics. The upshot of my overall investigation will cast some light on the intriguing problem uncovered by Dreier. In the next pages, I will venture answers for the following seven questions: • • • •
What exactly is creeping minimalism? What precisely is a minimal notion? Why is creeping minimalism so captivating and pervasive? Does creeping minimalism affect the configuration of the debate between realists and nonrealists? • What engenders commitment to minimal entities? • What is the ontological status of minimal entities? • Does the existence of minimal entities bear on the fate of the dispute between realists and nonrealists?
The following summarizes my understanding of the phenomenon of creeping minimalism and of its broad consequences for the confi guration of the debate between realists and nonrealists in metaphysics (i.e., my answers to the fi rst four questions above). The sweeping minimalist attitude is essentially one that prompts the generalized use of a framework of minimal, semantical, and ontological notions. This framework is built up through the logical regimentation of a holistic web of semantical and ontological notions implicit in ordinary language. Minimal notions are metaphysically unloaded, as they replicate corresponding folk notions the understanding of which does not require any philosophical training or background. Minimal notions are also platitudinous in the sense that they reflect uncontentious principles implicit in ordinary language. An important function of minimal notions (and of their correlated ordinary notions) is enabling the explicit formulation of blind generalizations. Creeping minimalism is compatible with conceptual pluralism. For the sweeping minimalist can concede that, in certain fields, we possess or can defi ne notions that represent thicker versions of minimal notions; these notions depend on nonplatitudinous principles and may allow for genuine philosophical explanations. The sweeping minimalist attitude is captivating because it relies, ultimately, on familiar notions we already use informally in ordinary discourse. The sweeping minimalist attitude is pervasive because minimal notions are holistically related to one another in the same way as the folk notions that they replicate are holistically related to one another. Accordingly, we cannot use one minimal notion if we do not use many other minimal notions. When creeping minimalism takes place, configuring the debate between realists and nonrealists as a dispute between realists and noncognitivists becomes unfruitful.
126
Luca Moretti
Yet this debate can still be construed as a dispute between realists and antirealists, or—with some restrictions—as a dispute between realists and error-theorists. The following summarizes my understanding of the nature of minimal entities and of the consequences of their existence for the fate of the dispute between realism and nonrealism (i.e., my answers to the last three questions above). Minimal entities are those that fall under minimal notions. Deflationary principles included in the framework of the sweeping minimalist, which replicate platitudinous ordinary principles, ensure ontological commitment to minimal entities. Disallowing such a commitment would impoverish the expressive power of language, as it would prevent the explicit formulation of blind existential generalizations. It is proper to the nature of minimal entities that only deflationary notions of reference can explain our ability to refer to them—these notions of reference can establish no link between language and the extralinguistic world. Minimal entities are thus intrinsic to language. Minimal entities are nothing but reifications or hypostatizations of linguistic constructs. Creeping minimalism is compatible with ontological pluralism because the sweeping minimalist can in principle accept that the world is populated by both minimal entities and, in certain fields, ontologically more substantive entities (where the latter are the extensions of thicker versions of minimal notions that the sweeping minimalist can in principle accept). Although the existence of minimal entities raises minor difficulties for certain forms of error-theory, it has no apparent implication for the acceptability of realism and antirealism. As a result, allowing the existence of these entities has no significant consequence for the debate between realism and nonrealism as long as this debate is construed as a dispute between realists and antirealists. This chapter aims to substantiate the following general thesis: creeping minimalism does not render nonrealism indistinguishable from realism. For when creeping minimalism takes place, it is still possible to distinguish with clarity and precision between realist and nonrealist positions— precisely, positions that entail realism about certain entities, positions that entail antirealism about the same entities, and positions that entail an error-theory about those entities. This also constitutes my answer to Dreier’s challenge. This is the structure of the chapter: §§2 and 2.1–2.3 provide my interpretation of the phenomenon of creeping minimalism and of the notion of a minimal entity via analyzing and “assembling” views put forward by Horwich, Wright, and Schiffer. Precisely, §2 introduces the topic; §2.1 focuses on Horwich’s Minimalist theory of truth and reference; §2.2 focuses on Wright’s notions of minimal truth and of minimal truth-aptitude; and §2.3 focuses on Schiffer’s pleonastic ontology. §3 analyzes the general configuration assumed by the debate between realism and nonrealism once creeping minimalism has taken place. §4 investigates the ontological status of minimal entities and the bearing of their introduction on the debate between
On Creeping Minimalism and the Nature of Minimal Entities 127 realism and nonrealism. §5 draws the conclusions of the chapter, in which Dreier’s problem is addressed.
2
THE NATURE OF CREEPING MINIMALISM
I consider in succession Horwich’s Minimalist theory of truth and reference, Wright’s minimalism about truth and truth-aptitude, and Schiffer’s pleonastic ontology. I do not attempt any substantive defense (or refutation) of these influential philosophical views; any fair and accurate assessment of them would require a chapter far longer than this. Rather, I try to cast light on the theoretical connections among these positions and on their contribution towards the creation of the sweeping minimalist stance. I show that once these three positions are suitably interpreted and developed, “assembling” them together in the right way produces almost naturally creeping minimalism.
2.1 Horwich’s Minimalist Theory of Truth and Reference Although Horwich is not the “father” of the sweeping minimalist stance, he can surely be counted as one of its “grandfathers.” Horwich aims to answer the question: what is truth? The answer comes in terms of the Minimalist (notice the capital letter “M”) theory of truth, presented in Truth (1998a, 1st ed. 1990)5 as a variant of the broader deflationary conception of truth (1998a, ix). Perhaps earlier deflationists such as Frege, Ramsey, Ayer, and Strawson might also be considered grandfathers (or great-grandfathers!) of creeping minimalism. One difficulty with this attribution is that the views of these authors involve—or seem to involve—that the predicate “is true” connotes no property at all.6 This clashes with the creeping minimalist thesis
5 I refer here to the second edition of this book, though the passages I am going to consider have undergone no substantial change or no change at all between the two editions. For further clarification of Horwich’s view, see also Horwich (2001). 6 Frege (1891, 1918), Ramsey (1927), and Ayer (1935, 1936) accepted versions of the redundancy theory of truth. Roughly, they agreed in holding that asserting that S is true (where “S” stands for a sentence) is just asserting that S, and is not asserting something (i.e., a property) of the proposition that S. Strawson (1950) and Ayer (1963) accepted the performative (or also expressivist) conception of truth, according to which the expression “is true” is used, not to make descriptions or attributions of properties, but only to perform quite different speech acts, like endorsing and agreeing. More recent versions of deflationism may have played a role in the development of creeping minimalism. However, this would not seem to be the case with the prosentential theories defended, for instance, by Grover et al. (1975). The prosententialists interpret expressions such as “it is true” and “that is true” as prosentences—namely, as linguistic tools for achieving anaphoric cross-reference to
128
Luca Moretti
that the truth predicate—as well as most (if not all) predicates—connotes a property. On Horwich’s Minimalist theory, the truth predicate connotes “a property of some sort” (1998a, 37) though no substantive property—for example, any complex or naturalistic property (1998a, 10–11, 37) or any type of justification or idealized justification (1998a, 56–62). I return to this point at the end of this section. Horwich believes that the conceptions of truth that presuppose substantive and complex accounts of truth cannot survive any serious scrutiny (1998a, 8–10). The central problem is that for any substantive theory that seems to explain what the truth of the propositions in a given domain is, there will in general be many other domains in which that explanation will face counterexamples. Horwich contends that the truth predicate exists principally to fulfil certain logical functions—basically, to enable the explicit formulation of generalizations in situations of partial information (briefly, the formulation of blind generalizations) (1998a, 2–4, 31–33, 37).7 Suppose I want to endorse what John said to Katie, though I do not know what he said. If I knew it, I could simply reassert what John said, but this is not possible now. I can however express my attitude by stating that what John said to Katie is true. Or suppose that, because I believe that the pope is always sincere and infallible, I intend to accept all the pope’s claims. Although I cannot assert everything the pope said and will say, I can certainly state that everything the pope says is true. On Horwich’s view, what enables truth to have these logical functions is simply the fact that all uncontroversial8 instances of the equivalence schema, [ES] It is true that S if and only if S, where “S” stands for a sentence,9 hold valid. Consider again the pope example. Without appealing to a principle like [ES], to accept all claims of the
sentences asserted formerly (just as pronouns are devices for achieving anaphoric cross-reference to names uttered previously). Prosententialists believe that the expression “is true” is no genuine predicate (see Kirkham 1997, 326–327), with the consequence that it cannot possibly connote any property. 7 Without the truth predicate, this logical function could be satisfied by a complex language that allows for substitutional quantification. Horwich explicitly admits that the truth predicate, strictly speaking, could be eliminated from our language (1998a, 124–125). Yet the truth predicate allows for a language that is simpler than it would otherwise be (1998a, 4 n. 1, 31–33). 8 Uncontroversial instances are those that are nonparadoxical or in some sense “pathological.” Typically, paradoxical instances of [ES] are those associated with the liar paradox (Horwich 1998a, 40–43). 9 Precisely, to include true sentences inexpressible in current English, Horwich asserts that “S” stands for sentences of any possible extension of current English (1998a, 18 n. 3).
On Creeping Minimalism and the Nature of Minimal Entities 129 pope, I would need to use a turn of phrase that alludes to a potentially infinite conjunction of conditionals such as: If the pope says that Sydney is big, then Sydney is big, and if the pope says that my pizza is indigestible, then my pizza is indigestible, and if . . . and so on. Through [ES], the potentially infi nite conjunction becomes equivalent to: If the pope says that Sydney is big, then it is true that Sydney is big, and if the pope says that my pizza is indigestible, then it is true that my pizza is indigestible, and if . . . and so on. Now, we can express the same content as above by a finite universal generalization, that is: For every x, if the pope asserts that x, then it is true that x. More colloquially: everything the pope says is true. Similarly, by appealing to [ES] we can express the potentially infi nite disjunction John said that S 1 to Katie and S1, or John said that S 2 to Katie and S 2 , or John said that S 3 to Katie and S 3, or . . . and so on, in terms of the following finite existential generalization: There exists an x such that John said that x to Katie, and it is true that x. Colloquially: what John said to Katie is true. Horwich takes truth to be a property of propositions: namely, the referents of the that-clause in [ES] (1998a, 16–17).10 Horwich considers his theory to be minimal in the sense of being conceptually or metaphysically minimal: truth has, according to it, a minimal “essence” or “nature” (1998a, 6). For the entire conceptual role of truth can be explicated on the sole grounds of the uncontroversial instances of [ES] (1998a, 5, 20–23), where the latter constitute all and only the axioms of the Minimalist theory of truth (1998a, 6, 11, 17–20).11 Truth does not have any hidden nature or deeper structure awaiting our discovery: the
10 The precise metaphysical nature of propositions is largely irrelevant for the formulation of the Minimalist theory (1998a, 16–17 and ch. 6). Horwich also formulates a version of the Minimalist theory for utterances (1998a, 98–103). 11 Clearly, the Minimalist theory cannot be explicitly formulated but only implicitly specified (Horwich 1998a, 11, 20, 25–31).
130 Luca Moretti Minimalist theory provides a complete explanation of this notion and of the correlated property (1998a, 4–5, 11, 23–25). In particular, there is no explanation (in any interesting sense of “explanation”) of why [ES] holds good. [ES] is conceptually basic (1998a, 50–51).12 The Minimalist theory of truth does not defi ne truth contextually in terms of other notions—for instance through the notions of reference or of a fact. Yet Horwich believes that truth and reference are so intimately related “that the rationale for a minimal account of truth will equally well motivate a minimal account of reference” (1998a, 113). The Minimalist account of reference, sketched in 1998a, is fully developed in Meaning (1998b). Horwich argues that the expression “refers” connotes no complex property (or relation) and that any attempted naturalistic or conceptual reduction of the notion of reference is probably doomed to fail (1998a, 113–117, 1998b, 115–118). Horwich is persuaded that the expression “refers” principally exists just like the predicate “is true”— because it simplifies our language (1998a, 115, 1998b, 121–123). The use of “refers” is fi xed by a schema the instances of which hold trivially valid; precisely: [R] For every x, the concept of a refers to x if and only if a is identical to x, where “a” stands for a singular term, and the concept of a is the propositional component expressed by “a” (1998a, 116, 1998b, 118–120). [R] is explanatorily basic just as [ES] is (1998a, 116 n. 6, 1998b, 120).13 An important function of the notion of reference is—by analogy with [ES]— that of enabling the explicit formulation of blind generalizations.14
12 The Minimalist theory gives no analysis of the notion of truth in terms of sufficient and necessary conditions (Horwich 1998a, 10, 20, 33–36) but provides a defi nition of the truth predicate in terms of use by specifying the basic regularity that governs its overall deployment (1998a, 34–35). (This conception of meaning is fully developed in Horwich’s 1998b). The meaning of the truth predicate is fi xed by our disposition to accept all uncontroversial instances of [ES] and to use this predicate always in accordance with that disposition (1998a, 33–36). 13 Horwich contends that the meaning of the expression “refers” is fi xed by our disposition to accept the instances of R and to use “refers” always in accordance with that disposition. Horwich (1998a, 116, 1998b, 29) suggests an analogous deflationary characterization of the notion of being true of (or being satisfied by). Putnam (1978, 128) has put forward a similar deflationary notion. 14 Suppose that, though I know nothing about what John said, I believe he referred to Plato. Suppose I want to express my belief. If I had no notion of reference (nor any equivalent notion such as being true of or being about), I would need to assert a potentially infi nite disjunction of propositions like the following:
On Creeping Minimalism and the Nature of Minimal Entities 131 The principle [R] can be reformulated as follows to apply directly to singular terms: [R*] For every x, “a” refers to x if and only if a is identical to x. where «“a”», on the left-hand side stands for the name of a singular term type exemplified by the replacement for the token “a” on the right-hand side.15 The contribution by Horwich and, more generally, by deflationists about truth to the creation of the creeping minimalism framework comes in terms of the substantiation of the thesis that the notion of truth can be characterized by principles as elementary and platitudinous as [ES]. Deflationism about truth has probably been the seed from which the plant of creeping minimalism has sprung up. For the idea that the notion of truth is platitudinous is suggestive of the sweeping minimalist’s conviction that many other important notions in philosophy are also so. Indeed, Horwich himself has made a crucial step to substantiate this conviction by putting forward his Minimalist notion of reference. In §2.3 I suggest that this notion (or one strictly analogous) is also part of the sweeping minimalist’s framework, and it is vital to ensure ontological commitment to minimal entities in general.
The proposition asserted by John embeds the concept of Socrates’ greatest student and Socrates’ greatest student is identical to Plato, or the proposition asserted by John embeds the concept of the writer of Republic and the writer of Republic is identical to Plato, or . . . and so on. By appealing to [R], this potentially infi nite disjunction becomes equivalent to the following: The proposition asserted by John embeds the concept of Socrates’ greatest student and the concept of Socrates’ greatest student refers to Plato, or the proposition asserted by John embeds the concept of the writer of Republic and the concept of the writer of Republic refers to Plato, or . . . and so on. This potentially infi nite sentence can now be explicitly expressed by a fi nite existential generalization, that is: There exists a concept that is part of the proposition asserted by John and that refers to Plato. Colloquially: John spoke of Plato. Without the Minimalist notion of reference, this function could be satisfied by allowing substitutional quantification (Horwich 1998a, 115–116). Further functions of minimal reference are illustrated in Horwich (1998b, 121–123). 15 To cope with issues of context-sensitivity and other difficulties, the valid use of [R*] should be constrained by further stipulations that I leave implicit here.
132
Luca Moretti
Finally, Horwich’s explanation of the logical function of truth and reference as devices to enable blind generalizations can be generalized to clarify an important function of the whole framework of sweeping minimalism. I will suggest in §§2.2 and 2.3 that other important minimal notions can be thought of as having the same logical role. One reason why Horwich’s deflationism harmonizes with the spirit of creeping minimalism is that it embeds the claim that the truth predicate connotes a property, though no substantive property. This is the precise sense in which Horwich is inclined to think of truth as a property: “Is true” is a perfectly good English predicate—and (leaving aside nominalistic concerns about the very notion of “property”) one might well take this to be a conclusive criterion for standing for a property of some sort. (Horwich 1998a, 37; see also 141–143) On this basis, Horwich would probably assert that his deflationary notion of reference identifies a corresponding thin property (or relation) too. This notion of a property would seem to coincide with the deflationary notion of a property elicited by the sweeping minimalist attitude and characterized in terms of only platitudinous principles (more on this in §2.2). Deflationism about truth can be targeted with various objections. For instance, it can be argued that on a deflated notion of truth it becomes impossible to explain why truth has intrinsic value independently of its practical utility; how true propositions can correspond to external facts; how the truth of a proposition depends on the referential properties of its parts; how truth can play any explanatory role—for example: explaining the achievement of practical goals, explaining the empirical success of scientific theories, explaining the nature of logical deduction, and so on. Horwich’s defensive strategy is articulated: he argues that some of these features do not pertain to truth at all and that some other features are actually possessed by Minimalist truth. Moreover, some of the functions traditionally attributed to truth would be explainable, for Horwich, on the grounds of the Minimalist theory in conjunction with other theories—such as a theory of human psychology, a theory of reference, a logical theory, and so on (1998a, 20–25 and chs 3–7).16 It is difficult not to be skeptical when faced with these claims. For it is implausible—if not incredible—that a notion of truth as vacuous as the Minimalist’s can actually have any of the substantive properties traditionally attributed to truth and can actually fulfil any of the substantive functions typically attributed to truth—if the
16 For criticism of Horwich’s Minimalism see, for instance, Dodd (2002), Gupta (2001), O’Leary-Hawthorne and Oppy (1997), Kirkham (1997, 339–349), Field (1992), and Devitt (1991).
On Creeping Minimalism and the Nature of Minimal Entities 133 fulfi lment is claimed to obtain after conjoining the Minimalist theory of truth with another theory, it is reasonable to expect that the latter, and not the Minimalist theory, will do all of the essential jobs. Could (a version of) Minimalism accommodate a more substantive conception of truth? I want to suggest it can. Horwich is a monist about truth. The Minimalist theory is meant to characterize a notion of truth that is assumed to be the only one we possess, and this thin notion is unsuitable to satisfy substantive functions and properties that many philosophers believe truth has to fulfil. An irenic attitude might be that of embracing some form of pluralism about truth. For example, a pluralist might concede that the Minimalist notion of truth is nothing but the platitudinous notion that we only use for the unsubstantial functions described by Horwich while recognizing that we may also possess further notions of truth, proper to certain areas of discourse, that fulfil more specific and substantive functions. (Consider, for instance, a form of pluralism according to which the notion of truth proper to, say, moral sentences is just platitudinous, whereas the one suitable to scientific sentences reflects a complex and naturalistic property.) This view is not prima facie absurd or implausible. Truth is certainly a fundamental notion in philosophy, but pluralism about key notions is customarily endorsed in many intellectually respectable disciplines. In physics, for instance, there are various notions of energy (kinetic, potential, gravitational, etc.) and various notions of force (electric, magnetic, electromagnetic, gravitational, etc.).17 In the same way, in philosophical logic, various notions of necessity (logical, nomological, metaphysical) are typically accepted. Prominent metaphysicians of the past were also pluralists about other central notions—for instance, Aristotle, Aquinas, and other great figures of scholasticism were pluralists about the very notion of being. A form of pluralism might thus apply to the notion of truth too.18 Although Horwich would most likely not accept any interpretation aimed at reconciling his own view with alethic pluralism, I suggest that this
17 Often, scientific notions belong to the same family because they share the same central principles. For instance, in classical physics, any force must satisfy the second principle of dynamics. We obtain gravitational force by adding Newton’s law of gravitation, alternatively we obtain electric force by the addition of Coulomb’s law, and so on. To produce pluralism about truth, the equivalence schema might be assumed to play the most central role, in analogy with the second principle of dynamics. In this view, any notion of truth will have to satisfy [ES]. Different notions of truth will obtain after adding further qualifying principles to [ES]. (I am in debt to Huw Price for suggesting this analogy). This form of pluralism does not entail the oddity that the notion of truth is equivocal or ambiguous but, rather, that it is susceptible of different specifications. 18 For a survey of the problems of contemporary pluralism about truth, see Lynch (2004).
134
Luca Moretti
is the reading of Truth that many sweeping minimalists adopt or tend to adopt. The sweeping minimalist attitude does not elicit the belief that there is just one notion of truth, which is a minimal notion. Rather, the sweeping minimalist attitude seems to rely on the assumption that we possess a web of minimal notions that includes the notion of truth. This is prima facie compatible with the possibility that we also possess more substantive notions of truth (and of reference). Besides, consenting to some form of pluralism about truth will shield sweeping minimalists from some of the objections that trouble Horwich’s Minimalism.
2.2
Wright’s Minimal Truth and Minimal Truth-Aptitude
Pluralism about truth of the sort just outlined appears to be a component of Wright’s minimalism. Wright is a closer parent of the sweeping minimalist stance than Horwich (if parenthood can be a matter of degree). In Truth and Objectivity (1992), Wright puts forward a broad philosophical perspective called minimalism (notice the small “m”), which has further been clarified especially in Wright (2001).19 Wright’s minimalism is “an account kindred to the spirit of [Horwich’s] own [Minimalism]” (1992, 23 n. 15). In particular, Wright accepts—as one of the central ingredients of minimalism—an enriched version of Horwich’s theory of truth as characterizing the most elementary or minimal concept of truth we possess (Wright 1992, 13 n. 13, 21 n. 15). This concept of truth corresponds to a deflationary or thin property of truth in the same sense in which Horwich’s Minimalist truth corresponds to a deflationary property (Wright 2001, 753–754, 783 n. 5). Yet whereas Horwich is a monist about truth, Wright is a pluralist: he contends that there are or may be different truth predicates additional to the minimal truth predicate, and thus various thicker notions of truth proper to different areas of discourse (Wright 1992, 24–25, 37–38).20 To qualify as a notion of truth, any predicate ought at least to satisfy the requisite indicated by Horwich—that is, the instances of [ES]—and further basic principles—or platitudes—about truth (Wright 1992, 38). These requirements characterize the minimal truth predicate. Wright’s minimal notion of truth
19 A significant theoretical change between the two papers is that, in the fi rst minimalism is described as entailing pluralism about both the notion and the property of truth (see, for instance Wright 1992, 25), and in the second, minimalism is claimed to entail pluralism about only the property of truth but not the notion of truth (Wright 2001, 751–754). I fi nd the latter position puzzling and—as far as I can understand it—implausible. See also below, note 22. 20 Wright (1992, 12–24) delivers an argument to the effect that Minimalism— and, more generally, deflationism about truth—is “an inherently unstable position.” For a reply see Khlentzos (2004, 25–29). Horwich himself has countered this objection in the Postscript of Truth (see 1998a, 143–144).
On Creeping Minimalism and the Nature of Minimal Entities 135 is thus not conceptually (or metaphysically) minimal in the same sense as Horwich’s is. For Wright’s minimal notion has a more complex, though platitudinous, structure. Yet Wright’s notion can still be considered to be conceptually minimal in the sense of being the most elementary of a plurality of notions of truth that (according to Wright) we do or can possess. The focus of Truth and Objectivity is on the issues of realism and antirealism (1992, 1). Following Dummett and Putnam, 21 Wright believes that the features of the notion of truth that best suits a given area of discourse determine whether that discourse is to be considered realist or antirealist. Wright suggests that the debate between realism and antirealism will properly be developed only after both parties have accepted the metaphysically neutral framework of minimalism, which allows for the defi nition of the minimal notion of truth and of the correlated minimal property of truth (1992, 27, 13, 76). Wright believes that metaphysically thicker notions of truth can possibly be obtained and that the debate about the realist commitment of a given discourse should be developed as a debate concerning the presence or absence in that discourse of further features that constitute these more substantial notions of truth (1992, ch. 3, 2001, 752).22 For example, the additional feature of truth of being potentially evidence-transcendent would give the discourse to which such a notion of truth applies a realist connotation (1992, 77–78). 23 To appreciate Wright’s crucial contribution to the development of the sweeping minimalist attitude, we need to go deeper into the framework of
21
Mainly Putnam (1981). Wright (2001, 751–754) claims that pluralism concerns the property of truth but not the notion of truth. Wright maintains that the same platitudinous notion of truth may denote different properties in different contexts and that what counts as a thick property of truth in a discourse must be something that explains why the truthplatitudes hold for that discourse. Wright draws an analogy between the concepts of truth and of water. The surface features of water typically described by the concept (or Putnam’s [1978, 98] stereotype) of water (e.g., being a colorless, odorless, tasteless liquid) are explained by the property of being composed of H 2O molecules (with auxiliary assumptions). Analogously, the property of a sentence “S” corresponding to a given mind-independent fact could explain why “S” satisfies truth-platitudes. Wright admits, however, that this analogy is imperfect because investigating the nature of water is an a posteriori task, and investigating the property of truth for a given discourse is a matter of “further conceptual reflection” (2001, 753; Wright’s emphasis). What is puzzling is how it could be conceptually true that the notion of truth can denote different properties in different discourses if this variability is not reflected by the notion of truth itself (see also Lynch 2004, 387). 23 The intuitive idea behind this is that if a sentence proves evidence transcendently true, what makes the sentence true is something that is independent of our cognitive operations (Wright 1992, 4). 22
136
Luca Moretti
minimalism. For Wright, the mere acceptance of [ES] does not suffice to characterize in full the minimal notion of truth; for this purpose, further principles are to be included (2001, 751). The minimal notion of truth is contextually defi ned by a set of “very general, very intuitive a priori laws” (1992, 72) that “chime with ordinary a priori thinking about truth” (2001, 759), which Wright calls truth-platitudes (2001, 759, 1992, 34–35). Some of the truth-platitudes are expressed by schemata; others are expressed by well-formed sentences. Wright advances no comprehensive list of truth-platitudes (1992, 35, 72; see also 2001, 760) but includes among them [ES]24 — which holds for propositions—and the version of [ES] for sentences, or disquotational schema, namely: [DS] “S” is true if and only if S, where «“S”», on the left-hand side stands for the name of a sentence type exemplified by the replacement for the sentence token “S” on the righthand side.25 Other truth-platitudes are, for instance, the following: [a] The proposition that S is true if and only if it is true that S;26 [b] Asserting the proposition that S is asserting that the proposition S is true;27 [c] “S” is truth-apt if and only if “not-S” is truth-apt;28 [d] Some proposition may be true and not justified, and vice versa; [e] “S” is true if and only if “S” corresponds to the facts; [f] Truth is absolute (i.e., not a matter of degrees). (See 1992, 35, 72, 2001, 760.)29
24 Wright appears less prone than Horwich to reject (apparently) problematic instances of [ES]. See Wright (1992, 61–64) and (2001, 782 n. 1). 25 It is unclear whether Wright considers “S,” in [DS], to be a Quinean “eternal sentence.” If he does not, the valid use of [DS] should be constrained by stipulations that settle arising problems of context sensitivity. 26 It is not completely clear to me whether Wright believes that the that-clauses refer univocally to propositions. At any rate, the fact that Wright explicitly considers [ES] as a schema for propositions commits him to [a]. 27 More generally: any propositional attitude to the proposition that S is an attitude to its truth (Wright 2001, 760). 28 More generally: aptitude for truth is preserved under standard logical operations (Wright 2001, 760). 29 Wright also considers the platitude—which he believes to be “more contentious” (1992, 72) than others—stating that truth is stable (i.e., if something is true, it is always so). In this chapter I lay no great weight on this truth platitude as defi nitive of minimal truth. If this platitude is accepted, a suitable antirealist
On Creeping Minimalism and the Nature of Minimal Entities 137 In [DS] and in the other principles above, “S” stands for a sentence that counts as at least minimally truth-apt (Wright 1992, 27–28). I elucidate this notion below.30 Wright believes that a set of platitudes about truth, which “collectively constrain and locate the [truth predicate] and that sufficiently characterize some of its relations with other [predicates] and its role and purposes” (2001, 759), will adequately illuminate the notion of truth. 31 Importantly, predicates or notions other than that of truth that figure in the platitudes are also conceived of as deflationary. Consider for instance platitude [e]. For Wright, the expression “the facts,” in it, could harmlessly be replaced with “reality,” and the whole of [e] could be rephrased for instance like this: “S” is true if and only if reality is as “S” says it is, without any loss of content (1992, 25–27). If the notions of a fact and of correspondence were characterized in terms of precise metaphysical theses, replacements and reformulations of this sort would result in some loss of content. In developing the theoretical framework of minimalism, Wright has broadened the family of defl ationary notions to include at least the notions of a fact, of correspondence with reality, and of truth-aptitude. 32 This enlargement is likely to have been the spark that triggered a chain reaction: as a result of the reception of Wright’s minimalism, philosophers may have started believing that a whole web of defl ationary notions somehow stems from a deflationary notion of truth (I will outline an argument
notion of truth can plausibly be thought of in terms of superassertibility (1992, 44–70), where a proposition is superassertible if and only if its assertion is “justified by some (in principle accessible) state of information and [would remain] justified no matter how that state of information might be enlarged upon or improved” (1992, 47). 30 There are logical links between some of these platitudes. For example, each instance of [DS] logically follows from the corresponding instance of [ES] in which the same sentence S occurs in conjunction with the platitudinous assumption that “S” is true if and only if what “S” says (or expresses) is true and the assumption that “S” says that S (Wright 1992, 24). Similarly, the instances of [b] are entailed by the corresponding instances of [ES] and other platitudinous assumptions (Wright 2001, 670). 31 Like Horwich, Wright is skeptical about the possibility of giving a reductive analysis of the minimal notion of truth, that is, one that provides sufficient and necessary conditions for the application of the truth predicate (Wright 2001, 759–760, 784 n. 15). 32 It is not clear to me whether Wright (1992, 2001) uses the term “proposition” in a deflationary or a thick sense.
138
Luca Moretti
to corroborate this belief shortly). This might have been the most crucial step in the formation of the sweeping minimalist attitude. An influential interpretation of Wright’s position—often attributed to Jackson (1994)—describes minimalism as “a common ground approach to circumscribing the platitudes which constrain [the notion of truth—where] the platitudes are those assumptions which all theorists irrespective of their particular metaphysical commitments are prepared to make” (Divers and Miller 1994, 15). This interpretation of Wright’s view seems to me not fitting. How could we formulate platitudes of this sort? Following suggestions by Burgess (1997, 262) and Divers and Miller (1994, 15 n. 4), the set of a priori principles about truth that “all theorists irrespective of their particular metaphysical commitments are prepared to make” should plausibly include philosophical disjunctions and conditionals that shun any specific ontological commitment; for instance, the following: [i] Fregean thoughts are bearers of truth, or Russellian propositions are bearers of truth, or beliefs are bearers of truth, or sentence-types are bearers of truth, or sentence-tokens are bearers of truth, or utterances are bearers of truth, or . . . (and so on to exhaust all known possibilities). [ii] If there exist Fregean thoughts, then any true Fregean thought denotes the true. A problem is that neither [i] nor [ii] looks like a truth-platitude such as those presented in Wright’s examples, which are much more immediate and easier to grasp.33 The basic reason seems to me this: accepting [i] or [ii] presupposes the knowledge of what specific philosophical doctrines (e.g., the theory of Fregean thoughts) say about truth—knowledge typically requested by any theorist (or, more precisely, metaphysician) of truth—but accepting what Wright calls truth-platitudes does not appear to require any sort of specialist knowledge. I believe that with the expression “truth-platitude,” Wright refers to principles more elementary than [i] and [ii]. Roughly, he refers to those principles about truth that any speaker who can be said to understand the ordinary meaning of the predicate “is true”34 is ready to accept as a priori correct.35
33 Surprisingly, neither Divers and Miller (1994) nor Burgess (1997) seems to notice this problem. 34 Because the predicate “is true” has more than one meaning (consider “this sentence is true” and “this painting is not true”), I specifically refer to the meaning of “is true” ruled by the principle—among others—that to assert something is to assert it as true. 35 This specific interpretation of Wright’s truth-platitudes seems presupposed in Lynch (2004).
On Creeping Minimalism and the Nature of Minimal Entities 139 For example, a speaker who could be said to understand the ordinary meaning of “is true” and had no philosophical training would not typically accept [i] and [ii] as a priori correct, but she would plausibly endorse all instances of [ES] (in which “S” appears to her truth-apt) as a priori correct. In sum, Wright’s minimal notion of truth would seem to aim to provide an elementary logical model (or elementary rational reconstruction) of the folk or ordinary notion of truth, which is not metaphysically loaded.36 The sweeping minimalist attitude elicits massive use of deflationary notions in semantics and ontology. I conjecture that the nature of many of these minimal notions can be characterized in the very same way as Wright’s minimal truth according to my interpretation of it; namely, these notions are elementary regimentations of their ordinary counterparts, which depend on platitudinous principles implicit in our ordinary language. This conjecture—if correct—casts some light on why the sweeping minimalist attitude is almost irresistible and easily creeps up on us: this attitude essentially consists in yielding to the temptation to deploy in philosophy familiar and comfortable notions we already use in ordinary discourse. Consider also that introducing minimal notions of this kind to philosophical discourse requires only a negligible theoretical effort, as these notions are virtually already available.
36 An objection might be that a speaker who understands the ordinary meaning of “is true” and had no philosophical training might not accept the instances of [ES] with “S” replaced with a philosophical sentence, as she might not understand its meaning, though that sentence would appear truth-apt to philosophically trained speakers. This might be the case with the replacement for “S” that yields the following instance of [ES]:
If it is true that there exist Fregean thoughts, then there exist Fregean thoughts. The same problem arises with other replacements for “S” in [ES]—and in other principles about truth—the understanding of which requires some specific training (for instance, scientific). I believe that Wright would accept all or most of these instances as truth-platitudes. To address this problem, my characterization of a truth-platitude should be refi ned, perhaps in this way: a truth-platitude is a principle about truth that any speaker who can be said to understand the ordinary meaning of the predicate “is true” is ready to accept as a priori correct unconditionally or conditionally under the assumption that all its embedded sentences are truth-apt. For example, a speaker who understands the ordinary meaning of “is true” would accept the above instance of [ES] if she were assured that the embedded sentence “there exist Fregean thoughts” is truth-apt. Notice that if the same speaker were provided with the same assurance and also with the assurance that the sentence “any true Fregean thought denotes the true” is truth-apt, she would typically not accept [ii] as a priori correct.
140
Luca Moretti
I will now sketch an argument to the effect that if we possess a minimal notion of truth that aims to be a logical regimentation of its folk counterpart, we will possess many minimal notions of the same sort. It is hard to deny that there are a priori links between the notion of truth and various semantical and ontological notions we use ordinarily—for instance, between the notion of truth and the notions of a proposition and of a fact.37 It appears also correct or very plausible that if a speaker did not acknowledge the existence of many of these a priori links, that speaker could not be credited with understanding the ordinary concept of truth. Such a priori links—perhaps the most central of them—should consequently be interpreted as definitive of or essential to the folk notion of truth. If the minimal notion of truth is a model of the folk notion of truth, these a priori connections should be described by some of the platitudes that contextually defi ne minimal truth. Two platitudes that describe these a priori links seem to be the following schemata: [e*] It is true that S if and only if the proposition that S corresponds to the fact that S;38 [g] For every x, it is true that a is P if and only if there exists an x such that “a” refers to x and x is P, where “a” stands for an expression having the logical (or syntactical) role of a singular term, «“a”» stands for the name of that singular term, and “P” stands for an expression having the logical (or syntactical) role of a predicate.39 If [e*] and [g] are logical reconstructions of principles implicitly accepted by any speaker who understands the folk notion of truth, then the notions of a proposition, correspondence, a fact, and reference appealed to in these platitudes must be as thin as possible. For these notions reflect ordinary concepts the possession of which does not require any sort of special knowledge or training. The semantical and ontological notions appealed to in [e*] and [g] can only be characterized in terms of platitudinous principles: these notions are minimal in the very same sense as minimal truth is. The same reasoning can now be reiterated to apply to some of these
37 Consider that any competent speaker can ordinarily move back and forward between “it is true that it is raining” and “it is a fact that the proposition that it is raining is true.” 38 [e*] is a specification of the platitude [e] in its version for propositions. 39 An interesting issue I cannot address in this chapter concerns the possible restrictions to constrain the proper replacements for “P” in this and other platitudes to avert paradoxes that result from allowing higher order predicates and higher order properties. On this, see for instance Schiffer (1996, 164–166), Schiffer (2003, 67–68 and ch. 5) and Hofweber (2006, 169, 185–187).
On Creeping Minimalism and the Nature of Minimal Entities 141 minimal notions—for example, the one of a fact. A central platitude that regiments the folk notion of a fact is certainly this: [h] It is a fact that S if and only if S. A platitude that regiments an essential link between the folk notion of a fact and the folk notions of an individual and of a property is plausibly the following schema: [i] It is a fact that a is P if and only if the individual (or object) a has the property of being P. If [i] is a logical reconstruction of a principle implicitly accepted by any speaker who understands the folk notion of a fact, then the notions of an individual and of a property appealed to in [i] must be as thin as possible. For these notions reflect ordinary concepts the possession of which requires no special knowledge or training. These notions of an individual and of a property must be minimal. In this way we quickly arrive at a whole web of minimal notions. This sketched argument casts some light on why the sweeping minimalist attitude is so pervasive (or so invasive): we cannot possess and deploy one minimal notion if we do not possess and deploy many other minimal notions. An interesting question is why our natural language contains thin semantical and ontological notions of the type just illustrated. An explanation is that these deflated concepts help to formulate explicitly blind generalizations. The arguments made by Horwich substantiate this claim in relation to the folk notions of truth and reference, and similar arguments apply to other semantical and ontological notions implicit in ordinary language (I give an example of it in the next section). Minimal notions resulting from regimenting these folk notions retain this important logical role—they satisfy the function of helping to formulate blind generalizations in regimented language. This is certainly one reason why we want them also in philosophical language. Wright believes that the minimal notion of truth can be made thicker by supplementing the truth-platitudes with substantive philosophical principles about truth. This thesis can plausibly be generalized to the other minimal notions—for example, the one of property and of an individual—that contribute to define minimal truth contextually and that, as I have conjectured, are part of the framework of creeping minimalism. These notions are characterized in terms of platitudinous principles only, which say nothing about whether, for instance, properties coincide with universals or with sets of tropes, or whether, for example, individuals (conceived of as concrete particulars) are just bundles of properties or are also constituted by independent substrata. For none of these alternative views is entailed by the platitudes definitive of the minimal notions of a property and of an individual. Minimal notions
142
Luca Moretti
do not seem to have the function of providing any interesting philosophical explanation. Yet the sweeping minimalist as such does not seem to be committed to maintaining that all of our semantical and ontological notions are just minimal. She could in principle concede that in certain areas of discourse we also possess, or we can define, thicker versions of these notions that rely on substantive philosophical assumptions that may allow for genuine philosophical explanations.40 Let us consider now the minimal notion of truth-aptitude, which plays a central role in both creeping minimalism and Wright’s minimalism. For Wright, the truth-platitudes hold valid of any sentence S that counts as at least minimally truth-apt. The problem of how to characterize truthaptitude is not central in Horwich’s investigation,41 but this issue is crucial in Wright (1992). Wright introduces his minimal notion of truth-aptitude in opposition to various forms of irrealism—such as expressivism and quasi-realism42 in metaethics, and instrumentalism in philosophy of science—which rely on more substantive notions of truth-aptitude (1992, 35–36). Expressivists and quasi-realists claim, for example, that the sentence “Buster Keaton was blown away by a tornado” is truth-apt, and the sentence “Buster Keaton was a pious man” is not truth-apt, though it may
40 Against forms of conceptual pluralism of this type, some philosopher might insist that the function of semantical and ontological notions is only that of helping to formulate blind generalizations and that these notions cannot provide any deep philosophical explanation. In this view, the alleged need for philosophical explanations would just arise from misunderstanding the essentially superficial character of semantical and ontological notions in general. (I am grateful to Amie Thomasson for making this point in discussion). Yet it is not evident to me that this conception of semantical and ontological notions is correct or even plausible—the burden of proof is on those who believe so. Furthermore, some sort of conceptual pluralism would seem to be already accepted in everyday discourse. For example, we generally have the tendency to distinguish between physical (or scientific) facts from facts of a different type—for instance, moral or aesthetic. And we generally assume that facts of the fi rst type are more objective or “real” than facts of the second type. It is unclear why philosophers should reject or revise these pluralist intuitions instead of trying to develop and clarify them. 41 More precisely, this problem is not central in the philosophical agenda of Truth. Stoljar (1993) contends that Horwich, in Truth, accepts however a specific form of minimalism about truth-aptitude called syntacticism, according to which any sentence with the superficial syntactical features of a declarative sentence is truth-apt. This contention is questioned by Jackson et al. (1994, 191 n. 6). Indeed, Horwich (1993, 1994) explicitly focuses on the problem of truth-aptitude, but the role played by these two short papers in the development of creeping minimalism is less relevant than that of Truth. 42 I refer, specifically, to Blackburn’s (1984) version of quasi-realism.
On Creeping Minimalism and the Nature of Minimal Entities 143 appear to be so. They believe that moral sentences cannot be asserted, as their contents are not assertoric; consequently, moral sentences are not true or false. Expressivists and quasi-realists claim that when we utter “Buster Keaton was a pious man,” we typically do not aim at describing any feature of Buster Keaton, but we do something else. For instance, we express our reverence for Buster Keaton’s actions. Likewise, instrumentalists claim that certain theoretical sentences are not actually truth-apt and that scientists use them in nonassertoric ways to produce observational consequences. Wright contends that although no form of irrealism is satisfactory, it would be less problematic and philosophically more fruitful to maintain that if certain sentences appear to be truth-apt (in a qualified sense of appearing), such sentences are in fact truth-apt at least in the elementary sense of being predicable of the minimal truth predicate (1992, 10–11, 27–29 and 74–75).43 For Wright, a sentence “S” is truth-apt if “S” satisfies surface constraints of syntax and discipline that characterize assertoric discourse (1992, 27–28). The constraints of syntax only require “S” to be declarative in form (hence, if a sentence “S” is truth-apt, for instance, “S” can occur as the antecedent of a meaningful conditional, “S” possesses a significant negation, “S” can occur in meaningful sentences expressing propositional attitudes—e.g., “I believe that S”—and so on). The constraints of discipline only require S to be subjected to communally acknowledged standards of proper use (1992, 29).44 This conception of truth-aptitude is often called disciplined syntacticism (Jackson et al. 1994). Many moral and scientific sentences will satisfy Wright’s light constraints of truth-aptitude. Disciplined syntacticism has been the target of several objections but none of them has yet proven compelling or sufficiently persuasive.45
43 Wright presses especially the so-called embedding problem. This problem was originally raised in Geach (1972) and Searle (1969, 136–141). 44 Notice that the same constraints of minimal truth-aptitude put forward by Wright (1992) were independently introduced by Boghossian (1990). 45 Here are some of these criticisms. M. Smith (1994b) and Jackson et al. (1994) have raised important objections to the adequacy of disciplined syntacticism in metaethics. Smith (1994b) argues as follows: if “S” is any moral sentence, the sincere utterance of “S” expresses a moral commitment that indicates how the utterer is motivated to act. If “S” is (minimally) truth-apt, the utterance of “S” will normally correspond to the assertion of “S.” It also appears platitudinous that
[B] If one asserts that S and one is sincere, one believes that S. The problem is that no belief in itself can be a motivational state (Hume’s principle). Thus, if “S” is (minimally) truth-apt, “S” cannot express any moral commitment and cannot be a moral sentence. Divers and Miller (1994) respond that Hume’s principle is controversial and thus not platitudinous and that, if Hume’s principle is true, it applies to a thick notion of belief, which is plausibly not the one
144 Luca Moretti Wright’s minimal notion of truth-aptitude can be said to be conceptually minimal—just as his minimal notion of truth can—in a relative sense of this expression. It would seem to be the most elementary notion of truthaptitude we can accept, in the sense of placing the most liberal possible constraints on what it is to count as truth-apt (see Burgess 1997, 260).46 Wright has apparently made no explicit argument to link the minimal notion of truth with the minimal notion of truth-aptitude. The minimal conception of truth-aptitude gives rise to the minimal conception of truth because any minimally truth-apt sentence, if true at all, is true at least in a minimal sense (M. Smith 1994b, 2). The reversed claim—that minimalism about truth involves minimalism about truth-aptitude—has however been questioned. For example, Jackson et al. (1994, 289) urge that the mere assumption that the truth predicate is implicitly defined by the appropriate instances of: [ES] It is true that S if and only if S does not involve by itself any commitment to claiming that, for instance:
the minimalist is committed to just on the grounds of [B]. Wright (1998) seems to reject [B] as a platitudinous principle. Jackson et al. (1994) put forward an argument somewhat similar to Smith’s (1994b). Any adequate analysis of the notion of truth-aptitude has to sustain relevant platitudes about the notion of belief. These platitudes include [B] and further principles such as: beliefs are states designed to fit the facts and beliefs tend to evolve in a rational manner under the impact of information. No minimal notion of truth-aptitude can sustain principles like the latter two. Therefore, no such notion is adequate. In response, Divers and Miller (1995b) contend that a minimal notion of truth-aptitude can sustain principles like these if properly interpreted in a platitudinous manner, and that, on the other hand, nonplatitudinous principles about belief cannot constrain the adequacy of a minimal notion of truth-aptitude. (This reply counts also as a response to M. Smith [1994a], written against Divers and Miller [1994]). Blackburn (1998) and, independently, Timmons (1999) recognize that moral discourse is superficially truth-apt but argue that it is not representational. If this were true, there would be a sense in which minimalism about truth-aptitude is not incompatible with the forms of moral expressivism asserting that moral discourse is minimally truth-apt but not representational. Wright (1998) and, independently, Dreier (2004) respond that though criticism of this kind seems to presuppose a “robust” notion of representation, it is very hard to question that superficially truth-apt discourse is also superficially representational. 46 Jackson et al. (1994, 291–293) have made a convincing case that the position called syntacticism—which appears even weaker than disciplined syntacticism—is incoherent. According to syntacticism, any sentence with the superficial syntactical features of a declarative sentence is truth-apt independently of whether it satisfies any constraints of discipline.
On Creeping Minimalism and the Nature of Minimal Entities 145 It is true that torture is morally wrong if and only if torture is morally wrong. For the initial assumption says nothing about whether the replacement of “S,” in [ES], with “torture is morally wrong” will yield an appropriate instance. Rather, if “torture is morally wrong” is not truth-apt, the replacement will be inappropriate. The above argument certainly works for some minimal notions of truth, but—I believe—it probably does not work when applied to a minimal notion of truth that is intended to be a simple regimentation of the folk notion of truth.47 For our folk notion of truth would seem to apply just to sentences that satisfy elementary constraints of syntax and discipline of the same kind as those proper to disciplined syntacticism (i.e., superficially declarative sentences subject to communally acknowledged standards of acceptance and rejection). Consequently, the platitudes that contextually defi ne minimal truth and the related minimal notions should apply to sentences lightly truth-apt in this very same sense. If this is correct, the minimal notion of truth—as long as it is meant to be a regimentation of the folk notion of truth—goes along with the minimal notion of truthaptitude that qualifies disciplined syntacticism. The sweeping minimalist has to possess this minimal notion of truth-aptitude, for this is the one that individuates the sentences suitable to the schematic platitudes that defi ne many minimal notions. As our folk notion of truth is to some extent fuzzy, incoherent, and easily yields logical paradoxes, we can obtain a viable model of it only if we revise and amend our natural linguistic practices to some extent. It might be argued that a consequence of this is that the sentences to which this minimal notion will end up applying will satisfy constraints of truth-aptitude less liberal than those proposed by Wright: for instance, constraints not fulfi lled by moral sentences in general. This might be true, but it is not obviously so. The burden of proof is on those who make this claim. Clearly, if we aim at defining a notion of truth that satisfies many substantive philosophical assumptions, we will end up with a demanding notion of truth-aptitude unsuitable to many kinds of sentences. This would however be no elementary regimentation of our folk notion of truth. Wright’s minimalism has set the stage for the rise of creeping minimalism by introducing a deflationary notion of truth defi ned contextually in terms of platitudes that link this concept with other deflationary notions. In doing this, Wright has broadened the family of thin notions available to the sweeping minimalist. Wright’s minimal notion of truth seems to amount to nothing but an elementary refi nement of the folk notion of truth.
47 As the predicate “is true” has different ordinary meanings (see n. 33), I specifically refer here to the folk notion of truth that links truth with assertion.
146
Luca Moretti
I have conjectured that many or most sweeping minimalists conceive of the minimal notion of truth in these exact terms. I have suggested that accepting a minimal notion of truth consisting in an elementary regimentation of the folk notion of truth involves accepting many (if not all) of the additional minimal notions that play a key role in creeping minimalism, including the minimal notion of truth-aptitude.
2.3
Schiffer’s Pleonastic Ontology
The sweeping minimalist attitude not only involves the possession of and the propensity to use a whole battery of minimal notions, it also engenders commitment to the existence of the entities that fall under these minimal notions—namely, commitment to the existence of minimal entities. For instance, a sweeping minimalist disposed to assert that Ribot is an energetic penguin will be prepared to claim not only that it is a fact that Ribot is an energetic penguin and that Ribot has the property of being an energetic penguin but also that this fact and this property actually exist. Stephen Schiffer’s work on pleonastic entities may have suggested the basic conceptual devices to justify this commitment. Schiffer’s most influential paper on pleonastic entities is probably “Language-Created Language-Independent Entities” (1996).48 The conception of pleonastic entities has been clarified and extensively developed in relation to propositions in Schiffer (2003). Schiffer argues that propositions, properties, states, and events are pleonastic entities—that is, entities whose existence is typically secured by something-from-nothing transformations. The latter are a priori conceptually valid inferences that take us from a true sentence in which no reference is made to a thing of a certain kind to a sentence in which there is a reference to a thing of that kind (1996, 149, 2003, 2, 51).49 Pleonastic entities can also be characterized by saying that their existence typically supervenes on the
48 The view proposed in Schiffer (1996) was already outlined in Schiffer (1994). Schiffer (1996, 167 n. 9) recognizes that his conception of pleonastic ontology was influenced by the view about meaning put forward in Johnston (1988). Forms of deflationary realism similar to Schiffer’s, in various philosophical fields, have been put forward by Azzouni (1991), Barber (1998), Resnick (1997), and Stalnaker (1996). 49 Valid interpretations and criticisms of Schiffer’s conception of pleonastic entities are given in Sainsbury (2005) and, especially, in Thomasson (2001). Reading the latter paper has been crucial for many of the claims I make in this section and in §4 of this chapter. Two important papers for the analysis of the somethingfrom-nothing feature and issues of pleonastic ontology are Hofweber (2005, 2007). Hofweber’s interesting and sophisticated theses are often at odds with mine and would deserve thorough analysis and discussion in a separate paper.
On Creeping Minimalism and the Nature of Minimal Entities 147 premises of something-from-nothing inferences (1996, 159, 2003, 61). 50 I will give examples of these inferences below. Pleonastic entities have no hidden and substantial nature waiting to be uncovered. The essential truths about them are fully determined by the linguistic practices that are constitutive of the concepts of them together with other necessary a priori truths applicable to things of any kind51 (1996, 160–164, 2003, 61–63). It is not clear to me whether Schiffer believes that the “essential truths” about pleonastic entities are just platitudinous principles, 52 and—if he believes so, in what sense of “platitudinous”—for example, in the sense of being part of an allegedly shared metaphysical background or in the sense of constituting a mere regimentation of the folk theory of those entities. Despite this, Schiffer’s central claims straightforwardly apply to minimal entities—precisely, to entities defi ned in terms of platitudinous notions conceived of as regimentations of correlate folk notions. Indeed, it is nearly irresistible to interpret Schiffer’s arguments as attempting to substantiate the claim that we are committed to the existence of minimal entities of the latter type. The something-from-nothing inferences come, in general, in two consequent steps that I will call nominalization (as we infer a sentence including a noun phrase) and quantifi cation (as we infer an existential generalization). 53 The following is a typical something-from-nothing inference that secures the existence of properties. Consider a truth-apt sentence with the form:
50 For a very precise defi nition of a pleonastic entity, see Schiffer (2003, 56–57, 62–63). Two important properties of pleonastic entities are that if we add them to a theory, we conservatively extend that theory, and that they do not disturb the preexisting causal order. 51 Such as if x is identical to y, then whatever property x has, y has, and vice versa. 52 A reason for doubting it is, for instance, that Schiffer claims that a property, “assuming that there be such a thing, is abstract, or immaterial, in that it doesn’t occupy space or have any other physical properties” (1996, 150). These characteristics of a property appear to me not to be platitudinous in any sense. 53 These two expressions are borrowed from Hofweber (2005). I will present here a cut down version of Schiffer’s conception of pleonastic propositions and properties. Schiffer (1996, 159–160) delivers two condensed arguments to the effect that propositions and properties exist necessarily. Only the argument about properties, not the one about propositions, is restated in Schiffer (2003). For limitations of space, I cannot examine these arguments here. Iacona (2003, 340–343) has argued—persuasively, in my opinion—that the argument for the necessary existence of propositions is invalid or unsound. In Moretti (unpublished), I contend that Iacona’s considerations can be applied to undermine Schiffer’s argument for the necessary existence of properties too.
148
Luca Moretti [1] a is P,
(Nominalization). After nominalizing the predicate “P” with “the property of being P,” we infer [1]’s pleonastic equivalent: [2] a has the property of being P. (Schiffer 1996, 149 and 159)54 (Quantification). From [2], we then infer: [3] There is something possessed by a; that is, the property of being P. (1996, 151) The following is a typical something-from-nothing inference that secures the existence of propositions. Consider a truth-apt sentence: [4] S. (Nominalization). After nominalizing “S” with “the proposition that S,” we infer [4]’s pleonastic equivalent: [5] The proposition that S is true. (1996, 149–150, 159)55 (Quantification). From [5], we then infer: [6] There is something true; that is, the proposition that S. (1996, 160) What are the grounds of the inferential step of nominalization? Schiffer emphasizes that it is hard to override the natural conviction that all correlated instances of [1] and [2] share the same truth-value as a matter of necessity and that also all correlated instances of [4] and [5] share the same
54 According to Schiffer, the existence of pleonastic entities cannot always be secured by something-from-nothing inferences (2003, 51). For there are sentences including a singular term whose referent is a property that is not entailed by sentences not including singular terms whose referents are properties. For example, the sentence “humility is a virtue” (or “the property of being humble is a virtue”) cannot be entailed by any sentence that contains no term referring to humility (2003, 70). 55 I am simplifying. To be more accurate, Schiffer obtains [5] from [4] via fi rst inferring “that S is true” (or “it is true that S”) from [4] through the equivalence schema (1996, 160, 2003, 71–72), and then by arguing that the that-clauses refer univocally to propositions if they refer at all (1996, 150–151, 2003, 12–15, 92–95). I fi nd the latter thesis questionable; for instance, there are contexts in which thatclauses appear prima facie to refer to facts and not to propositions (see Moffett 2003). I cannot go deeper into this delicate issue here.
On Creeping Minimalism and the Nature of Minimal Entities 149 truth-value as a matter of necessity (1996, 152). For it seems to be a conceptual truth that a is P if and only if a has the property of being P and that S if and only if the proposition that S is true (2003, 61). In particular, the mere possession of the notion of a property enables us to infer [2] from [1] and vice versa, and the mere possession of the notion of a proposition enables us to infer [4] from [5] and vice versa. 56 For our concern here, notice that the equivalence between all correlated instances of [1] and [2] holds valid for any minimal notion of a property characterized at least in terms of the platitude: [j] a has the property of being P if and only if a is P. On the other hand, the equivalence between all correlated instances of [4] and [5] holds valid for any minimal notion of a proposition characterized at least in terms of the Equivalence Schema plus the platitude: [a] The proposition that S is true if and only if it is true that S. Consider now the inferential step of quantification—that is, the inference from [2] to [3] and the inference from [5] to [6]. Schiffer emphasizes that the expression “the property of being P” (when P is replaced with a predicate) has in many contexts the syntactical function of a singular term (1996, 150, 2003, 61). Schiffer also argues that the expression “the proposition that S” (when “S” is replaced with a truth-apt sentence) has in many contexts the syntactical function of a singular term (1996, 150– 151, 2003, 12–15, 92–95). 57 In particular, the expression “the proposition that S” (when “S” is replaced with a truth-apt sentence) does play the syntactical role of a singular term in [5] (i.e., “the proposition that S is true”). On the other hand, [2] (i.e., “a has the property of being P”) admits of innocent reformulations in which the expression “the property of being P” (when “P” is replaced with a predicate) does play the syntactical role of a singular term (2003, 63). For instance, this happens in the equivalent reformulation of [2], [2*] The property of being P is instantiated by a,
56 Schiffer (1996, 52, 2003, 89) stresses that it is the conviction that the step of nominalization rests on conceptual truths that prima facie prevents us from endorsing an error-theory of pleonastic entities. If instances of [1] and [4] are true and the respectively correlated instances of [2] and [5] conceptually follow from them, there is no room for errors: these instances of [2] and [5] must be true too. 57 More exactly, Schiffer fi rst argues that the expression “that S” functions as a singular term and then that “that S” refers to the proposition that S, if it refers at all. See also n. 55.
150
Luca Moretti
deduced from [2] via the following platitude: [k] The property of being P is instantiated by a if and only if a has the property of being P. Schiffer stresses that we normally accept that “[existential] quantifiers can bind variables that occupy the positions of the singular terms in question” (1996, 151), which implies that we normally accept that [2*]— and so [2]—entails: [3] There exists an x such that x is instantiated by a, and x is identical to the property of being P. (Colloquially: there is something possessed by a; that is, the property of being P), and that [5] entails: [6] There exists an x such that x is true, and x is identical to the proposition that S. (Colloquially: here is something true; that is, the proposition that S). 58 The general intuitive principle implicitly appealed to by Schiffer to ground both these inferences seems to be the following: something has a given feature if and only if there is something that has that feature. The following schema expresses the same principle in a slightly more regimented way: [E] b is Q if and only if there exists an x such that x is Q, and x is identical to b.59
58 Schiffer (1996, 51–52), and in much more detail Schiffer (2003, 74–87), also argues that a given proposition exists if a sentence expressing a propositional attitude towards that proposition is true. For example, the proposition that S exists if the sentence “I believe that S” is true. Schiffer contends that what is special about sentences expressing propositional attitudes is that we can determine their truth independently of identifying the referent of their that-clause. That their thatclauses do refer follows from the truth of these sentences. Schiffer (2003, 87–89) also argues that the that-clause included in a sentence obtained via a somethingfrom-nothing inference (e.g., “that S is true”) and the that-clause included in a corresponding sentence expressing a propositional attitude (e.g., “I believe that S”) do refer to the same proposition (i.e., “that S”). As I aim to individuate a principle of ontological commitment for minimal entities in general, and not just for propositions, I do not analyze these arguments by Schiffer here. 59 Notice that [E] is conceptually analogous to the syntax priority thesis introduced by Wright (1983) as the result of applying Frege’s context principle to reference (or “bedeutung”). The syntax priority thesis is employed by Wright and other
On Creeping Minimalism and the Nature of Minimal Entities 151 (In which “b is Q” stands for a truth-apt sentence, “b” stands for an expression that has the syntactical function of a singular term, and “Q” stands for an expression that has the syntactical function of a predicate). We can use [E] to infer instances of its left-hand side from instances of its righthand side: we fi rst establish that there exists an x that is Q and is identical to b, and from this we infer that b is Q. Yet, in the case of pleonastic (and minimal) entities, we proceed the other way around: we fi rst infer that b is Q on purely conceptual grounds through the step of nominalization (or through a relevant platitude). From it, we then infer that there exists an x that is Q and is identical to b. There might perhaps be reasons to question the validity of instances of [E].60 Notice however that colloquial versions of the instances of [E] that
contemporary neo-Fregeans to defend arithmetical Platonism. Wright formulates the syntax priority thesis as follows: Once it has been settled that a class of expressions function as singular terms by syntactic criteria, there can be no further question about whether they succeed in objectual reference which can be raised by someone who is prepared to allow that appropriate contexts in which they figure are true. (Wright 1992, 28) It is unclear to me whether Wright considers the syntax priority thesis to be a component of his minimalism about truth. At any rate, I believe that the syntax priority thesis and the discussions of recent neo-Fregeans may also have played a significant role in the development of creeping minimalism. 60 A worry might be that [E] commits the sweeping minimalist to asserting that fictional entities do exist. Precisely, the problem would be that many sentences having the form “b is Q,” where “b” stands for the name of a fictional entity, are minimally truth-apt because they fulfi l minimal constraints of syntax and discipline; consequently, these sentences can be replaced in [E]. The absurd consequence would then follow. Divers and Miller (1995a) have exposed a very similar concern about Wright’s syntax priority thesis (for the content of this thesis, see n. 58). I believe that a slightly modified version of Wright’s (1994) response to Divers and Miller might help to dispel the concern about [E]. Wright (1994, 1998) emphasizes that his minimalist view is compatible with the fact that a truth-apt sentence can be subject to different standards of acceptability than the ones that aim at warranting its truth. For example, truth-apt fictional sentences are also subject, in all probability, to standards of acceptability as conformity to the relevant fiction. A sweeping minimalist who accepts [E] can make the same claim as Wright. Consider now the truth-apt sentence “Godzilla is huge.” Although from “Godzilla is huge” and [E] we can derive logically “there is something huge; that is, Godzilla,” we are not justified in asserting the latter existential sentence because we are not justified in asserting “Godzilla is huge.” We would be justified in asserting “Godzilla is huge” if, for instance, we actually saw Godzilla out there, and we observed that this monster is huge. (Evidence of this type would plausibly satisfy the commonly
152
Luca Moretti
quantify over the entities that fall under the folk notion of a proposition and under the folk notion of a property (and, plausibly, under the folk notions of a fact and of an individual)61 are typically accepted as valid in ordinary language. (For example, we can ordinarily infer the sentence “there is something possessed by fi re; that is, the property of burning” from the sentence “the property of burning is possessed by fi re” and vice versa. Also, we can ordinarily infer the sentence “there is something false; that is, the proposition asserted by John” from the sentence “the proposition asserted by John is false,” and vice versa). Thus, the minimalist who aims to provide a regimentation of folk semantical and ontological notions should probably accept as valid at least the correlated regimented instances of [E]; and so will the sweeping minimalist. If these instances of [E] are valid, the step of quantification is justified on the whole. There is, indeed, a quite general reason why existential quantification over minimal entities should be allowed. As Hofweber (2006) has emphasized, quantification over these entities will increase the expressive power of our language by enabling the explicit formulation of generalizations in situations of partial information (2006, 157–158). To illustrate this point, let us focus on minimal properties. Suppose John knows nearly nothing about marmite and vegemite. Suppose, however, that he believes, for some reason, that marmite and vegemite share some property, and that he expresses his belief by asserting that marmite and vegemite share some property. The quantifier implicit in John’s assertion can be made explicit by formulating that assertion for instance as follows:
acknowledged conditions of acceptance that make the declarative sentence “Godzilla is huge” truth-apt). Yet, given that Godzilla does not exist, this cannot happen. On the other hand, the ordinary assumption that we are in some sense justified in saying “Godzilla is huge” should properly be intended in the sense that we are justified in uttering “Godzilla is huge” because it conforms to the content of the relevant fiction. The same applies to other fictional sentences. At any rate, if [E] is accepted, some restrictions on the expressions that can be replaced for “Q” should however be imposed. For example, if “b” is, again, “Godzilla” and “Q” is “a nonexistent entity,” “b is Q” is clearly true. From this, we can derive, through [E], the oddity that there exists a nonexistent entity, which sounds weird even in ordinary language. This is, of course, just one instance of the very complex problem of the negative existentials that I cannot analyze in this chapter. 61 For example, an informal instance of [E] that quantifies over the entities that fall under the folk notion of a fact may be that today is Saturday is a fact if and only if there is a fact: namely, that today is Saturday. An informal instance of [E] that quantifies over the entities that fall under the folk notion of an individual (or object) may be: the individual John is tall if and only if there is someone tall; that is, the individual John.
On Creeping Minimalism and the Nature of Minimal Entities 153 There exists an x such that x is a property of marmite, and x is a property of vegemite. Another way to formulate John’s assertion is via the following sentence: There exists an x such that x is instantiated by marmite, and x is instantiated by vegemite. It might not be obvious that the second generalization also quantifies over minimal properties. That it does so becomes apparent as we consider that it is platitudinous that, for every x, x is instantiated by a, if and only if x is a property of a. There are many equivalent ways to express the same existential generalization over minimal properties, where it is not essential that the term “property” explicitly occur in all of them (this is true mutans mutandis for other minimal entities). If existential quantification over minimal entities were not permitted, John could express his belief only with a turn of phrase that alludes to a potentially infi nite disjunction of sentences, for instance, this: Marmite is baldness-curative, and vegemite is baldness-curative, or marmite is shyness-curative, and vegemite is shyness-curative, or . . . and so on.62 Similar examples that apply to minimal propositions and to other minimal entities can easily be found.63 In conclusion, the accomplishment of an important function of minimal notions—that is, enabling the explicit formulation of blind generalizations—engenders commitment to the existence of minimal entities. As Schiffer (1996, 151–152, 2003, 89–90) observes, though the friend of pleonastic entities can probably vindicate the inference of [3] from [2] and the inference of [6] from [5], it might still be argued that the truth of instances of [3] and [6] ensure no defi nite commitment to the existence of properties and propositions, despite the fact that we can say—by asserting these sentences—that properties and propositions exist. Precisely, the argument would be to the effect that it is not necessary to assume that the expressions in these true instances of [3] and [6] that play the syntactical role of a singular term—namely, “the property of being P” and “the proposition that S” (where “P” is replaced with a predicate and “S”
62 That this potentially infi nite disjunction expresses the same content as the above generalizations can be shown by applying [j] and [k] to its disjuncts. 63 Consider for instance explicit existential generalizations like “there are facts about Hegel’s life we still do not know,” and “in China, there are two identical individuals.”
154 Luca Moretti with a truth-apt sentence)—actually refer. This is so because—it might be argued—the existential quantifiers in [3] and [6] can be thought of as nonobjectual quantifiers64 (for instance, substitutional quantifiers or internal quantifiers).65 Nonobjectual quantification would allow for the truth of instances of [3] and [6] without requiring the existence of entities in the domain of quantification that satisfy the open sentences “x is instantiated by a, and x is identical to the property of being P” and “x is true, and x is the proposition that S” (when “a” is replaced with a singular term, “P” is replaced with a predicate, and “S” with a truth-apt sentence). Schiffer gives a quite convoluted response to this potential objection that I am not sure I understand in full (2003, 90–91). I have, anyway, a simpler response that becomes straightforward when the objection aims to undermine reference to—specifically—minimal entities (i.e., entities defi ned in terms of regimentations of correlated ordinary notions). Consider that any competent speaker of ordinary language is certainly entitled to defl ationary semantical notions and that, consequently, the sweeping minimalist is also entitled to them (i.e., to their regimented versions). On these deflationary notions, it is hard to deny that the singular terms in the instances of [3] and [6] do refer and that the quantification in the instances of [3] and [6] is objectual. Consequently, the singular terms in these instances do refer (though in a deflationary sense), and the quantification in the same instances is objectual (though in a deflationary sense). Let us go into more detail. Consider again the platitudinous principle about reference introduced in §2.1: [R*] For every x, “b” refers to x if and only if b is identical to x. The conjunction of [3] and the relevant instance of [R*] entails: There exists an x such that x is instantiated by a, and “the property of being P” refers to x. On the other hand, the conjunction of [6] and the relevant instance of [R*] entails: There exists an x such that x is true and “the proposition that S” refers to x. This shows that the sweeping minimalist as such possesses the semantical tools to refer to minimal properties and minimal propositions.
64 Clearly, to license the inferences of [3] from [2*] and of [6] from [5], the quantifier in [E] should also be interpreted as nonobjectual. 65 For the notion of internal quantification, see Hofweber (2005).
On Creeping Minimalism and the Nature of Minimal Entities 155 A platitudinous principle characterizing the notion of satisfaction may be the following: [S] For every x, x satisfies “x is Q, and x is identical to b” if and only if x is Q, and x is identical to b. The conjunction of [3] and the relevant instance of [S] entails: There exists an x such that x satisfies “x is instantiated by a, and x is identical to the property of being P.” The conjunction of [6] and the relevant instance of [S] entails: There exists an x such that x satisfies “x is true, and x is identical to the proposition that S.” This shows that the sweeping minimalist as such possesses the semantical tools that permit objectual quantification in generalizations over minimal properties and minimal propositions. Principles [R*] and [S] can plausibly be employed to ensure reference to and objectual quantification over other minimal entities, such as individuals and facts. The claim that mere deflationary semantical principles enable the sweeping minimalist to refer to minimal entities and to apply objectual quantification to them should not appear surprising or odd. For minimal entities are themselves deflationary entities—they are mere hypostatizations of our linguistic practices. Rather, it would make little sense to try to apply more substantive semantical principles to cheap entities of this kind—for instance, a causal notion of reference or one that postulates a sui generis reference relation.66 Minimal entities are essentially such that only deflationary semantical notions can properly apply to them. As I show in §4, this claim entails a decisive consequence for the issue of the ontological status of minimal entities. It is worth emphasizing that Schiffer is a monist—and not a pluralist— about properties and propositions. What Schiffer (1996 and 2003) aims at is providing a truthful account of properties and propositions tout court— in his account, properties and propositions turn out to be just pleonastic entities. Apparently, for Schiffer, there cannot be properties and propositions other than these (and this plausibly holds for states and events too). I do not see, however, why philosophers affected by sweeping minimalism
66 I refer to the implausible conjecture—for instance considered and dropped by Putnam (1981)—that our minds have special semantical powers, inexplicable in naturalistic terms (or in both naturalistic and normative terms), which allow us to refer to external things.
156 Luca Moretti should necessarily subscribe to Schiffer’s ontological monism. Minimal entities are “shallow” entities in the sense of being characterized in terms of platitudinous principles only, which do not allow for genuine philosophical explanations. Yet it is not self-evident that all semantical and ontological entities are just minimal (the onus of proof would seem to be on those who make this claim). Philosophers who believe that philosophical explanations are possible can entertain the sweeping minimalist attitude. For the sweeping minimalist is not committed to rejecting the thesis that there exist also thick entities, characterized in terms of substantive principles.67 The sweeping minimalist can prima facie be an ontological pluralist; she could accept that the world is populated by both thin entities, which are useful in virtue of their logical function, and—at least in some ontological domains—thick entities, which also function to give philosophical explanations. Consider for example Armstrong’s (1983) account of laws of nature grounded in the assumption that only substantial properties—or universals—possess the (nonplatitudinous) feature of instantiating relations of nonlogical or contingent necessitation. A philosopher affected by creeping minimalism could prima facie accept this account. She could maintain that, for any given predicate, the corresponding property is either unsubstantial (i.e., it coincides with a minimal property) or substantial (i.e., it coincides with a universal in Armstrong’s sense). The contribution by Schiffer to the creation of the creeping minimalist framework comes in terms of the explanation of the typical way in which the minimalist can become committed to the existence of minimal entities: that is, via using something-from-nothing inferences. I have argued that these inferences can be thought of as grounded in central platitudes for the introduction of minimal notions, principles that govern the logical role of the existential quantifier, and deflationary semantical principles. I have suggested that all these principles are probably included in the conceptual framework of sweeping minimalism, for they all can be interpreted to be logical regimentations of correlated assumptions implicit in our natural languages. Résumé Let us recap the fi ndings of the last three sections. I have described the sweeping minimalist attitude as essentially one that prompts a massive use of deflationary notions in semantics and ontology, where these notions are defi ned through the regimentation of corresponding folk notions implicit in
67 In the next section, I argue that the sweeping minimalist can be an ontological realist, in the sense that she can coherently believe that given entities exist and are mind-independent. From this it seems to follow that a sweeping minimalist who embraced a form of ontological pluralism could plausibly accept that thick entities such as facts, propositions, and properties are mind-independent.
On Creeping Minimalism and the Nature of Minimal Entities 157 natural language. Minimal notions of this sort are metaphysically unloaded and platitudinous. This interpretation of creeping minimalism is somewhat corroborated by the fact that it explains (or at least it casts some light on) two perceptible features of this philosophical tendency: its strong appeal and its pervasiveness. The sweeping minimalist attitude is captivating, in this interpretation, because it consists essentially in deploying in philosophy familiar notions we already use in ordinary discourse and that we can introduce to technical language with only a negligible theoretical effort. Sweeping minimalism is pervasive because minimal notions are holistically related to one another as the folk notions that they replicate are. I have suggested that the framework of creeping minimalism can be constructed as a natural extension of the framework of Wright’s minimal truth. I have also suggested that Horwich’s explanation of the logical function of deflationary truth and deflationary reference as devices to enable blind generalizations can be generalized to clarify the logical role of many other minimal notions. I have argued that deflationary principles included in the framework of the sweeping minimalist ensure ontological commitment to minimal entities; principles of this kind have been presupposed or explicitly introduced in papers by Schiffer and Horwich. I have emphasized that the very accomplishment of the function of minimal notions of enabling blind generalizations engenders commitment to the existence of minimal entities. Finally, I have suggested that Wright’s pluralism about truth can be extended to other minimal concepts deployed by the sweeping minimalist and to the correlated minimal entities.
3 REALISM AND ITS ANTAGONISTS IN THE AGE OF CREEPING MINIMALISM In this section, I analyze the consequences of the rise of creeping minimalism for the general configuration of the debate between realists and nonrealists in metaphysics. This enquiry is not only interesting in itself but it also paves the way for the discussion of the next section, in which I investigate the nature of minimal entities and what follows from the assumption that these entities exist for the dispute about realism and nonrealism. The reader might worry that the upshot of the next section may prove relevant for what is discussed in this section, as entertaining the sweeping minimalist attitude involves allowing the existence of minimal entities, and this might in turn have implications for the general configuration of the debate on realism and nonrealism. (Indeed, I will argue in the next section that there are implications in this sense when nonrealism is construed as a form of error-theory). The proper way to understand the enquiry of the present section therefore is this: I will investigate the consequences of creeping minimalism for the general configuration of the debate between realists
158 Luca Moretti and nonrealists while neglecting—for the moment—the possible implications that result from allowing the existence of minimal entities. On this understanding, the investigation of the next section constitutes the completion—or at least, the intensification—of the one of this section. As we have seen in §2.2, Wright thinks of realism as a doctrine essentially concerning the nature of truth. This way of characterizing realism is however highly controversial.68 Devitt (1997) and Horwich (1998a, 53–60)69 —among many others—have argued that realism is essentially an ontological doctrine, which is quite independent of any specific notion of truth. Since I find the ontological characterization of realism more natural and immediate than the alethic characterization, I stick to the former in the discussion of the next few pages. In accordance with the ontological characterization, one is a realist about entities of type X if and only if one believes [I] that X-entities do exist and [II] that X-entities (and at least some of their features) are mind-independent (Miller 2005).70 The general notion of mind-independence can be spelled out in various ways—for instance, as independence of anyone’s beliefs, theories, linguistic practices, and so on. (I give examples of more precise notions of mind-independence on pp. 165–166.) If the existence condition [I] or the independence condition [II] is not satisfied, nonrealism about X-entities obtains. In contemporary metaphysics, irrealists—such as expressivists, quasi-realists. and instrumentalists on the one hand and error-theorists on the other—typically reject the existence dimension, and antirealists typically accept the existence dimension but reject the independence dimension (Miller 2005).71 From an ontological point of view, expressivists, quasi-realists, and instrumentalists about X-entities—let us call them all noncognitivists about X-entities—are committed to rejecting the thesis that these entities exist. For, according to these philosophers, the sentences that appear to fulfil constraints of truth-aptitude and that appear to be about X-entities are not really truth-apt and so cannot be about X-entities. The consequence is that,
68 Notice that this does not undermine Wright’s input to the creation of the creeping minimalism framework, which appears largely independent of this controversial conception of realism. 69 On this issue, see also Horwich (1996). 70 This characterization of realism appears to rule out realism about mental states at the outset, as these entities are trivially mind-dependent, in the sense that my mental states cannot exist without my mind. A way out may be that of imposing that realism about mental states obtains only if these states do not prove minddependent in some more substantive sense of mind-dependence. I cannot go deeper into this very complex issue here. An interesting paper on this problem is Vinueza (2001). 71 Here, I do not follow closely Dreier’s (2004, 25) characterization of the different kinds of nonrealism, which I fi nd slightly confusing.
On Creeping Minimalism and the Nature of Minimal Entities 159 for them, claiming that there are X-entities makes no sense (or it cannot be interpreted literally). What happens when noncognitivists become sweeping minimalists? They will concede that the sentences that appear to fulfil constraints of truth-aptitude and that appear to be about X-entities are at least minimally truth-apt and are about X-entities in a deflationary sense of being about something. Once this step is made, they will no longer be entitled to maintain that the thesis that there exist X-entities is meaningless. The noncognitivists who fall prey to creeping minimalism appear committed to maintaining that X-entities do exist whenever they can warrantedly assert certain sentences about X-entities—for instance, sentences in subjectpredicate form in which the name of an X-entity features as the subject. For this follows from the acceptance of deflationary principles that ensure ontological commitment, such as the above [E], [R*], and [S]. Given that some of the sentences about X-entities (e.g., moral properties, numbers, etc.) that entail the existence of the latter via deflationary principles typically satisfy the standards of assertibility that inform the relevant discourses, these philosophers are often committed to maintaining that X-entities do exist.72 There are two distinct questions here: one is whether noncognitivists affected by creeping minimalism still remain noncognitivists, and the other is whether noncognitivists affected by creeping minimalism can still challenge realism. Concerning the fi rst question, O’Leary-Hawthorne and Price (1996) have argued that one can properly be said to be a noncognitivist about X-entities on the grounds of the specific functional role one attributes to the sentences about these entities. This would allow philosophers who acknowledge that sentences about X-entities are minimally truth-apt to qualify, at least in some cases, as noncognitivists.73 This consideration is certainly interesting but does not answer by itself the second question. Noncognitivists about X-entities affected by creeping minimalism might perhaps insist that the X-entities they are committed to accepting as existent
72 To be more accurate, sophisticated expressivists and sophisticated quasi-realists seem to believe that certain sentences can at the same time be minimally truth-apt but nondescriptive. When sophisticated expressivists and sophisticated quasi-realists become sweeping minimalists, they should however recognize that the same sentences are descriptive in some deflationary or minimal sense (see n. 45). 73 O’Leary-Hawthorne and Price focus on expressivism. They contend that what is distinctive about expressivism—independently of issues of truth-aptitude—is its functional pluralism. According to them, “most generally construed, non-cognitivism is a doctrine about the functions of parts of language. . . . [T]he non-cognitivist’s essential claim is that the function of ethical discourse is different from that of, say, scientific discourse, in some philosophically significant respect—in such a way, for example, as to make attempts to reduce ethical talk to scientific talk inappropriate” (1996, 285). Plausibly, as an expressivist goes sweeping minimalist, she can still be considered to be an expressivist (and so a noncognitivist) in this sense.
160
Luca Moretti
are just defl ationary and that, because of this, they are not fully real. It is however dubious that the noncognitivists as such do possess the conceptual apparatus to elucidate, beyond rhetoric, the meaning of the expressions “fully real” (or a similar expression) and “deflationary” to explain why deflationary entities are not fully real. For instance, a noncognitivist might try to play on the thesis that deflationary entities are not fully real because they are not mind-independent. The problem is that our noncognitivist should give arguments for the conclusion that deflationary entities are not mind-independent, and this would mean settling the debate between realists and nonrealists as a dispute between realists and antirealists rather than between realists and noncognitivists, as it is the antirealist who essentially questions realism via questioning the independence dimension. In that case, the challenge to realism would come from the antirealist side, and not from the noncognitivist side. Alternatively, a noncognitivist affected by creeping minimalism could perhaps attempt to rebuff her commitment to the existence of X-entities by substantiating the thesis that we all have, for some reason, a propensity to apply erroneously the justification standards of the sentences about X-entities, with the consequence that the sentences about X-entities that appear warrantedly assertible are in fact not so. The problem is that this strategy would amount to settling the debate between realists and nonrealists as a dispute between realists and error-theorists rather than between realists and noncognitivists, as it is the error-theorist who essentially questions realism (via questioning the existence dimension) by adducing the presence of generalized errors in our factual claims. In that case, the challenge to realism would come from the error-theorist side and not from the noncognitivist side. Indeed, Dreier (2004, drawing from Fine, 2001; O’Leary-Hawthorne and Price, 1996; and Gibbard, 2003) contends that expressivists and quasi-realists can still challenge realism even while entertaining the sweeping minimalist attitude. Briefly, expressivists and quasi-realists should argue that even though elucidating what a moral judgment is may involve describing it as representational and cognitive in character (e.g., judging that Hitler is evil consists in believing that the fact that Hitler is evil obtains), the deepest explanation of what a moral judgment is can only describe it as nonrepresentational and noncognitive in character. In a few words, this explanation would say that making a moral judgment consists of nothing more than being in “a [mental] state that plays a certain noncognitive psychological role, a role more like desire than it is like factual belief” (Dreier 2004, 39).74 A consequence one should draw from this explanation being the deepest available is—if I get Dreier right—that moral discourse is essentially noncognitive, with the effect
74 Dreier emphasizes that, according to this view, “expressivism doesn’t merely attribute a distinctive functional role to [moral] judgment but also makes the stronger claim that having the functional role constitutes being the [moral] judgment” (2004, 38).
On Creeping Minimalism and the Nature of Minimal Entities 161 that the claim that there are moral properties and moral facts should not be taken seriously. Although Dreier’s proposal is interesting and encouraging, a reason to be unhappy with it is that the notion of “consisting of nothing more than” is not perspicuous75 —perhaps it is even more obscure than the notions it is meant to illuminate. Suppose, for example, we want to analyze this notion in terms of possible worlds semantics. If making a moral judgment consists of nothing more than being in a (given) noncognitive mental state, then—plausibly—there should be a possible world in which one makes a moral judgment without being in any cognitive mental state. This would explicate, to some extent, why cognitive features are not constitutive of moral judgments. Unfortunately, this analysis is not available to the sweeping minimalist. The problem is that the philosopher entertaining the creeping minimalist attitude will very plausibly consider the (nonparadoxical) instances of the following schemata as a priori—and so necessarily—true: If one judges that S, one judges that it is true that S; If one judges that S, one judges that it is a fact that S. (where “S” stands for a minimally truth-apt sentence). The sweeping minimalist will thus conclude that, in any possible world, if one judges that Hitler is evil, one judges that it is true and that it is a fact that Hitler is evil. But this would seem to show that factual and cognitive features are constitutive of moral judgments. Unless some substantive clarification is given, I cannot see Dreier’s suggestion as satisfactory. In conclusion, it seems to me that when creeping minimalism comes into play, the challenge raised by noncognitivism to realism loses most of its original bite (at least, when realism is defi ned in ontological terms through the theses [I] and [II]). When creeping minimalism takes place, to construe the debate about realism and nonrealism as a dispute between realists and noncognitivists would probably be fruitless (or too problematic). For it becomes very hard to understand in what sense noncognitivism (or the position noncognitivism is turned into) still constitutes a challenge for realism. Of course, mixed positions—for example, O’Leary-Hawthorne and Price’s noncognitivism plus antirealism, or plus error-theory—could probably still challenge realism. But in that case, what would accomplish the essential task of denying the realist doctrine would be, not noncognitivism but, rather, antirealism or error-theory. In sum, a mixed position would be a form of nonrealism because of its antirealist or its error-theory component and not because of its noncognitivist component.
75
Dreier himself acknowledges this as a (nondevastating) problem (2004, 35).
162
Luca Moretti
Setting aside noncognitivism, realism can still be questioned either on the error-theory side (via denying the existence dimension) or on the antirealist side (via denying the independence dimension). Traditionally, error-theorists about X-entities have argued that truth-apt and warrantedly assertible sentences about X-entities are systematically false because no Xentity actually exists.76 For instance, Mackie (1977) questioned the existence of moral properties by casting doubt—principally—on their alleged ontological features; namely, by contending that it is strongly implausible that entities with the queer feature of being objectively prescriptive actually exist. Field (1989) questioned the existence of numbers conceived of as Platonic entities by arguing that—roughly—there cannot be any adequate epistemology of entities of this sort, for there is no acceptable explanation of how our knowledge of them could be reliable. Let “S(X)” stand for a minimally truth-apt and contingent sentence that purports to describe—in accordance with its surface grammar—entities of some kind X. For many different kinds of entities (e.g., kangaroos, protons, desires, stars, etc.), the sweeping minimalist is apparently uncommitted to the a priori truth of the instances of the following schema: If it is warrantedly assertible that S(X), then S(X).77 A consequence is that the sweeping minimalist could in principle accept an error-theory about all these entities without facing any immediate conceptual incoherence or logical inconsistency. Precisely—supposing that the relevant entities are of kind X—the sweeping minimalist might try to argue that there are compelling reasons (ontological, epistemological, or whatever) for believing that no X-entity actually exists. The consequence would be that no warrantedly assertible sentence that entails the existence of X-entities could prove (even minimally) true. A typical tacit presupposition of arguments of error-theorists about X-entities is that these entities would exist mind-independently if they existed at all.78 This presupposition appears acceptable by the sweeping minimalist too, who can prima facie maintain that mind-independent entities exist or
76
Precisely, this would hold for the sentences that entail that X-entities exist. The same would seem to hold for the reverse schema: If S(X), then it is warrantedly assertible that S(X). 78 For example, if the existence of X-entities were thought of as mind-dependent—precisely, dependent on our warranted sentences about them—the traditional arguments of error-theorists would no longer make. For many sentences that entail the existence of X-entities (e.g., moral properties and numbers) are plausibly warrantedly assertible, with the consequence that the error-theorist would be forced to conclude that X-entities do exist and that the sentences that entail the existence of these entities are true and not false. 77
On Creeping Minimalism and the Nature of Minimal Entities 163 may exist (more on this below). At any rate, though typical, this presupposition is not strictly necessary for the formulation of an error-theory.79 A general limit of the applicability of the error-theory approach is that, though it proves appropriate in the domains where it is reasonable to cast doubts on the existence of the target entities—for example, in metaethics, philosophy of mathematics and philosophy of theoretical science80 —the error-theory approach cannot apply globally, that is to say, to the whole of reality. For no sensible person would endorse a comprehensive error-theory asserting that nothing at all exists! (Further restrictions to the applicability range of the error-theory strategy that depend on allowing the existence of minimal entities are suggested in the next section.) Let us turn to antirealism. The antirealist approach can probably apply to most ontological fields.81 Indeed, antirealist arguments against realism have been put forward to apply globally. For instance, the arguments for internal realism made in Putnam (1978 and 1981) can be interpreted as aiming to show that reality as a whole is mind-dependent. What consequences follow from the introduction of creeping minimalism for the dispute between realists and antirealists? Dreier (2004, 29–30) seems to contend that when creeping minimalism breaks through, appealing to the notion of mind-independence to characterize realism and to distinguish it from antirealism produces conceptual puzzlement so that it is no longer effective.82 The problem would arise from the fact that many things prove mindindependent in accordance with some everyday sense of “mind-independence.” Dreier assumes that the sweeping minimalist can use these trivial
79 For instance, as I have suggested before, a noncognitivist affected by creeping minimalism could try to rebuff her commitment to the existence of X-entities by substantiating the thesis that we all have a propensity to apply incorrectly the justification standards of the sentences about X-entities, with the result that whenever a sentence of this type appears warrantedly assertible it is in fact not so. This error-theory strategy does not require assuming that X-entities are mind-independent. Indeed, Wright (1992, 86–87) includes this strategy among those available to error-theorists about mind-dependent entities. 80 For instance, van Fraassen’s (1980) constructive empiricism can be seen as a form of error-theory about unobservable entities grounded, principally, in the theory underdetermination thesis. 81 Notice that noncognitivist positions have also been proposed to apply globally to the whole of reality. See for instance Macarthur and Price (2007). In accordance with my arguments before, I tend to believe that as creeping minimalism breaks through, the advocates of such global positions could effectively challenge realism only if they resorted to antirealist arguments. 82 More exactly, Dreier writes that trying to distinguish between real and nonreal entities on the basis of a distinction between substantive senses of mind-independence and trivial senses of it “doesn’t look very promising” (2004, 30).
164 Luca Moretti notions of mind-independence together with her minimal semantical and ontological notions. This is indeed plausible. Dreier gives this example: it is a commonsense truth that slavery would not cease to be morally wrong if only everyone started believing that slavery is not so.83 The sweeping minimalist can say that the moral wrongness of slavery is independent—in this trivial sense—of everyone’s beliefs and so of everyone’s mind (Dreier 2004). Similar examples can be found in other fields. For example, it appears true that, say, penguins would not stop populating the South Pole, and the earth would not stop orbiting around the sun if we only ceased believing so, or if we had no knowledge of these facts at all. These things—and many others—can be said to be mind-independent in such trivial senses by the sweeping minimalist. Thus, if realism is characterized through notions of mind-independence, it seems that the antirealist who falls prey to sweeping minimalism can assert just what the realist can assert, though it is dubious that the antirealist actually becomes a realist. For it is implausible that realism (about, say, moral wrongness, penguins, and many other things) could be obtained in such a cheap way. I think that Dreier is too quick in drawing his skeptical conclusion. I believe that the problem just exposed can be dissolved in this way: what is at stake here is the characterization of realism as a philosophical position rather than a commonsense view. Ordinary ascriptions of mind-independence do not characterize realism conceived of as a philosophical position simply because realism, in philosophy, is typically not characterized in terms of these ordinary ascriptions but in terms of more qualifi ed and thus more sophisticated ascriptions of mind-independence (I give some examples below).84 Furthermore, the mere fact that there exist ordinary or commonsense notions of mind-independence does not disqualify the more sophisticated notions of mind-independence typically used in philosophy to characterize realism from having this function. Why should this ever happen? Finally, ordinary ascriptions of mind-independence can possibly sustain some forms of “realism”—perhaps everyday or ordinary forms of realism—but these positions would not coincide with the forms
83 Precisely, for Dreier, what is commonly accepted is that “slavery would not be morally wrong, if only we slavery-haters mellowed out a little” (2004, 29–30). 84 Consider for instance the commonsense claim that penguins would not stop populating the South Pole if we only ceased believing so. A more qualified claim that might provide a philosophical sense of mind-independence characterizing realism could be this: it would still be the case that penguins populate the South Pole even if a rational agent in sufficiently good (or ideal) epistemic conditions would believe that penguins do not populate the South Pole. (Notice that this claim is philosophically more controversial than the original.) The notion of mind-independence presupposed here does not seem to be part of our ordinary conceptual background.
On Creeping Minimalism and the Nature of Minimal Entities 165 of realism that are central to the debate on realism and antirealism in contemporary metaphysics.85 To strengthen my point consider also that, in the contemporary debate on realism and antirealism in metaphysics, the antirealist is typically supposed to accept ordinary ascriptions of mind-independence as true. In other words, it is generally assumed that the antirealist—and not just the realist—should avoid revising ordinary beliefs, on pain of rendering her philosophical position implausible. The antirealist should be able to endorse claims normally made in everyday life, which include—very plausibly— commonsense ascriptions of mind-independence like the above ones.86 Consequently, these ascriptions cannot by themselves entail realism, according to the philosophical sense of “realism.” A couple of examples might clarify and support these points. A notion of mind-independence deployed by metaphysicians to characterize realism—at least in certain ontological fields—is that of potential evidencetranscendence (Wright 1992, 77–78). Accordingly, one is a realist about entities of type X if one believes that [I] X-entities exist and that [II*] they are potentially evidence-transcendent. Condition [II*] can be spelled out by stating that the existence of X-entities is not essentially (or necessarily) knowable.87 This allows the existence of X-entities to be only contingently knowable. The complementary characterization of antirealism says that one is an antirealist about entities of type X if one believes that [I] X-entities exist and that [~II*] these entities are evidence-constrained—namely, that their existence and, plausibly, their characteristic features are essentially (or necessarily) knowable.88 Typically, the claim
85 It is worth emphasizing that the philosophical position called commonsense realism typically does not state that everyday notions of mind-independence like those considered above defi ne or characterize realism. “Commonsense realism” typically refers to the position according to which the world of commonsense is real—for instance, ordinary objects with their ordinary features (e.g., colors, shapes, etc.) exist really. 86 It might be helpful to notice that idealist positions stating that everything is just mental play no role—or no significant role—in contemporary antirealism. 87 This condition might in turn be spelled out by saying—for instance—that there is a possible world in which there exist X-entities and such that it is not the case that if there were a rational agent in good (or ideal) epistemic conditions for judging whether there exist X-entities, she would believe that there exist X-entities. 88 This condition might in turn be spelled out by saying—for instance—that (i) in all possible worlds, there exist X-entities if and only if there were a rational agent in good (or ideal) epistemic conditions for judging whether there exist X-entities, she would believe that there are X-entities, and (ii) being in good (or ideal) epistemic conditions for judging whether there exist X-entities does not entail by itself that there exist X-entities.
166 Luca Moretti is that it is a priori true that the existence and the characteristic features of X-entities are knowable. These characterizations of realism and antirealism appear prima facie suitable to apply to middle-sized objects and so to penguins. Both the realist and the antirealist about penguins and their habits will agree with common sense that penguins would not stop populating the South Pole if we only ceased to believe so or if we had no knowledge of these facts. Yet the realist about penguins and their habits will not be a realist because of these trivial convictions—she will be so because she believes that penguins and their habits exist, and they are potentially evidence transcendent. On the other hand, the antirealist about penguins and their habits can share the trivial convictions of the realist, for these convictions are compatible with the antirealist’s qualifying thesis about penguins and their habits. In other words, the thesis that the existence of penguins and their habits is essentially knowable is compatible with the conviction that we could (mistakenly) change our beliefs about these things and with the conviction that we could have known nothing at all about these things.89 In metaethics, Harman (1977), Wiggins (1987), and Sturgeon (1988)— among others—consider variants of a criterion for moral realism that can be interpreted as based on a notion of mind-independence alternative to the one described above. In a crude version of this criterion, one is a moral realist if one believes that [I] moral facts or properties exist and that [II**] they figure ineliminably in the best explanation of how moral experience is produced.90 Condition [II**] seems to provide the alternative notion of mind-independence. For, intuitively, the feature of a fact or property of figuring ineliminably in the best explanation of how a certain experience is produced marks out that fact or property as a piece of the furniture of reality we can interact with and distinguishes that entity from what should rather be considered to be a mere “linguistic projection” or a reification of our concepts. The complementary condition qualifying moral antirealism can be formulated as follows: one is a moral antirealist if one believes that [I] moral facts or properties—in some deflationary sense of “facts” and “properties”—do exist, and that [~II**] they do not figure in the best explanation of our moral experience.91
89 Or, at least, there are several straightforward qualifications of “knowable” available to the antirealist that entail that the former claim is compatible with the latter two. 90 For criticism of criteria for moral realism of this type, see for instance Wright (1992, 176–201). 91 Suppose, for instance, I see John kill Bob for the mere fun of it and that I come to believe, as a reaction, that John’s act is horribly wrong. The moral realist will explain my belief by mentioning—among other things—the moral fact that John’s act is wrong. The moral realist will contend that I was somehow sensitive to
On Creeping Minimalism and the Nature of Minimal Entities 167 Using this framework, the debate between moral realists and moral antirealists can plausibly proceed independently of questions concerning the capacity of moral facts to be potentially evidence-transcendent. Also, with this framework, both the moral realist and the moral antirealist appear entitled to endorse commonsense ascriptions of mind-independence about moral facts and moral properties. Thus, both the realist and the antirealist can accept that slavery would not cease to be morally wrong if only everyone started believing (mistakenly) that slavery is not so. The difference is that the realist will assert that the fact that slavery is morally wrong figures ineliminably in the best explanation of our moral experience, and the antirealist will deny it. The sweeping minimalist can plausibly endorse characterizations of realism and antirealism that come in terms of the notions of mind-independence discussed above. Clearly, to do so, she has to acknowledge that, together with everyday notions of mind-independence, we also possess philosophically more substantive notions of mind-independence, but this does not seem to be a problem for the sweeping minimalist. For, as I have said, creeping minimalism is prima facie compatible with pluralism about notions and entities.92 To recap, in this section I have accepted a generic characterization of realism according to which one is a realist about certain entities if and only if one believes that [I] these entities exist and [II] they are mind-independent. This ontological characterization of realism is intuitive and less controversial than alethic characterizations of realism. I have argued that, as creeping minimalism takes place, framing the debate between realists and nonrealists as a dispute between realists and noncognitivists is probably fruitless. I have suggested that there are two further ways to frame this
the moral wrongness of that act. She will insist that, without mentioning that moral fact (and that moral property), no satisfactory explanation of why I entertained my belief in that context is possible. On the other hand, the moral antirealist will accept that the sentence “John’s act is morally wrong” is justifiable; consequently, the moral antirealist will be able to assert both “it is a fact that John’s act is morally wrong,” and “John’s act has the property of being morally wrong,” in accordance with defl ationary senses of “fact” and “property.” The moral antirealist will, however, explain my belief that John’s act is horribly wrong by adducing—among other things—my sensitivity to nonevaluative features of John’s act (and of the situation I witnessed), and she will consider redundant to this explanation mentioning the fact that John’s act is morally wrong and the property of being morally wrong possessed by John’s act. 92 Further substantive notions of mind-dependence/independence are plausibly available to the sweeping minimalist—for instance, the important notion of judgment-dependence/independence. For a careful analysis of this notion, see Wright (1992, 108–139) and Miller (2005).
168
Luca Moretti
debate that look interesting and productive: one consists in construing it as a dispute between realists and error-theorists, and the other consists in construing it as a dispute between realists and antirealists.
4 THE ONTOLOGICAL STATUS OF MINIMAL ENTITIES AND THE DEBATE BETWEEN REALISM AND NONREALISM I have shown in §2.3 that the sweeping minimalist becomes committed to the existence of minimal entities through, typically, performing something-from-nothing inferences. Indeed, this will happen quite often, as the sweeping minimalist will frequently be in circumstances that justify sentences from which a something-from-nothing inference can stem. Does any crucial consequence for the debate between realism and nonrealism follow from the claim that minimal entities do exist? Answering this question presupposes clarifying the ontological status of minimal entities. I substantiate the thesis that minimal entities are language-dependent, in the strong sense that they do not exist outside of language. Minimal entities are nothing but linguistic posits—that is, hypostatizations or reifications of linguistic constructs introduced via certain linguistic practices. From this it follows—I contend—that the thesis that minimal entities do exist has no significant consequence for the debate between realists and nonrealists, at least when this dispute is framed as one between realists and antirealists. Schiffer has put forward an interesting explanation of why minimal entities should be considered to be linguistic posits. (This explanation has originally been formulated to apply to pleonastic entities but—if correct—it could apply to minimal entities too). Briefly, Schiffer contends that: There is an important difference between, on the one hand, “linguistic posits” like . . . [pleonastic] properties and propositions and, on the other hand, those entities that are not linguistic posits, entities enjoying the highest degree of independence from our linguistic and conceptual practices. The difference is that the essences of the latter can be discovered by a posteriori scientific investigation, whereas those of the former can’t be discovered in any such way. Whatever belongs to their essence can be read off the something-from-nothing linguistic practices that posit them in our ontology. (Schiffer 1996, 161) (In this passage, Schiffer evidently refers to Kripke’s and Putnam’s celebrated views that the essential features of real things and kinds of real things can be determined by a posteriori investigation.) Schiffer’s suggestion amounts to the claim that what marks out the language-dependent nature of a minimal entity is the fact that its essential
On Creeping Minimalism and the Nature of Minimal Entities 169 features can be known a priori by reading them off the relevant linguistic practices. I find Schiffer’s suggestion questionable. Consider a full-blooded realist about certain entities of type X. She surely believes that X-entities enjoy “the highest degree of independence from our linguistic practices.” Yet such a realist does not appear prima facie committed to believing that the essential features of these language-independent entities “can be discovered by a posteriori scientific investigation.” For she could believe, instead, that the essential features of the X-entities can be known a priori through the knowledge of relevant analytical sentences (or of the correlated linguistic practices). Entertaining this belief requires endorsing the thesis that there are analytical sentences and the thesis that analytical sentences can be known a priori to be true of language-independent facts. This does not seem particularly problematic. Philosophers—prominently, Boghossian (1996, 2003)93 —have given powerful arguments to support both theses.94 In conclusion, Schiffer claims that what marks out the language-dependent nature of minimal entities is the fact that the essential features of these entities can be known a priori. This claim is dubious, as it can be argued that this very characteristic could be shared by fully language-independent entities.95
93 Boghossian (1996, 2003) gives quite persuasive arguments to salvage the notion of analyticity from Quine’s famous criticism and to show that analytical sentences can be known a priori to be true of fully objective facts. Precisely, Boghossian distinguishes between metaphysical and epistemic analyticity. A sentence “S” is metaphysically analytical if and only if “S” is true solely in virtue of its meaning, and “S” is epistemically analytical if and only if a grasp of the meaning of “S” suffices to justify the belief that “S” is true. Boghossian argues that only the notion of metaphysical analyticity but not that of epistemic analyticity succumbs to Quine’s objections. He also argues that, although metaphysically analytical sentences are supposed to owe their truth to their meanings only, without any contribution from the facts, epistemically analytical sentences are true of independent facts. For Boghossian, epistemically analytical sentences divide into Frege-analytical and Carnapanalytical sentences. A sentence is Frege-analytical if and only if its analyticity is to be explained by the fact that it is transformable into a logical truth by substitution of synonyms for synonyms. A sentence is Carnap-analytical if and only if it is part of an implicit definition of certain of its component terms. The platitudes characterizing minimal entities can probably be interpreted as Carnap-analytical. 94 The clear articulation of my reply to Schiffer has greatly benefited from discussion with Tommaso Piazza. I am especially indebted to him for the reference to Boghossian’s work. 95 Someone might suggest alternative reasons to substantiate the claim that minimal entities are linguistic posits. For instance, Schiffer (2003, 62–63) emphasizes that the introduction of pleonastic—and thus minimal—entities in our ontology will not change the preexistent causal order. Accordingly, it might be thought that minimal entities are linguistic posits because they do not satisfy the eleatic
170 Luca Moretti To obtain a persuasive clarification of why minimal entities are nothing but linguistic posits, we have to shift from epistemology to semantics. That minimal entities are language-dependent becomes clear as we realize that the deepest and most complete explanation of our ability to refer to thin entities of this type can be given by making use of a mere defl ationary notion of reference; for example, a notion of reference grounded in principle [R*] introduced in §2.1 and used in §2.3. Indeed, it seems to be correct to say that reference to minimal entities can be explained only if a deflationary semantical notion of this type is made use of. For it would make no sense to try to use any substantive—for example, naturalistic—notion of reference to explain it. Intuitively, if we could use complex semantical notions to explain reference to minimal entities, the latter would also possess a complex nature that would be described by principles well more substantive than mere platitudes. For instance, a causal notion of reference could apply to minimal entities if the latter had causal powers (like tables, cats, and electrons). Yet that minimal entities have causal powers is not stated by the platitudinous principles that define the nature of these objects. Consequently, minimal entities are very plausibly causally inert.96 (I return to this shortly.) If minimal entities were independent of our linguistic practices, the deepest and most complete explanation of our ability to refer to them would certainly have to involve a different and more complex notion of reference. For we would have to clarify how our words can hook up to entities extrinsic to language. Yet [R*] and deflationary notions of reference in general establish no link between language and the extralinguistic world. The conclusion we should draw is that minimal entities cannot but be intrinsic to language. As Thomasson (2001, 323–324) has emphasized, to claim that minimal entities depend on our linguistic practices does not mean, however, to claim that minimal entities are created by our linguistic practices. For the mere existence of those practices does not bring about, by itself, the existence of minimal entities. Reality has to collaborate to some extent to make true the basic sentences from which the appropriate something-from-nothing inferences start off.97
criterion according to which everything that really exists makes a difference to the causal powers of something (Armstrong 1997, 41). A problem is that the eleatic criterion is not unanimously accepted. For instance, realists about possibilia and abstract entities typically reject it. 96 It seems plausible that whether entities of a given type are (completely) causally inert or have causal powers must be something essential to those entities and not merely contingent. 97 That minimal entities are language-dependent in the sense explained plausibly entails that minimal entities are mind-dependent. It is very important to appreciate that mind-dependence, in this sense, does not coincide with being evidence-constrained, for it is not absurd to think of minimal entities the existence of
On Creeping Minimalism and the Nature of Minimal Entities 171 The above consideration suggests that, in explaining reference to minimal entities, the sweeping minimalist may need to mention the attainment of conditions that are extrinsic to language. For instance, the full account of how we refer to the minimal property of being round involves maintaining that a sentence asserting that this property exists is true; for example, the sentence “there exists something instantiated by this ball; that is, the property of being round.” This in turn involves maintaining that the more basic sentence “this ball is round”—which implies the former—is true. Finally, contending that “this ball is round” is true requires substantiating the claim that the truth conditions of this sentence—which may be extrinsic to language—actually obtain. Thomasson has suggested (in personal communication) that if one appeals to the attainment of conditions that are extrinsic to language in the explanation of reference to minimal entities, one can probably develop a semicausal (or hybrid) account of reference to these entities. The upshot would be that it is false that minimal entities are just intrinsic to language. For a semicausal account of reference would just explicate how terms that refer to minimal entities hook up to things extrinsic to language. I am quite skeptical about the possibility of developing a semicausal account of reference to minimal entities.98 What is a semicausal theory of reference? Pure causal theories of reference face the qua problem, which arises because, as long as a pure causal connection allows a speaker to refer to something, it cannot in general determine the exact sort of thing she refers to. This holds for both singular terms and sortal terms. (See for instance Devitt and Sterelny 1999, 63–65, 72–75.) For example, suppose I want to ground the reference of the term “cat” through a causal link by pointing to my cat Birba. What makes it the case that “cat” refers just to
which is only contingently but not essentially knowable. Suppose, for instance, I’m a realist about this ball, and I believe that it is potentially evidence-transcendent (though contingently knowable) that this ball is round. Through a pleonastic transformation, I am committed to believing that it is potentially evidence-transcendent that this ball has the property of being round. This could in turn be clarified by saying—for instance (see n. 87)—that there is a possible world w at which this ball has the property of being round and such that it is not the case that if there were a rational agent in good (or ideal) epistemic conditions to judge whether this ball has the property of being round, she would believe that this ball has the property of being round. Notice that the expression “the property of being round” does not refer to anything that exists at w language-independently. Minimal entities are intrinsic to language because the actual and the possible referents of the terms that refer to them are intrinsic to language. 98 Although I will not pursue this issue here, semicausal accounts of reference have been found problematic in different respects. For typical objections and some interesting replies, see Thomasson (2007, 48–53).
172 Luca Moretti the natural kind cats rather than, say, to the natural kind felids, or to the individual Birba, or to a spatial or temporal part of her? Devitt and Sterelny suggest that we could answer this question by endorsing a semicausal or hybrid theory of reference according to which, basically, those who intend to ground the reference of a term by pointing to something must have in mind some descriptive category under which that thing falls. And “the grounding will fail if the cause of the perceptual experience does not fit the general [descriptive] term used to conceptualize it” (1999, 80). Thomasson (2007) has developed in detail and enhanced Devitt and Sterelny’s initial suggestion (Thomasson 2007, 38–45). Thomasson argues that, on a semicausal theory of reference: To disambiguate whether or not singular and general terms refer, and if so to what they refer, these must have some minimal associated conceptual content. This conceptual content establishes the sort of individual (or kind) that the term is to pick out, outlining frame-level conditions for the existence and identity of the referent, if any. (Thomasson 2007, 161–162) An interesting result is that: The conceptual content in these terms is enough to ground certain conceptual truths based on interrelations among these terms, since the conditions for application of one term (or set of terms) may be necessary, sufficient, or otherwise relevant to the application of the other, or the satisfaction of the truth-conditions for one sentence may be sufficient for the application of the new nominative term. (Thomasson 2007, 162) An example of how the satisfaction of the application conditions of a term is sufficient to satisfy the application conditions of another term concerns the predicates “house” and “building” (Thomasson 2007, 162). It seems correct to say that, if the former predicate applies to a given entity, the latter must apply to it too, for everything denoted by the latter predicate is denoted by the former predicate. The result is that if “building” denotes only language-independent entities, “house” must denote only languageindependent entities too. Thomasson recognizes that, in the case of minimal (or pleonastic) entities, it is typically the satisfaction of the truth-conditions for one sentence that suffices to satisfy the application conditions of a term referring to entities of this kind (2007, 162). For example, the satisfaction of the truthconditions of the sentence “this ball is round” is sufficient to satisfy the application conditions of the singular term “the property of being round,” in the sense that this term is guaranteed to refer (2007, 163). Thomasson would seem to believe that the existence of conceptual links of this type is
On Creeping Minimalism and the Nature of Minimal Entities 173 sufficient to ground a semicausal account of reference to minimal entities. I believe that this is probably false. As we have seen in §2.3, the sweeping minimalist is plausibly committed to accepting the existence of conceptual links of this type. The difference between the sweeping minimalist as such and Thomasson is that the sweeping minimalist as such thinks of the relation of reference to minimal entities only in deflationary terms, but Thomasson seems to believe that this relation can in certain cases be identified with something more substantive. What Thomasson has plausibly in mind is that when the truth-conditions of the sentence “this ball is round” are satisfied by some objective state of affairs—that is, by a state of affairs fully independent of language—the abstract singular term “the property of being round” can be argued to refer to the property of being round—whatever this entity might be—through a semicausal connection and so through (also) a causal link. The problem with this claim is that as long as the property of being round is a minimal entity, which involves that its nature is characterized entirely a priori in terms of general and platitudinous principles like those considered in §§2.2 and 2.3, we are not allowed to assert that this entity can sustain causal links. And if we conceive of the nature of this property as something ontologically more substantive—by introducing metaphysical assumptions— this entity will no longer be minimal. For example, if the referent of the abstract singular term “the property of being round” was identified with the class of all round-tropes, where each of these tropes was conceived of as provided with causal powers, we could probably work out a semicausal account of reference to this property. But a property characterized in this way is no longer minimal! Analogous considerations apply to the other minimal properties and to other minimal entities.99 Now, let me clarify how the thesis that there are minimal entities can be settled into the realist’s metaphysical background without evident difficulties. The existence of minimal entities is typically ensured by inferences based on certain platitudes. These inferences are such that, from a true sentence containing no term that refers to a minimal entity, one gets a sentence embedding a term that does refer to a minimal entity. For example, a sweeping minimalist can conclude that the property of being approximately cylindric is instantiated by Ribot so that there exists a given property, if and only if it is true that Ribot is approximately cylindric. Hence,
99 For instance, the mere fulfi lment of the language-independent truth-conditions of “this ball is round” does not seem to commit the sweeping minimalist to maintaining that the individual term “that this ball is round” and the predicates “is a fact” respectively refers to and denotes facts through a semicausal link. For the platitudinous notion of a fact is not characterized in terms of principles that enable minimal facts to enter into causal relations.
174
Luca Moretti
minimal entities typically exist if and only if reality collaborates to make true certain relevant basic sentences.100 Let “S” be a true sentence containing no term that refers to a minimal entity. Suppose “S” is “Ribot is approximately cylindric” (where Ribot is a penguin). A realist about penguins is committed to maintaining that the truth-conditions of “S” obtain and that what fulfi ls these conditions is mind-independent in some nontrivial sense of mind-independence. As our realist goes sweeping minimalist, she will be able infer from “S” the sentence “S 1” = “Ribot has the property of being approximately cylindric” and the sentence “S 2” = “there is something instantiated by Ribot; that is, the property of being approximately cylindric.” Given that, for the sweeping minimalist, “S,” “S 1,” and “S 2” can be logically inferred from one another, the realist will be committed to holding that these three sentences are true and share the same mind-independent truth-conditions. Notice however that the contents of “S,” “S 1,” and “S 2” are apparently very different. For “S” says nothing at all about properties, “S 1” says something of a property, and “S 2” asserts the existence of that property. Let us also assume that our realist about penguins is an austere nominalist who claims
100 There might, however, be minimal entities depending on principles entailing that they exist necessarily (i.e., in all possible worlds). As I have emphasized (n. 53), Schiffer’s arguments for the claim that pleonastic propositions and pleonastic properties exist necessarily are dubious. Yet, in Moretti (2007a), I have independently argued, from less disputable premises, that the framework of Wright’s minimalism—and thus of creeping minimalism—entails that at least tautological and analytical propositions exist necessarily. I have also argued that this consequence commits the minimalist to the thesis that tautological and analytical propositions are real entities in that they exist mind-independently, because they exist in possible worlds where there is no speaker or rational agent. (Schiffer [2003, 50, 59–60] argues for the same conclusion in relation to all propositions). I no longer claim that the minimalist is committed to the thesis that tautological and analytical propositions are real, at least in the sense explained. An antirealist about propositions could probably accept that propositions exist in all possible worlds and so in those where there is no rational agent. For the mind-dependence of propositions could be defi ned—in accordance with what is suggested in n. 87—in terms of a counterfactual condition that does not require the existence of a rational agent. Roughly, there are propositions if and only if; if there were a rational agent in good (or ideal) epistemic conditions to judge whether there are propositions, she would believe that there are propositions (Moretti 2007a, 32). Brogaard and Salerno (2005) have made a case that counterfactual defi nitions of this sort are unacceptable because they entail a conditional fallacy. In Moretti (2007b), I have however shown that Brogaard and Salerno’s argument is inconclusive.
On Creeping Minimalism and the Nature of Minimal Entities 175 that no property exists. How could such a realist reconcile all these apparently inconsistent claims? Consider fi rst that many realists endorse a commonsense view according to which our descriptions often include among their components linguistic constructs that reflect no objective feature of the world. Accordingly, many realists believe that the same reality can often be described in different and alternative ways via incorporating linguistic constructs that reflect no objective feature of the world in our descriptions or replacing some of the constructs of this type already in use with different constructs of the same sort. These considerations plausibly apply also to the mind-independent states of affairs that, according to realists, fulfi l the truth-conditions of “S.” Suppose that one of these realists—let us call her Mary—is both a philosopher and an anatomist. Mary gives at least two alternative descriptions of the truth-conditions of “S”: in certain cases, she uses “S” itself to describe them and, on other occasions, she uses the sentence “S1*” = “the sum of the undetached parts of Ribot is approximately cylindric” (imagine that, for an anatomist, it is sometimes more natural to speak of a sum of organic parts rather than a whole). Suppose fi nally that from “S1*,” Mary is prone to infer back and forward the sentence “S 2*” = “there is something approximately cylindric: that is, the sum of the undetached parts of Ribot.” Mary considers “S 2*” to be an innocent reformulation of “S1*,” and so as a further alternative way to describe the truth-conditions of the original sentence “S.” As a philosopher, Mary believes that the expression “the sum of the undetached parts of Ribot” is just a linguistic construct that, if interpreted literally, refers to nothing that exists in the independent world. Yet, when Mary infers “S 2*,” she just asserts that the sum of the undetached parts of Ribot does exist. To make her inferential practices coherent with her ontological convictions, Mary can stipulate that the expression “the sum of all undetached parts of Ribot” refers just in a deflationary sense of “refers.” She can argue that, consequently, the referent of this expression exists only intrinsically to her language. This entity could be described as a reification or hypostatization of her linguistic construct “the sum of all undetached parts of Ribot.” Realists affected by creeping minimalism could treat minimal entities in the very same way as Mary does with her linguistic posit. Indeed, it seems plausible to me that once a realist comes to realize that the only notion of reference suitable to minimal entities is platitudinous, she will naturally conclude that these entities are nothing but reifications of linguistic constructs (or something like this). Let us go back to our austere nominalist realist who falls prey to creeping minimalism. Our philosopher accepts the sentence “S” = “Ribot is approximately cylindric” as mind-independently true. Consider now how she deduces “S 2” from “S.” First, she derives from “S,” via nominalization, “S1” = “Ribot has the property of being approximately cylindric.” The step of nominalization is just one that introduces a linguistic construct (in this case: “the property of being approximately
176 Luca Moretti cylindric”). The realist can interpret this linguistic construct as one that picks up nothing that exists in the independent world. The realist then derives from (a platitudinous reformulation of) “S1,”101 via quantification, “S 2” = “there is something instantiated by Ribot: that is, the property of being approximately cylindric.” The realist can interpret the step of quantification as one that reifies or hypostatizes the linguistic construct introduced formerly. She can at this point maintain that “S,” “S1,” and “S 2” are equivalent descriptions of the same mind-independent state of affairs and that the difference in factual content between these sentences is just apparent. For it arises from the introduction—in “S1” and “S 2”—of a construct that reflects no objective feature of reality. The realist can emphasize that asserting “S 2” does not commit her to holding that the property of being approximately cylindric exists objectively, for this entity is just a reification of a linguistic construct. Realists can plausibly deal with the other minimal entities considered in this chapter in a similar fashion. The foregoing discussion shows that the thesis that there are minimal entities can be settled into the metaphysical background of the realist without apparent difficulties. Allowing the existence of minimal entities does not seem to undermine the realist position, but it does not appear to bolster its plausibility either. Using minimal notions and allowing quantification over minimal entities will increase the expressive power of the language of the realist, but this apparently involves nothing about the truth or the plausibility of the content of the realist doctrine. Let us now turn to the two forms of nonrealism that, as I have argued in §3, are apparently unharmed by the phenomenon of creeping minimalism—namely, antirealism and the error-theory. Let us consider fi rst antirealism. As minimal entities are intrinsic to language, the thesis that there are minimal entities can plausibly be settled into the metaphysical background of the antirealist without problems. For allowing the existence of these reifications of linguistic constructs will result in, so to say, a mere extension of the stock of furniture of mind-dependent reality. Furthermore—to parallel the realist position—the existence of minimal entities would seem, on the one hand, not to undermine the antirealist position but, on the other, not to enhance its plausibility either. Let us turn to nonrealism construed as an error-theory. In this case, too, allowing the existence of minimal entities might appear harmless. The error-theorist’s typical claims are negative: she asserts that the target entities do not exist and that sentences that aim at describing them are false. It might appear implausible that a something-from-nothing inference stemming from sentences like these could yield problems for the error-theorist. Indeed, some difficulty emerges when the objects constituting the targets of
101 Namely, “the property of being approximately cylindric is instantiated by Ribot.”
On Creeping Minimalism and the Nature of Minimal Entities 177 the error-theory have a counterpart among the minimal entities. Consider for example an error-theorist about propositions who becomes a sweeping minimalist. She will assert, as an error-theorist, that propositions do not exist. Yet, as a sweeping minimalist, she will be committed to asserting that many propositions do exist. For example, given her statement that propositions do not exist, she will be committed to claiming, via platitudinous inference, “there exists something true: that is, the proposition that propositions do not exist.” This does not turn the error-theorist into a realist, as the proposition that propositions do not exist is a minimal entity, which—as I have argued—exists only language-dependently. The snag is, however, that error-theorists as such are not conceptually equipped to make this rejoinder. For this reply relies on—plausibly—an antirealist argument (briefly, the error-theorist will have to distinguish between thick propositions and thin propositions and argue that thin propositions are language-dependent and thus not fully mind-independent). Analogous examples can be found concerning other entities—for instance, facts and properties. In conclusion, when an error-theorist goes sweeping minimalist, she can no longer target objects that have counterparts among the minimal entities by only applying error-theory arguments—she ought to appeal to antirealist arguments too. The phenomenon of sweeping minimalism has thus some interesting consequences at least for the debate between realists and error-theorists. In this section, I have attained two important results: fi rst, I have established that minimal entities are just intrinsic to language; second, I have given reasons to believe that the introduction of minimal entities in ontology is inconsequential for the fate of the dispute between realists and antirealists. A conclusion that can be drawn now, on the grounds of this and the former section, is that the phenomenon of creeping minimalism has no significant consequence for the debate between realism and nonrealism, as long as this debate is framed as a dispute between realists and antirealists.
5 CONCLUDING REMARKS AND BACK TO DREIER’S PROBLEM In the fi rst part of the chapter (i.e., §§2, 2.1–2.3), I ventured an explanation of the ramping up phenomenon that Dreier calls creeping minimalism by interpreting it, not as a doctrine, but as a widespread tendency. On this interpretation, creeping minimalism coincides with a propensity to make a massive use in philosophical discussion of elementary regimentations of folk semantical and ontological notions—that is, of concepts implicit in natural language that have functions other than giving illuminating explanations. One important function of many minimal notions is—I have suggested—that of enhancing the expressive power of language by allowing the explicit formulation of generalizations. Sweeping minimalists are not necessarily dogmatic and may embrace forms of pluralism about notions
178 Luca Moretti and entities, in the sense that they may believe that we also possess or can defi ne thicker and more controversial versions of our minimal notions for the purpose of achieving genuine philosophical explanations. I have suggested that the sweeping minimalist attitude might have been triggered by independent works by Paul Horwich (and other deflationists about truth), Crispin Wright, and Stephen Schiffer. These philosophers do not aim to provide elementary regimentations of ordinary notions (at least, they do not say so explicitly), but this interpretation of their works emerges quite naturally. I have shown that the views defended by these authors can be “assembled” together to produce a conspicuous part of or even the entire framework apparently presupposed by sweeping minimalism. In the second part of this chapter (i.e., §§4, 5), I investigated whether entertaining the sweeping minimalist attitude produces any significant consequence for the debate between realism and nonrealism in metaphysics; this involved looking into the ontological status of minimal entities. I characterized realism about X-entities as the ontological (or metaphysical) position according to which X-entities exist and are mind-independent. An interesting result concerns alterations in the “geography” of the possible nonrealist positions brought about by the rise of creeping minimalism. When creeping minimalism breaks through, the irrealist perspectives, which challenge realism by questioning the existence dimension, face difficulties. Precisely, a serious problem that affl icts the noncognitivist positions, like expressivism and instrumentalism, is that it becomes hard to understand in what sense these positions still constitute a challenge for realism. Probably, when creeping minimalism takes place, noncognitivism (i.e., the position it turns into) can no longer be considered to be a form of nonrealism. A minor difficulty, which affects the error-theory, is that certain possible forms of error-theory that target specific objects (e.g., propositions) could still stand up only with supplementation by antirealist arguments. A significant—though perhaps expected—outcome of my investigation is that creeping minimalism has no substantial consequence for the acceptability of realism and antirealism, in the sense that neither the realist nor the antirealist can take advantage of the framework of creeping minimalism to the detriment of their opponent. This result is important because the realist–antirealist construal of the debate about realism and nonrealism is apparently the most comprehensive of all in terms of possible applications—disputes between realists and antirealists can possibly be developed to apply globally, to the whole of reality. Creeping minimalism leaves disputes like these untouched. Another important result is the defi nite answer I have given to the question of the ontological status of minimal entities. These entities are language-dependent in the strong sense that they do not exist outside language because it is proper to the nature of these entities that only deflationary notions of reference can explain our ability to refer to them—these notions of reference can establish no link between language and the extralinguistic
On Creeping Minimalism and the Nature of Minimal Entities 179 world. This does not mean, however, that minimal entities are just created by our linguistic practices. Reality has to collaborate to fulfil the conditions of truth of the basic sentences that entail that there are minimal entities. At the end, after this long journey, let us return to the fascinating problem uncovered by Dreier that originally elicited my investigation into creeping minimalism and the nature of minimal entities. Dreier’s puzzle is the following: the sweeping minimalist appears to be entitled to assert everything the realist can assert; so, when a nonrealist goes sweeping minimalist, she will be entitled to assert everything the realist can assert. Yet it is strongly implausible that the nonrealist will actually become a realist. For it is absurd to believe that realism could be obtained by just implementing the vacuous principles of creeping minimalism. How can we make sense of this? We are now in a position to unravel Dreier’s puzzle. This is my solution: when creeping minimalism takes place, many forms of nonrealism can still be distinguished from realism because it is false that the sweeping minimalist is entitled to assert everything the realist can assert; consequently, it is false that, when a nonrealist goes sweeping minimalist, she is entitled to assert everything the realist can assert. The following examples clarify my point. Consider fi rst a realist—or Platonist—about numbers. The Platonist about numbers can be a realist about the entities that coincide with facts about numbers (e.g., the fact that 3 has the property of being odd) and properties of numbers (e.g., the property of being odd). Because the sentence “3 is odd” is warrantedly assertible, a sweeping minimalist can assert: “it is a fact that 3 has the property of being odd.” Does this make the sweeping minimalist indistinguishable from the Platonist? No. For the Platonist who is a realist about facts and properties of numbers is committed to claiming that such facts and properties are mindindependent in a nonplatitudinous sense of this term. Yet the sweeping minimalist as such is not entitled to this claim, for her conceptual framework does not include nonplatitudinous notions of mind-independence. A Platonist about numbers may, however, not want to be committed to the existence of entities such as facts and properties so that she may consider facts and properties to be mere reifications of linguistic constructs (i.e., minimal entities). Is the sweeping minimalist who asserts “it is a fact that 3 has the property of being odd” indistinguishable from a Platonist of this kind? No, because the Platonist about numbers is anyway committed to claiming that numbers are mind-independent in a nonplatitudinous sense of it, and the sweeping minimalist as such is not entitled to this claim. Consider now an antirealist about numbers who becomes sweeping minimalist. She will be entitled to assert sentences like: “it is a fact that 3 has the property of being odd.” Does this or similar entitlements turn the antirealist into a Platonist about numbers? No. For this and similar entitlements do not commit the antirealist to the thesis that numbers (or facts about numbers and properties of numbers) are mind-independent in a nonplatitudinous sense.
180
Luca Moretti
Consider fi nally an error-theorist about numbers. She will contend, basically, that all warrantedly assertible (or provable) sentences about numbers are false, for there are no such entities. Consequently, for her, “3 is odd” is false. As this error-theorist goes sweeping minimalist, she will be able to claim that the sentence “it is a fact that 3 has the property of being odd” is false too (and similar claims involving minimal notions will be allowed through different something-from-nothing inferences). But this does not seem to raise evident difficulties. These examples teach that creeping minimalism does not render nonrealism indistinguishable from realism. When creeping minimalism takes place, it is still possible to distinguish with clarity and precision between positions that entail realism about X-entities, positions that entail antirealism about X-entities, and positions that entail an error-theory about Xentities. This completes my answer to Dreier’s problem.
10 Hempel’s Dilemma
1
Daniel Stoljar
1
INTRODUCTION
In the course of a commentary on Nelson Goodman’s Ways of Worldmaking, Carl G. Hempel formulated a problem for physicalism that has come to be known as “Hempel’s dilemma.” He wrote: I would add that the physicalistic claim that the language of physics can serve as a unitary language of science is inherently obscure: the language of what physics is meant? Surely not that of, say, eighteenthcentury physics; for it contains terms like ‘caloric fluid’, whose use is governed by theoretical assumptions now thought false. Nor can the language of contemporary physics claim the role of unitary language, since it will no doubt undergo further changes too. The thesis of physicalism would seem to require a language in which a true theory of all physical phenomena can be formulated. But it is quite unclear what is to be understood here by a physical phenomenon, especially in the context of a doctrine that has taken a decidedly linguistic turn. (Hempel 1980, 194–195. See also Hempel 1969) Hempel here was contrasting Goodman’s position with that of Otto Neurath, the member of the Vienna Circle most responsible for formulating and promoting the thesis of physicalism (see Neurath 1931a, 1931b). For Neurath, or anyway for Neurath as interpreted by Hempel (see Hempel 1949), physicalism was the linguistic doctrine that every statement is equivalent in meaning to—that is, is synonymous with—some physical statement. And, on its face,
1 Material on which this chapter is based was presented at the Truth and Reality Conference at the University of Otago, New Zealand, the University of Birmingham (UK), Cambridge University (HPS), and the Philosophy Society, Australian National University. I am very grateful for the very good comments I received on the chapter from those audiences. I am also grateful to Andy Egan for written comments on a previous draft.
182
Daniel Stoljar
Hempel’s point is limited to a criticism of physicalists of Neurath’s type. But of course few if any physicalists are of that type now. Nowadays, physicalism is the metaphysical doctrine that, not necessarily but as a matter of fact, every instantiated property is necessitated by, or supervenes on, some physical instantiated property.2 So it might appear that, even if Hempel is right about the inherent obscurity of Neurath’s physicalism, this is not something with which a contemporary physicalist need be overly concerned. For most writers on this issue, however, the problem Hempel is posing is a problem not simply for a linguistic version of physicalism but for the metaphysical version as well. Geoffrey Hellman, for example, writing a few years after Hempel, describes it as “perhaps the most serious objection to all efforts in formulating physicalism” and formulates it in the following influential and often-quoted passage: [C]urrent physics is surely incomplete (even in its ontology) as well as inaccurate (in its laws). This poses a dilemma: either physicalist principles are based on current physics, in which case there is every reason to think they are false; or else they are not, in which case it is, at best, difficult to interpret them, since they are based on a “physics” that does not exist—yet we lack any general criterion of “physical object, property, or law” framed independently of physical theory. (Hellman 1985, 609) Moreover, in supposing that the problem is a problem for physicalists in general, Hellman has been followed by many others. Hempel’s dilemma has become, as Jeffrey Poland puts it, the “the stock objection” (1994, 157) to the idea that physicalism admits of a clear formulation. What should we say about the stock objection? My answer will be that Hempel’s dilemma is fallacious because its main disjunctive premise is false. So far as I can determine, to the extent that standard discussions of the dilemma focus attention on this premise at all, it seems to be assumed that it is an obvious or even a logical truth—in Hellman’s words, “either physicalist principles are based on current physics . . . or else they are not.” However, far from being an obvious or logical truth, the disjunctive premise is, at least when properly understood, a substantial falsehood. Once the premise is rejected, as I think it should be, we are faced with a number of difficult problems about the concept of the physical; but one problem we are not faced with is Hempel’s dilemma. The chapter is organized as follows. In §2 I set out what I take Hempel’s dilemma to be. In §§3–6 I develop and defend my objection. In §§7–8 I
2 Of course, there is a very large philosophical literature on what it means to say one fact or set of facts supervenes on or entails another, but I operate here with a rough and ready notion. That should be enough for my purposes.
Hempel’s Dilemma
183
compare this objection with two others. I end the chapter, in §9, by briefly examining the motivation that Hempel himself offered for formulating his dilemma, a motivation that lies in ideas about philosophical methodology associated with the so-called linguistic turn.
2
THE DILEMMA FORMULATED
We have so far talked informally about Hempel’s dilemma—what is it exactly? No doubt the precise formulation is a matter of debate, but in my view it is reasonable to work with the following version of the argument: H1. If physical properties are by defi nition the properties expressed by the predicates of a current physical theory, physicalism is false. H2. If physical properties are by defi nition the properties expressed by the predicates of an ideal physical theory, we don’t know what physicalism says. H3. Either it is the case that physical properties are by defi nition the properties expressed by the predicates of a current physical theory, or it is the case that physical properties are by defi nition the properties expressed by the predicates of an ideal physical theory. HC. Ergo, either physicalism is false, or we don’t know what it says. Understood this way, Hempel’s dilemma is a valid argument for a prima facie significant conclusion. What reasons are there for the premises? I will address this question by focusing on each premise in turn.
2.1
Support for H1
We may think of a physical theory as a theory that aims to provide a complete inventory of the properties and relations required in the explanation of ordinary physical objects and related phenomena. But surely current physics does not provide a complete inventory. That is, surely we are in the midst of an ongoing investigation, rather than being at the end or close to the end of the investigation. On the other hand, if current physics is indeed incomplete, presumably there is a possible future physical theory that is complete, or at least is more complete than current physics is. Let us now imagine such a theory and, in particular, imagine that a predicate of such a theory expresses a property— call it “F.” If physical properties are by definition those that are expressed by the predicates of current physical theory, F is not physical. This is not to say that F is spooky or ethereal nor that F conforms to any paradigm we might have of what a nonphysical property is; in fact, as we have no idea what F is, F conforms to no paradigm at all. It is just to say that if physical properties are by definition contemporary physical properties, F is not physical. Now, that F is not physical is not by itself a problem for physicalism. The beauty of Venice is a property of Venice, but it is not a physical property
184
Daniel Stoljar
of Venice. But that is no problem for physicalism, for physicalism is not the thesis that every property is a physical property. It is rather the thesis that every instantiated property is necessitated by some instantiated physical property—and this formulation permits the instantiation of some nonphysical properties. However, though F not being physical is not by itself a problem for physicalism, there is a related consideration that is. For not only is F not physical; it is also not necessitated by any physical property. For consider: F was introduced as a property not on the list of properties given by current physics. But “not being on the list” in the relevant sense, means, I take it, not only “is numerically distinct from” but also “is not necessitated by” any property on that list. But now it does indeed follow that physicalism is false: F is a property that is neither physical nor entailed by anything physical; physicalism entails that there is no such property. In sum, the reason for H1 is that current physics is incomplete. If current physics is incomplete, then, if physical properties are by defi nition those expressed by contemporary physics, physicalism is false.
2.2
Support for H2
Suppose that the Peircean limit of inquiry is reached at C.O.B. on November 5, 3027. Presumably the scientists leaving work on that fateful day—of course they may not know that the day in question is the fateful one—will have formulated some physical theory, and the theory in question will say at least that there are various properties, let us call them “F,” “G,” and “H,” which do various things. If physical properties are by defi nition the properties expressed by the predicates of this ideal physics, then the physical properties are F, G, and H. Similarly, if physical properties are by defi nition the properties expressed by the predicates of ideal physics, then physicalism tells us that F, G, and H entail every other (instantiated) property, either singly or in combination. But what then does physicalism say? For those of us who are not at the Peircian limit the answer seems to be that we have no idea. Physicalism says that every property supervenes on some complex of F, G, and H. But because we don’t know what F, G, and H are, we don’t know what physicalism says. Of course we can in a sense name these properties and refer to them; I have just done so at least schematically. But naming is different from knowing. Even if we can name the properties, we still don’t know what they are, nor do we know what physicalism says. In effect, we are like the people in Gareth Evans’ example who can refer to Julius as “the actual inventor of the zip” but do not know who Julius is (Evans 1982). Suppose then we don’t know what physicalism says—what follows? Well, if we do not know what it says, we are in no position to know it, no position to deny it, no position to believe or disbelieve it with justification. Nor are we even in a position to speculate about whether it is true. In fact the whole project of rationally assessing physicalism—providing reasons
Hempel’s Dilemma
185
for and against it, declaring oneself for it or against it, saying it is a good bet at least in the long run—seems to presuppose that we know what it is, at least in outline. But if that presupposition is false, physicalism is unworthy of assessment. In sum, the reason for H2 is that, in view of the incompleteness of current physics, we don’t know what an ideal physics says. If we don’t know what ideal physics says, then, if physical properties are by defi nition the properties expressed by the predicates of ideal physics, we don’t know what physicalism says.
2.3
Support for H3
H1 says that if physical properties are defined one way, something bad happens, and H2 says that if physical properties are defined some other way, something different but also bad happens. H3 tells us that these methods of definition exhaust the field: either physical properties are defined in terms of a theory that is current or they are defined in terms of a theory that is ideal. Why believe H3? Later on I discuss Hempel’s motivation for H3. If H3 is false, as I think it is, this motivation must be less persuasive than Hempel, at least, thought it was, and I think that an analysis of it bears this out. For the moment, however, I will content myself with making two points about the case for H3. The first is that, although it is true that we may extract from Hempel’s discussion a consideration in favor of H3, in point of fact most contributions to this debate fail outright to provide such considerations and pass over this part of the argument without comment. As I noted before, many people seem to think H3 is just obvious.3
3 Evidence for this claim derives in part from the fact that the standard responses to the dilemma assume that H3 is true. For example: one group of philosophers, such as Mellor (1973; see also Crane and Mellor 1990) and Daly (1998), takes the dilemma to establish what it purports to, viz., that physicalism has no true formulation that is worthy of assessment and draws the moral that discussions of physicalism in philosophy of mind that concern the doctrine are without point or are incoherent. Another group, such as Melnyk (1997, 2003) and Hellman (1985) himself, agrees that the dilemma establishes what it purports to but draws a different moral, namely that physicalism can be philosophically significant even if false. A third group, such as Smart (1978) and Ravenscroft (1997), denies that the dilemma establishes its conclusion, arguing either that contemporary or near contemporary physics is in fact true or at any rate true as far as philosophy of mind goes. And a fourth group, such as Poland (1994), J. Wilson (2006), and Dowell (2006), agrees that the dilemma does not establish its conclusion but argues instead that a version of physicalism based on some ideal or future physics may, contra the dilemma, be worthy of assessment after all. These proposals all raise interesting issues on their own terms but what they all share is the tendency to agree with H3.
186
Daniel Stoljar
Why do people think that it is just obvious? I am not sure. However, as I mentioned at the outset, it may be that they confuse it with a logical truth. One possibility is that they confuse H3 with the truth that physics is either current or it is not. But this can hardly be the premise that is in operation in Hempel’s dilemma, if only because it says nothing about how physical properties are to be defi ned. A more likely possibility is that H3 is being confused with H3*: H3* Either it is the case that physical properties are by defi nition the properties expressed by the predicates of physical theory that is current, or it is not the case that physical properties are by defi nition the properties expressed by the predicates of physical theory that is current. But again it is implausible that this is the main premise of Hempel’s dilemma. If this were the disjunctive premise of Hempel’s dilemma, the second premise would have to be, correlatively, this: H2* If it is not the case that physical properties are by defi nition the properties expressed by the predicates of a current physical theory, we don’t know what physicalism says. But we have no reason for believing H2*! In particular, the reason we had for believing H2 is not likewise a reason for believing H2*. The reason for believing H2 was that we did not know what ideal physics says. But we know perfectly well what various noncurrent physical theories say, and so we know perfectly well what various versions of physicalism say if they are defi ned with reference to those theories. (We know perfectly well what physicalism says if it is defi ned with reference to the physics of the nineteenth century, for example.) What this suggests is that in order to formulate Hempel’s dilemma we need to operate with a contrast between the physical theory that we currently adopt, on the one hand, and the physical theory that will be true in the ideal limit, on the other. But then it would seem that in order to formulate Hempel’s dilemma, we need H3 and not H3*. The possibilities about why people think H3 is obvious that I have just mentioned are speculative; as I said, I don’t know why people think this. But in any case—and this is the second of the two points I said I want to make about H3—they are wrong if they do. H3 is a substantial, and so not an obvious, claim. In particular, H3 might be false in either of two ways. First, it could be that physical properties are not defi ned by a physical theory at all. Second, it could be that, even if physical properties were defi ned by a physical theory, they would not be defi ned by one that is either current or ideal.
Hempel’s Dilemma 3
187
A COUNTEREXAMPLE TO H3
So far I have set out Hempel’s dilemma and said something about the plausibility of its premises. Turning now to criticism, my own focus—as I have indicated—will be on the disjunctive premise of the argument, viz., H3 Either it is the case that physical properties are by definition the properties expressed by the predicates of physical theory that is current, or it is the case that physical properties are by defi nition the properties expressed by the predicates of a physical theory that is ideal. But why is H3 false? Well, let us focus fi rst on a very simple version of a view about physical properties—call it “the theory view”—that is suggested by the fi rst disjunct of this premise: (1) F is a physical property if and only if F is a property expressed by a predicate of a current physical theory. It is very easy to fi nd counterexamples to (1). Thus, consider impetus, the property attributed to physical objects by medieval impetus physics. Impetus, or having impetus, is not a property expressed by a predicate of a current physical theory. Yet having impetus is a physical property. Hence (1) is false.4 Is this just the fi rst premise of Hempel’s dilemma all over again? Not so, for imagine replacing (1) with (2):
4
The notion of impetus was expressed by the medieval philosopher and scientist Jean Buridan. He wrote: “A mover, while moving a body, impresses on it a certain impetus, a certain power capable of moving this body in the direction in which the mover set it going, whether upwards, downwards, sideways or in a circle. By the same amount that the mover moves the same body swiftly, by that amount is the impetus that is impressed on it powerful. It is by this impetus that the stone is moved after the thrower ceases to move it; but because of the resistance of the air and the gravity of the stone, which inclines it to move in a direction opposite to that towards which the impetus tends to move it, this impetus is continually weakened. Therefore the movement of the stone will become continually slower, and at length, the impetus is so diminished or destroyed that the gravity of the stone prevails over it and moves the stone down towards its natural place” (qtd. in Crombie 1963, 67). The concept of impetus has also been of interest to developmental psychologists, some of whom suggest that common sense physics conforms to the principles of impetus physics (see, e.g., McCloskey 1983). Construed as a counterexample to proposals about how to defi ne physical properties, impetus is mentioned but not developed in my ‘Physicalism’ (2001).
188
Daniel Stoljar
(2) F is a physical property if and only if F is expressed by a predicate of an (the?) ideal physical theory. On the assumption—which I will adopt—that the ideal physical theory is a true physical theory, the example of impetus refutes (2) as a defi nition of what it is to be a physical property just as surely as it refutes (1). For, though the true physical theory will tell us which physical properties are instantiated, and what they do, it will not tell us about impetus for the simple reason that impetus is not instantiated. Hence impetus is not a physical property by this defi nition. It might be objected that (as we indicated before) we have no idea what ideal physics says. If so we are in no position to say that it will not mention impetus—maybe impetus is mentioned in ideal physics after all! But this objection doesn’t get to the heart of the matter, as so far developed. The heart of the matter as so far developed is that any true physical theory will tell us only about the physical properties that happen to be instantiated in this world. But it is very plausible that there are physical properties that are not instantiated in this world; impetus is my example of such a property, but any property like this will do. On the other hand, if there are physical properties not instantiated in the actual world, then (2) is false. Another way to put the point is to say that presumably the ideal theory will rule out any false theory. But properties postulated by false physical theories are nevertheless physical. So an ideal physical theory will not tell us what it is for a property to be physical. The impetus example tells us that (1) and (2) are both false. But it also tells us something very important about Hempel’s dilemma, viz., that its third premise, H3, is false. For consider: H3 tells us that physical properties are by defi nition those expressed by the predicates of current physics or ideal physics; in effect, H3 is just the disjunction of (1) and (2). But we have already seen that both (1) and (2) are false; Hence H3 is false too. So impetus tells us not only that (1) and (2) are false. It tells us also that H3 is false and that Hempel’s dilemma as stated collapses.
4
INDIVIDUAL THEORIES VERSUS KINDS OF THEORIES
I have argued by counterexample that H3 is false and, in consequence, that Hempel’s dilemma collapses. But arguments by counterexample only go so far, and you may already have thought of ways to respond. I come back to some of these ways in a moment, but fi rst I want to point out that, when we reflect on why the impetus case is persuasive, it emerges that our dispute with H3 concerns an issue of principle and not simply an issue of how to respond to one example. The persuasiveness of the impetus example is due, I think, to an unclarity in the theory view I have so far been ignoring. The unclarity has to
Hempel’s Dilemma
189
do with what is meant by “physical theory.” On the one hand, when one speaks of “physical theory” one might have in mind a particular physical theory, such as medieval impetus physics, or Einstein’s theory of gravity, or that physical theory, whatever it is, that scientists will formulate in the Peircian limit. On the other hand, when one speaks of “physical theory” one might have in mind a kind of theory of which each of these particular theories are instances. So, in particular, one might think of a “physical theory” as a theory of a certain general kind, say distinguished by its subject matter. One might distinguish here between theory-types and theory-tokens, but the phrase “theory-token” I fi nd grating, so I will speak instead about individual theories versus kinds of theories. Now, what interpretation of the phrase “physical theory” is at issue in Hempel’s dilemma? It seems clear that what is intended is the idea of a particular physical theory. Hempel asks “the language of what physics is meant?,” and he clearly means to be asking: the language of what particular physics is meant? Similarly, H3 says that physical properties must be defi ned either in terms of one particular physical theory—viz., current physics—or another particular physical theory—viz., ideal physics. In short, Hempel’s dilemma tacitly assumes that by “physical theory” one means “particular physical theory,” and so far we have taken over this assumption uncritically. However, on reflection this assumption is false, and it is impossible that the notion of a physical property might be defi ned in terms of any particular physical theory. For suppose you pick some particular theory, T, and say, “the physical properties are by defi nition the properties expressed by T.” It should be clear that there will always be some other particular theory T*, and if that other theory is similarly a physical theory, then the properties that are expressed by this other theory will also be physical. Hence one’s purported defi nition will fail. It is this fact, I think, that explains why the impetus example works. Both current physics and ideal physics are two particular physical theories; impetus physics is another. If you defi ne the physical properties by reference to current physics, then the existence of impetus physics will refute you. More generally, if you defi ne physical properties by reference to any particular physical theory, the existence of other particular physical theories will refute you. Hence, no version of the theory view that is founded on particular theories will be true. How might this conclusion be avoided? If the argument will be replicated no matter what particular physical theory you pick, the only response that is plausible at this point is to abandon particular physical theories for kinds of theory. But this means in turn that the only version of the theory view that has a chance of being true is the one that defi nes physical properties in terms of a theory of a certain kind, that is, the physical kind. But in turn, if this is right, we arrive again at the conclusion that H3 is false. The problem with H3 is that it presupposes a version of the
190 Daniel Stoljar theory view that tries to defi ne physical properties in terms of particular physical theories. As that presupposition is false so too is H3; and if H3 is false, Hempel’s dilemma collapses.
5
A “KIND OF THEORY” VERSION OF THE THEORY VIEW?
I have objected that any version of the theory view that defines physical properties in terms of a particular physical theory is false and that, in consequence, Hempel’s dilemma, which presupposes such a theory, is unsound. How might a proponent of the argument respond to this objection? One response starts from the point that I have not so far argued against any version of the theory view. In particular, I have left open the idea that a version of the theory view defi ned in terms of a kind of theory might be true. Might such a theory be exploited to resurrect Hempel’s dilemma? The answer to this question is “No.” The “kind of theory” version of the theory view is, I take it, something like (3): (3) F is a physical property if and only if F is expressed by a predicate of a theory of a certain kind, viz., the physical kind. An attractive feature of (3) is that it immediately avoids the impetus problem. Having impetus might have failed to be expressed by a predicate of a particular theory, but it could not have failed to be expressed by a predicate of a theory of a certain kind, viz., a physical theory. So impetus counts as physical by the lights of (3). But although (3) avoids the impetus problem, it is no help to the wouldbe defender of Hempel’s dilemma. Suppose a property is physical just in case it is expressed by a predicate of a theory of the physical kind. It does not follow that a theory of that kind must be either current or ideal. An ideal theory means, in this context, the physical theory that turns out to be true in the actual world. But this is a particular theory, not a kind of theory. So we are no closer to a defense of H3 than we were before. Of course, one might point out that, in addition to being no help to the would-be defender of Hempel’s dilemma, (3) faces a further problem. What is it for a theory to be of the physical kind? Take all the actual and possible theories that are physical; what is it that they have in common in virtue of which they are physical? These are obviously exceedingly difficult questions—it is certainly not obvious what kind of theory a physical theory is. But even though this is a problem, it is a problem that I want to set aside here. Whatever issues we confront when we confront the issue of explaining what kind of theory a physical theory is, Hempel’s dilemma will not be among them. For Hempel’s problem starts from the assumption that you must employ one of two particular theories to defi ne what it is to be a
Hempel’s Dilemma
191
physical property. My objection is precisely that this assumption is clearly no good.
6 A MORE SOPHISTICATED “PARTICULAR THEORY” VERSION OF THE THEORY VIEW? The fi rst response to our objection to H3 is to appeal to a “kind of theory” version of the theory view. The second response points out that I have so far considered only very simplified versions of the “particular theory” version of the theory view. Might not a more sophisticated version do better? This is obviously a very natural thought. Its only problem is that when we examine what some of these more sophisticated versions of the theory view are, it becomes apparent that the basic problem remains untouched. The first alternative involves adjusting the theory view so that the primary idea concerns not whether the theories in question are true and not whether they are physical but whether they have been formulated at some point in time: (4) F is a physical property if and only if F is expressed by a predicate of a formulated physical theory (that is, a theory formulated at some point in the present, past or future). This proposal does not face the impetus problem, for there is a formulated physical theory that contains a predicate that expresses impetus, viz., impetus physics. But a variant on our main example can be constructed to get around this. For take some property that we agree to be physical, say having a certain mass or impetus for that matter. Suppose mass were not expressed by a predicate of a formulated physical theory. Would it still not have been physical? For example, suppose that humans lost out in their competition with Neanderthals, and Neanderthals never developed sciences of their own. Then mass would not have been expressed by a predicate of a formulated physical theory. But surely having mass or impetus would nevertheless have been physical properties. A scientist might say: “if the physical constants of the universe were different, life would not have evolved.” For a philosopher to retort “more different than you realize—for the constants would not even have been physical” is just silly, but that is what (4) commits one to. The second alternative is to adjust the theory view so that it involves not simply what predicates are in fact contained in current physics but what predicates might be defi ned in terms of those that are contained in current physics: (5) F is a physical property if and only if: either F is expressed by a predicate of current physical theory, or F is expressed by a predicate defi nable in terms of the predicates of current physical theory.
192 Daniel Stoljar Like proposal (4), this avoids (or seems to) the impetus problem. The predicate “has impetus” is not a predicate of current physical theory, but perhaps it is defi nable in terms of those. But nevertheless (5) does not get around the principled objection that the impetus problem brought to light. Unless every physical property is expressed by a predicate definable in terms of current predicates, there will always be a property like impetus. But it is implausible that every physical property is so expressed—this would entail, for example, that current physics is a universal theory in the sense that this theory has the expressive resources of any physical theory. One might respond that it is implausible that current physics is a universal theory in this sense, but it is not implausible that ideal physics is such a theory; more generally, it is not implausible that there is a particular theory T which is such that T has the expressive power to capture every physical theory. However, there are two problems with this proposal. In the fi rst place, it seems extremely speculative to suppose that there is a physical theory like this—in particular, from the fact that a theory is ideal, i.e., is one we would adopt in the ideal limit, it does not follow that the theory is universal in this sense. 5 In the second place, to argue in this way is in effect to reintroduce the idea of a kind of theory. A physical theory is a kind of theory that is translatable into T. And, as we have seen, if there is such a theory, we no longer have any reason to believe H3. The final alternative—I mention it mainly because it itself is so often mentioned in the literature—is to appeal to resemblance: (6) F is a physical property if and only if: either F is expressed by a predicate of current physical theory or F is sufficiently similar to a property expressed by a predicate of current physical theory. The problem with this view is that it succeeds only by being coy about what the dimensions of similarity are. For example, is the property of having impetus sufficiently similar to the properties expressed by current physics to count or not? If not, then impetus will refute (6) just as it did (1) and (2). If so, then there is some dimension of similarity that ties current physical properties together with impetus. But what could that dimension of similarity be?
7
FALLIBILISM
Our objection to Hempel’s dilemma is that its third premise—H3—presupposes a particular version of the theory view, and no such theory could
5 In fact, I think it can be shown that there is no physical theory like this, owing to the impact of twin-earth examples on the topic of what it is to be a physical theory. But this point will have to be left to another occasion. See my Physicalism (forthcoming) and Sussman (1981).
Hempel’s Dilemma
193
possibly be true. But the interest of this suggestion would be limited if there were any number of other reasons to resist the dilemma. Are there such reasons? More generally, how does our objection compare with others in the literature? One objection is that Hempel’s dilemma is unsound because it fails to appreciate the sense in which the contemporary physicalist is a fallibilist in epistemology. The fallibilist in epistemology holds that, at least for any empirical belief, the evidence that one has for the belief does not entail that the belief in question is true. Fallibilists emphasize that at least in empirical matters, there is no certainty and no ruling out all possible alternatives. And the contemporary physicalist certainly is by nature a fallibilist. Indeed, the very idea that physicalism is a kind of empirical, but abstract, hypothesis is suggestive of fallibilism. However, while physicalists are fallibilists, what is the connection between fallibilism and Hempel’s dilemma? Well, as we have seen, the reasoning in support of both H1 and H2 presupposes that contemporary physics is incomplete. In the case of H1, the incompleteness of current physics supports the idea that if physicalism is defi ned in terms of it, then physicalism is false. In the case of H2, it supports the idea that we don’t know what an ideal physics will say. But perhaps a physicalist could insist that, in their view, physics is complete—it is simply that, compatibly with what we know, it might not be complete. In sum, fallibilism permits one to respond to Hempel’s dilemma by denying the incompleteness of physics. “It is not true that physics is incomplete,” you might say, “all that is true is that it might be; that is, it is consistent with our evidence that it is.” Without further development, this response to Hempel’s dilemma is unsatisfactory. It is true that both the completeness and the incompleteness of current physics are logically compatible with the truth of what physics tells us about the world now, and it is certainly true that physicalists are fallibilists and reasonably so. However, this does not mean that it is plausible to believe that current science is complete. On the contrary, there seem a rather large number of reasons for supposing the opposite: • There may be cosmological reasons: We may think of ourselves (or humans in general) as a occupying a very small pocket of the universe; it is amazing to think that cognitive mastery could be achieved from such a vantage point. • There may be evolutionary reasons: We may think of ourselves (or humans in general) as evolved creatures whose cognitive capacities have been shaped by millions of years of evolution; surely in that case it is implausible that minds like ours could form a complete picture. • There may be psychological reasons: We may think of our mind (or the human mind in general) as being subject to various kinds of barriers, constraints, or fi lters; surely it is implausible that minds of our kinds are likely to know all sorts of facts.
194
Daniel Stoljar
• There may be historical reasons: We may think that the science of our own day is like that of previous epochs in that it contains mistakes, confusions, wrong turns, approximations; the natural inference is that present circumstances are no different from previous circumstances in these respects, even if present science is better than, closer to the truth than, other sciences. • There may be methodological reasons: We may think that the practice of science or rational inquiry presupposes that we don’t know various types of facts; the very point of scientific inquiry seems to presuppose ignorance. In summary, although it is true that both the completeness and incompleteness of current physics are logically compatible with what science tells us about the world, it is not hard to motivate the hypothesis that current physics is indeed incomplete. So appealing to fallibilism will not help the physicalist respond to Hempel’s dilemma. Of course, one might respond that, even though these considerations are true enough in their way, they nevertheless do not actually entail that current knowledge is incomplete; hence it remains open for an optimist to insist in the face of them that current physics is complete. But at this point the issue seems to degenerate into a stalemate between the optimist and the pessimist about current physics (and current science or knowledge more generally). The optimist and the pessimist may agree about what is currently known and about all of the claims just reviewed. Yet the optimist may insist that current knowledge is complete (or very nearly so) and the pessimist may insist it is not. How do we decide this issue? So far as I can see there is no way (or at any rate, no practical way) to resolve it; on the other hand, it seems clearly a dispute about a matter of fact. If correct, that is an interesting observation that deserves further discussion. But for the moment, the important thing is the connection between the unresolvability of this dispute and the fallibilist response to Hempel’s dilemma. In particular, it would appear that fallibilism is not going to provide a successful response to Hempel’s dilemma, for the success of that response turns on an unresolvable dispute between the optimist and pessimist about current physics.
8
KNOWING ENOUGH TO KNOW
The fallibilist objection says that Hempel’s dilemma fails because, though it is possible that we have incomplete knowledge, in fact we have complete knowledge. The other objection that I want to consider takes a different tack. We have seen that the key observation in the dilemma is the incompleteness of current knowledge and that it is at least dialectically ineffective to deny this. Nevertheless, it might be objected that even if we acknowledge
Hempel’s Dilemma
195
that current physical theory is incomplete, the argument may be resisted. “Sure, current physics is incomplete, and in consequence we don’t know exactly what the true physics will say,” the proponent of this objection will say, “but nevertheless we know enough to know that the true physics will not contain sui generis psychological phenomena, and this is enough for the purposes of philosophy of mind.” In sum, the main thrust of this objection is to accept Hempel’s dilemma but to defang it, arguing that the conclusion does not have the significance that many have supposed. What can we say about this objection? To begin with, the objection does contain an important insight. It is common to suppose that Hempel’s dilemma is significant because, if it were sound, many famous discussions in philosophy of mind would lose their rationale. For example, take the knowledge argument in philosophy of mind. This is an argument to the conclusion that physicalism is false. And people who debate the argument often describe themselves as “physicalists” (if they don’t like the argument) or “antiphysicalists” (if they do). On the other hand, if Hempel’s dilemma is sound, the whole rationale of this discussion of the knowledge argument seems to have been removed. If Hempel’s dilemma is sound, we know, and on grounds that have nothing to do with the knowledge argument, that physicalism is either false or unworthy of assessment. What then would be the point of discussing whether some other argument establishes that it is false? However—and this is the insight embodied in the response to Hempel’s dilemma that we are now discussing—this reason for finding the dilemma significant is mistaken. It is true that discussions surrounding the knowledge argument go on in the name of physicalism, but it is much less clear that this is essential to the enterprise. For the knowledge argument would if successful tell us not only that phenomenal facts—such as the fact that I have a pain in my toe—are not entailed by physical facts. It would tell us that this fact is not entailed by any fact that is not itself a phenomenal fact; that is, it would tell us that phenomenal facts are not entailed by any other facts. But if that is right, physicalism construed as a thesis about the world drops out of the picture. It doesn’t matter for the purposes of the knowledge argument whether the contrasting facts are physical or not; what matters is that they are not phenomenal. If this is so, however, physicalism per se is inessential to the argument. And that means that, even if Hempel’s dilemma were sound, it would have a very limited impact on arguments such as the knowledge argument. So there is an element of truth in this “know enough to know” objection to Hempel’s dilemma. Nevertheless it remains in my view unsuccessful as a response. For the objection assumes that Hempel’s dilemma is significant, if it is, for one reason only, viz., because of its impact or alleged impact on arguments in philosophy of mind such as the knowledge argument. However, although this is certainly one way in which Hempel’s dilemma is significant, it is not the only way. In fact, the dilemma is incredible on its face (which is not to say that it might not be sound). Philosophers from Hobbes to Smart have endorsed physicalism; and philosophers from Descartes to C. D. Broad have
196
Daniel Stoljar
denied it. Surely it is incredible that these philosophers are as confused as they would have to be if the conclusion of Hempel’s dilemma were true. In sum, the reason for thinking that Hempel’s dilemma is significant (or at any rate of considerable interest), is this: it is an apparently obvious fact from the history of philosophy that physicalism is a substantive doctrine; hence it is not the case that it is either false, or we don’t know what it says.
9
CONCLUSION
My assumption has been that Hempel’s dilemma is an argument from premises of the form “A ∨ B,” “If A then C,” and “If A then D” to a conclusion of the form “C ∨ D.” Focusing on this structure in turn focuses attention on the main disjunctive premise of the argument, and what I have been suggesting is that when we do focus on this premise it emerges that it is false and so that Hempel’s dilemma is unsound. But it is important to be aware of the limits of my discussion. I have said nothing about the truth (or falsity) of the conclusion—only that Hempel’s dilemma is not a good argument for that conclusion. And I have said nothing about what the correct view of physical properties or facts is—only that one can conclusively dispose of Hempel’s dilemma without doing so. Rather than taking up these larger questions here, in this fi nal section I turn to Hempel’s own motivation for advancing the dilemma. A very common methodological move in twentieth-century philosophy is the attempt to restate metaphysical issues as linguistic ones: to move, as Quine famously put it, from talk of miles to talk of “miles.” In one way of looking at it, the theory view is precisely an instance of this strategy—for it trades in the question of what a physical property is for the question of what a physical theory or language is. To the extent that we think of this strategy—Quine called it “semantic ascent”—as the method of clarification in philosophy, we will be friendly to the theory view. It is this motivation that is most prominent in Hempel’s own discussion of the dilemma. In the last sentence of the fi rst quotation in §1, he says: “it is quite unclear what is to be understood here by a physical phenomenon, especially in the context of a doctrine that has taken a decidedly linguistic turn.” By “linguistic turn,” Hempel means, I think, exactly what Quine means by “semantic ascent.” In fact, the role of semantic ascent in Hempel’s dilemma seems to me to be strangely unnoticed in the literature. It is often pointed out that the remark on Goodman continues a theme developed in a much longer paper, published ten years earlier; indeed the 1969 paper is often cited together with the comment on Goodman as the source of Hempel’s dilemma as discussed here. But it is almost never pointed out that, in that earlier paper, Hempel’s (1969, 180) stated goal is to “reflect on the rationale of this linguistic turn and to explore some of its implications for current problems in the theory of reduction”—the phrase “the
Hempel’s Dilemma
197
linguistic turn” here again is used by Hempel for Quine’s semantic ascent. In fact, what he primarily refers to as a dilemma in that 1969 paper is not Hempel’s dilemma as discussed here but a larger dilemma that contains Hempel’s dilemma as a proper part. The larger dilemma goes roughly as follows: either we adopt the method of semantic ascent, and so adopt the theory view, or we do not. If we do adopt it then we face what I have called here (following tradition) Hempel’s dilemma. If we don’t adopt it, then, Hempel seems to suggest, we have no way to clarify metaphysical disputes about physicalism. It is this second claim that provides a motivation for H3. If the theory view is nothing less than the linguistic turn applied to the question of the physical, and if the linguistic turn is the method of philosophical clarification, then the thesis of physicalism will remain unclear unless and until we can bring to bear the linguistic techniques upon it. In my view, it is doubtful that this larger dilemma is successful. For one thing, why view semantic ascent as the method of clarification in philosophy, as opposed to one of a whole barrage of methods of clarification and explanation? For another, Hempel seems to run together the idea that one might clarify a notion by switching focus to the language we use to express it with the quite different idea that we might explain a notion in terms of the language we use to express it. Nevertheless, while it might be that the larger dilemma Hempel offers us fails, there is in my view something right in the thrust of his thinking here, something that connects with our earlier discussion. For suppose you attempt to explain what a physical property is in terms of a physical theory or, following Hempel, a physical language. The language in question must be either a particular language or a kind of language. Suppose that what you have in mind is a particular language. Then you will run into the problem I outlined earlier, namely that whatever particular physical language you pick, the existence of other particular physical languages will refute you. So suppose now that what you have in mind is a kind of language—a language of the physical kind. As we saw before, the notion of a language (or theory) of the physical kind is somewhat obscure. But it seems reasonable to say that, whatever precisely it is, a language will be of the physical kind just in case it can be used to talk about a particular subject matter, that is, the physical. (Similarly, a physical theory is a theory of a particular subject matter, i.e., the physical.) But now it is plain that, just as you can’t explain what a physical property is in terms of a particular language, you can’t do so in terms of a kind of language either. For consider: if a language is one of the physical kind because it has a certain subject matter, you would need to identify that subject matter before you identify the kind of language in question. But then one could not explain the subject matter in terms of the language in question; if anything, it is the other way around. The problem just outlined is close to both the versions of Hempel’s dilemma we have discussed in the chapter, but it is not quite either of them. Its resolution will have to be left for another occasion.
11 From Reality to Truth The Case of Electrons and Electron Theory Robert Nola
1
INTRODUCTION
Scientific realism is the ontological (or metaphysical) thesis that most of the (unobservable) entities (including objects, events, processes, properties, relations, etc.) of physical science exist mind-independently, that is, independently of our beliefs about them, our theories of them, the terms we use to refer to them, the language we use to discourse about them, our perception of them, etc. This is distinct from semantic theses often associated with scientific realism; these appeal to semantic items and say that most of the (physical) theories of science that we have proposed are true, or have a high degree of verisimilitude, or that many of the terms used in these theories do correctly denote. Both of these theses are distinct from a third epistemological thesis about whether or not physical theories can be items of knowledge (as distinct from mere belief), what truth-value we know the theories to have, what rational degree of belief we can have for the theories on the basis of evidence, whether or not we have made genuine discoveries of hitherto unknown entities, and the like.1 Finally, realism can be an axiological thesis about what we aim for in science, viz., truth or increasing verisimilitude, in contrast with other nonrealist aims such as empirical adequacy. Just how well any of these theses apply to putative scientific entities and theories of these entities can be a controversial matter. But when philosophers wish to reach for an example of an entity to illustrate ontological realism about the unobservables of science, more often than not they appeal to electrons. What could be more real in science than the mind-independent existence of electrons? They are an important scientific discovery about a really existing kind of thing in the world about which humanity had no prior inkling before the second half of the nineteenth
1 The different theses of realism are discussed more fully in Hellman (1983) and Part II of Devitt (1997).
From Reality to Truth
199
century. In contrast, whether or not our various electron theories are true, or have at least increasing verisimilitude over time, is a much less clear matter. And if this is unclear, some doubt might be cast on whether we are right to be strongly committed scientific realists about electrons. And this doubt is reinforced when it is argued that there have been phases of electron theory that differ so greatly from earlier theories (compare the classical with the various quantum theory phases) that it becomes unclear whether deeply different theories are about the very same entity or about different entities. What is argued here is that the existence of electrons is far better known than any theory of the electron. This was in fact so for all theories of the electron up to the fi rst few decades or so of the twentieth century. And it can be argued that this is still the case; even though we now know a lot more about electrons, electron theory is still under active “construction” (for example, the nature of spin is not well understood). To complicate matters there were rival particle and wave theories of the electron long before electrons were even called by the name “electron.” Even in a situation such as this, the realist claim that electrons exist can be correct while theories about the electron are largely false. Such a possibility is consonant with a distinction drawn by Hacking (1983, ch. 1, especially 27) between two kinds realism. The fi rst is an ontological thesis of realism about entities which says that the (unobservable) entities of science exist mind-independently (as opposed to entity antirealism which says that they are fictions, or logical or social constructs, or whose supposition is instrumentally efficacious, etc.). The second is a semantic thesis of realism about theories which says that theories have a truth-value independently of whether we know it or not (as opposed to an antirealism which says that theories lack a truth-value, but they may still be useful to work on, or be instrumentally efficacious, etc.). Hacking also proposes a sufficient (but not necessary) condition for an entity to be real: “if you can spray them they are real” (Hacking 1983, 22–23). As will be shown, the early discovery of electrons was through their being “sprayed” and then by tracking both their effects and what happens when they are manipulated. In fact the very experimental condition under which they could be sprayed and then tracked and manipulated was for many years the only way in which they could be identified and reidentified across different experimental situations. This is one good approach to establishing ontological realism about electrons. But there are alternative approaches, one of which uses inference to the best explanation to show that if a theory makes successful predictions, then the truth of the theory is the best explanation of that success. From the conclusion that the best explainer is true, it can then be inferred that the theory’s existence claims are correct. But the path to entity realism, viz., that electrons exist, goes via the path of realism about theories, viz., that the theories are (approximately) true. However
200 Robert Nola this strategy for establishing entity realism may not always be available (and this is so independently of whether one does or does not have qualms about the use of inference to the best explanation). In this strategy, if theory realism cannot be established then entity realism cannot be established. This turns out to be the case as the early theories of the electron reveal. Because none of the theories of the electron were true, then inference to the best explanation would not have been able to assist in establishing the conclusion of the entity realist, viz., that electrons exist. However, the existence presupposition of the otherwise largely untrue theories, viz., that there are electrons, was well known to be true. As I argue, it was known to be true along the lines suggested by Hacking’s “spraying” test which does not depend on the truth (or approximate truth) of any theory about the items being sprayed. The position of a constructive empiricist is not entirely clear on the status of electrons and electron theory. Constructive empiricists invite us to consider the empirical adequacy of theories rather than their truth (at the level of nonobservational claims). However, none of the early theories of the electron met the requirement of being empirically adequate; in fact many of them were clearly empirically inadequate for the very people who used them. Realists and constructivists differ over the axiology of science; the aim of science is not truth of theories about the unobservable but truth about all that can be observed. In contrast, there is some agreement over other kinds of realism: “With the realist I take it that a theory is the sort of thing that can be true or false, that can describe reality correctly and incorrectly and that we may believe or disbelieve” (van Fraassen 1989, 192). It evidently follows from this that one of the existential presuppositions of electron theory, viz., that electrons exist, does have a truth-value and that such an existential claim does or does not correctly describe reality. (Here we assume that the reality to which van Fraassen refers is not just observable reality but also unobservable reality.) But what may be left undecided is our knowledge of these truth-values, or knowledge of whether reality is correctly described. In the latter case what may be at stake is whether our commonsense metaphysics of substance, individuation, causality, etc. that limn our modes of representation also apply to the unobservable world that they purport to represent. 2 However entity realists can go some way to
2 For a fruitful exchange on issues to do with realism and constructive empiricism, see McMullin (2003) entitled “van Fraassen’s Unappreciated Realism.” He draws a useful distinction between O-theories (which are about van Fraassen observables that can be understood to include neutron stars and dinosaurs) and U-theories (which are about van Fraassen unobservables). Realists and constructive empiricists can agree about O-theories and issues of realism with respect to them; however they may share, from their different perspectives, the same qualms about the strictly unobservable items in the ontology of U-theories. See the reply to McMullin in van Fraassen (2003),
From Reality to Truth
201
bridging this gap and step in where constructive empiricists fear to tread, claiming that we do know the truth-value of the proposition that electrons exist; it has the value true and not false. §2 begins by considering the way in which scientists fi rst obtained knowledge (by description and not by acquaintance) of the existence of a kind of “something” which only much later was called “cathode rays” and even later “electrons.” The Ramsey–Lewis account of denotation fi xing that serves as the philosophical background to the history is discussed in §3, along with the presuppositions made about what the description denotes. Once cathode rays could be identifi ed across different experimental set-ups, different theoretical models of what was identified were proposed that strongly rivalled one another. The deep ontological rivalry of the various ether disturbance and particle models is discussed in §4. The discussion is continued in §5 which outlines Thomson’s “startling hypothesis”3 which gives rise to his “corpuscle” model. As the hypotheses animating these models, including Thomson’s, are false of what they purport to model, a modifi cation is proposed to the standard way of using the Ramsey–Lewis approach to fi x denotation. This turns on the experimental results drawn together by Thomson, not his theoretical model; these results provided new ways of identifying cathode rays outside their parochial context of cathode-ray tubes and led to the view that cathode rays were ubiquitous entities in nature. §6 moves from the nineteenth-century prehistory of the electron to the twentieth century with its relativistic and quantum theories of the electron.4 But even with these radical changes in theory, it is argued that there is still ontological continuity despite the recognition that electrons might not fit very well into our various traditional ontological categories. The case presented shows how, for much of the history of particle physics, we can be more assured of the claim of entity realism that electrons exist than we can be assured about the truth (or even approximate truth in many cases) of the various theories of electrons.
in which one of the concerns about U-theories is their mode of representation and the nature of the unobservable reality they purport to represent. 3 Thomson (1897a, 108) tells us: “The assumption of a state of matter more fi nely subdivided than the atom of an element is a somewhat startling one.” His startling hypothesis is placed in a theoretical context that many will now fi nd somewhat strange and, as seen in §5, is not at all like the particle models that were subsequently adopted, even by the later Thomson. 4 The history of the investigation of electrons in nineteenth- and twentieth-century physics is quite complex; only a very superficial account is given here to illustrate the various kinds of realism and how denotation can be fixed. For some of the more complex details see Dahl 1997, Buchwald and Warwick (eds.) 2001, and Arabatzis 2006.
202
Robert Nola
2 GETTING ACQUAINTED WITH ELECTRONS BY DESCRIPTION
2.1
The Initial Experimental Discovery of . . . What?
J. J. Thomson never claimed to be the discoverer of electrons: “The fi rst observer to leave any record of what are now known as the Cathode Rays [subsequently renamed “electrons”] seems to have been Plücker . . .” (Thomson 1897a, 104). Plücker who? Since the beginning of the eighteenth century, experimenters had been investigating electrical discharges in closed tubes that contained different gases; it was noted that sparking could take place over greater distances with lower pressures of the gas. Investigators were hampered by technical shortcomings such as their inability to produce an adequate supply of electricity for the discharge in the tubes, and to stop atmospheric leakage into the tubes at low internal pressures. Some of the best work was done in the late 1830s by Faraday. He was well aware that in the case of air, as it was pumped out, the sparking between the anode and cathode ceased and was replaced by violet streamers, which in turn were replaced at lower pressures by a pinkish glow that fi lled the tube. Below about 5 mm of mercury pressure (the pressure of one atmosphere being 760 mm of mercury) the pinkish glow concentrates towards the anode, a blue-violet glow appears at the cathode, and a dark space emerges between them. Faraday was able to investigate it, and it is now known as “the Faraday dark space.” Given the limitations of the apparatus he used, Faraday was unable to produce pressures below about one-thousandth of an atmosphere; so he ceased further work. By the second half of the 1850s, new technology became available for making improved coils for producing electric currents (due to Rühmkorff) and for making efficient pumps and preventing leakage (due to Geissler). Using these new coils and pumps, in 1857 Julius Plücker was able to get the pressure of the contained gas down to about one hundred-thousandth of an atmosphere, and other experimenters who took up his work were able to get the pressure down to a millionth of an atmosphere. At these low pressures totally new phenomena could be observed. As the pressure decreases the glow around the anode breaks up into striations; and then the blue-violet glow around the cathode breaks into two. One part adheres around the cathode while a new dark space emerges (subsequently called “the Crookes dark space” after William Crookes who investigated it). The Crookes dark space expands pushing the other part of the blue-violet glow, the Faraday dark space, and the striations towards the anode until the Crookes dark space fi lls the tube. Then from about 0.01 mm of Mercury and lower, a yellowish-greenish glow appears on the walls of the glass tube. This is a quite new and surprising phenomenon. Plücker was aware of Davy’s 1821 experiment which showed that an electric discharge could be deflected by magnets; and he was aware of Ørsted’s earlier work on
From Reality to Truth
203
the relationship between magnetic fields and electric currents. One of his tasks in conducting the experiment was to discover what effect magnetic fields would have on the new phenomenon in the Geissler tubes.5 In some experiments he used a flat cathode plate; but in others he used a point-like cathode. In the latter case, not only did a more concentrated glow appear on the glass tubes but also a line of light could be observed coming from the point-like cathode.6 This ray of light was readily deflected by horseshoe magnets, as well as the more concentrated glow7 that it appeared to cause on the surface of the glass. Given the outline of the experiment above, the following can be set out. (E) There is a repeatable experimental set-up concerning the apparatus (viz., Plücker’s use of Geissler’s tubes) which, under conditions C (of its electrical source and sufficiently low pressure beyond the emergence of the Crookes dark space), has a law-like disposition to produce phenomenon P (“P” for Plücker). The four central Plücker phenomena are as follows: (P1) There is, in conditions C, a colored glow (in Plücker’s experiment a greenish-yellowish light) on the glass of the tube (usually at the end opposite the cathode). (P2) In C, the patches of colored glow can be moved by a magnetic field. (P3) A point-like cathode, rather than a cathode plate, produces a visible beam and a less diffuse, more concentrated glow effect. The following manipulation effect can be observed as a special case of P3: 5 A large variety of Geissler tubes were used of different shapes and sizes, with or without bulbous ends, with cathodes and anodes placed in various positions, and the like. A commonly used simple tube is a small cylinder of glass about 30 cms long, sealed at both ends, with cathode and anode at opposite ends (each of which is linked to a source of electricity); the tube is also connected to an evacuation pump. 6 The later understanding of this phenomenon is that the more concentrated stream of cathode rays (= electrons) that were given off the cathode were able to impact with residual molecules of the remaining gas which is at a quite low pressure. Later experimenters placed a screen coated with sensitive material inside the tube so that as the cathode rays passed down the tube they grazed the coated screen, creating a visible indication of their path. 7 It was later discovered that the color of the glow was due to the chemical nature of the glass; so the color plays no role in the story to be told here.
204
Robert Nola (P4) With a point-like cathode, the observable ray coming from the cathode and the patch of colored glow it causes on the tube can be deflected by a magnet.
Plücker also noted a number of other phenomena arising from his experiments, but they do not play a central role in the identification of whatever it is that produced the observed effects (though a comment in §4.2 explains why (P 7) is to be excluded from the identifying description constructed below): (P5) The glow is independent of the anode position. (P6) The same glow occurs when several different metals are used as cathode and anode. (P7) A “sputtering” effect occurs in which some of the platinum of the cathode is torn off and deposited on the inside of the glass tube, slowly blackening it.
2.2
An Identifying Description Becomes Available.
Given claims (P1) to (P4), call these collectively “P”, an identifying description can be constructed which denotes a “something” arising in Plücker’s experimental set-up; the description is a version of Russell’s theory of descriptions but generalized to kinds rather than individuals: (1) There exists a unique kind k such that instances x of k in conditions (E) cause effects P (under manipulation of xs). (Or using “(¶–)” as the defi nite description operator: (2) (¶k)[(x) if x instantiates k & x is in (E), then x causes effects P (under manipulation)]. For brevity let us write (3) (¶k)[Pk]; that is, the (unique) kind that Ps (i.e., the unique “Plückerer”). In this case there is no explicit introduction of a name to refer to the “something” along the lines of a purely causal theory of name introduction (for a kind). First, no name is in fact being introduced; and second the “something” that is being denoted is not, and cannot be, in any direct perceptual causal contact with any person who might be construed to have introduced a name or who even introduced the description. Even though we humans have always been in some causal relation (proximal or distal) with these
From Reality to Truth
205
“somethings,” and earlier experimenters had also been in a causal relation with them through their particular experimental set-ups, the experimental situation that Plücker and his coworkers used to pick out the “somethings” provided precise and repeatable identification of the “somethings” across experimental set-ups. There are good grounds for claiming that a description like (1) was readily available to Plücker and his coworkers as they communicated about what was happening in the experimental setup. Also supposing that some such description is available offers the best explanation of their behavior. Once grounded in their activities, such an identifying description can then be communicated to the scientific community at large. This Plücker did by publishing his results.8 Physicists then had a recipe for constructing their own discharge tubes and replicating the experimental conditions (E) and the observed phenomena and then to begin experimenting on whatever is the kind of “something” that causes the phenomena. Note that the glow and the manipulability effect are important for fi xing the denotation of the description (2), or (3). If some other things y were to be given off the plate but they could not be manipulated and yet caused a glow effect, then they would not be the items picked out by (2). Again, if some further items z were to be given off the plate and they could be manipulated, but they do not cause the same glow effect, then they would not be the items picked out in (2). What is required to pick out the appropriate items x of kind K is both the glow and the manipulability effect as set out in conditions (P1) to (P4) taken in conjunction, as in (2). Hacking’s proposal “if you can spray them, then they are real,” provides a sufficient condition for something being real but not a necessary condition (pulsars and tectonic plates are real, yet we cannot spray them). All the experimenters who followed Plücker recognized that, when the current was turned on, “something” was “sprayed” off the cathode, and it ceased to be “sprayed” when the current was turned off. Not only that: what was sprayed off was evidently manipulable using the magnetic fields produced by horseshoe magnets. Hacking’s criterion can be generalized to say: “if x can be manipulated by manipulating some y then, as well as y being real, x is also real.” In Plücker’s experiments the “somethings,” though they are manipulated by being “sprayed off” the cathode when the current is on, are not used to manipulate anything; rather, magnetic fields are used to manipulate them. However, the manipulated and the manipulator must be equally real.9 But from this it does not follow that we need know anything of their intrinsic properties or natures or how the manipulations take place;
8
An English translation of some of his results can be found in Plücker (1858). Manipulation and intervention can be used to give an account of causation; see Woodward (2003). As such these relations, and especially what they relate, need to be understood realistically. 9
206 Robert Nola we need only know that they take place. The nature of magnetic fields was as unknown as the natures of the “somethings” (cathode rays) the fields were used to manipulate. Nonetheless, our manipulating the magnetic field enables us to manipulate the path of the “somethings.” In addition, the very conditions under which the “somethings” are identified and re-identified are part of the denotation fi xing generalized description (¶k)[Pk]. It is hard to see what could be more realist than this.
2.3
Presuppositions of the Identification
We cannot be readily assured that some unique kind is picked out by the generalized defi nite description (¶k)[Pk], let alone what kind it is or what nature it has. Certainly for many years after Plücker’s discovery scientists had no idea what they were working with beyond the experimental interactions they investigated. But the presupposition that there was one kind of thing they identified, and then re-identified in other experimental set-ups, was a natural one to make. Four comments can be made concerning the kind assumption. (I) The description picks out “something,” but it does not specify to what ontological category the “something” belongs. This was unclear at the time they were discovered, and remains unclear even now because we are not sure what kind of thing the “somethings,” viz., electrons, really are. The “somethings” were variously regarded as flow of electricity (whatever that may be), or beams or rays—though of what remained unclear. But at least there is a kind of “something” involved as there appear to be many instances of the kind arising in different experimental set-ups that are repeatably the same. The generalized description need not specify a kind in order to denote some item: the description “(¶k)[Pk]” still denotes even though it does not say what kind it denotes. In fact the description picks out its denotation by means of extrinsic (causal) properties and not any intrinsic properties. As well as “flows,” “beams,” or rays, a number of other broader candidate ontological categories have been proposed: substantial object such as a particle or a corpuscle (the view of J. J. Thomson); events or processes such as some sort of electromagnetic wave in the ether (the view of many nineteenth-century ether theorists); or as current string theory postulates, a ten-dimensional, fluctuating filament of energy; and so on. In this list can also be included vexing questions about the identity conditions of electrons raised in quantum mechanics10 and issues about the qua-problem raised for the denotation fi xing of names that lead some causal theorists of naming
10 For some of the problems about the identity of subatomic particles see French and Krause (2006).
From Reality to Truth
207
to adopt a modified causal-descriptivist account instead.11 What is important, however, about the identifying description (¶k)[Pk], is that it can perform its task of identifying the same (kind of) “something” across different experimental set-ups without having to say into what ontological category its denotation falls.12 The identification is made through the “somethings” standing in a particular causal nexus of the sort specified in (¶k)[Pk], whatever the ontological category of the “somethings” in that nexus. Resolving matters concerning ontology will be a task to be performed by both an advance in science and in the metaphysics of the categories of reality. But getting on with the task of the scientific investigation of realist items such as electrons does not require that we settle these issues fi rst, or even second or third—perhaps only towards the end of science. Their identification and re-identification can be successful without settling matters concerning ontological category or nature. (II) If there is a kind picked out, it need not be the lowest level of species that does the causing; it could be at a higher level of taxonomy such as a genus. There may be many species of the one genus each of which produces the same effect (but perhaps each does it in a slightly different way). Such is the case when different isotopes of the elements, already well known, such as oxygen or hydrogen, were discovered in the twentieth century. In such a case “hydrogen,” for example, was not deemed to be a term which lacked denotation; rather it denoted a genus of which there were several species. With the advance of science, refi nement of the denotation occurred when the terms “protium,” “deuterium,” and “tritium” or the labels “1H,” “2H,” “3H” were introduced.13
11 For an account of the problems that face both a purely causal account of reference fixing and any move towards a modified causal descriptivism, particularly the qua-problem, see Devitt and Sterelny (1999, §§4.5, 5.3). What is claimed here is that some of these problems can be bypassed using a description as a denotation fixer that identifies the items denoted, whatever their ontological category, by the causal and manipulation relations between these items and their observable effects. 12 In specifying the common work that realists and empiricists can do together in the philosophy of science, van Fraassen (2003, 490) singles out one area where there might be no common work possible: “Realists are engaged in ontology: do the categories of substance, accident, essence, haecceity, causality, determinism, and determinacy apply to certain (putative) sectors of reality?” But then he adds that “empiricists cannot (as I see it) in good conscience participate in it [traditional ontology].” The suggestion made in the text is that a combination of the advance of science and work in metaphysics can settle some of these questions without undercutting realism or, at least, a realism about electrons. 13 There are in fact seven isotopes of hydrogen; only the fi rst three occur naturally; the other four have been created in the laboratory and are highly unstable.
208
Robert Nola
(III) The unique kind presupposition is challenged in cases where the causes turn out to be quite heterogeneous; several quite distinct kinds of thing might give rise to a similar effect. (For example, there are different kinds of thing that have the same effect of poisoning but do this in different ways; they are still the same “functional” kind even though they are not a natural chemical kind.) But from this we should not conclude that the generalized description lacks denotation altogether. There are many cases of this in science. As an example, consider the term “atmosphere,” which we pick out by the causal role it plays in giving us certain feelings on a windy day, sustaining our lives when we breathe it, sustaining fi res, and the like. Though Aristotelians thought it to be a single kind of substance, we now know it to be a heterogeneous collection of different kinds. But in discovering this we did not conclude that there is no atmosphere. Here the idea of denotational refi nement can also come to play a role in the advance of science when a defi nite description is found to pick out a heterogeneous class (but without the description becoming denotationless). (IV) The unique kind assumption can be understood in a sufficiently relaxed way so that in cases where there is no one kind involved successful denotation is not undermined by the involvement of multiple or heterogeneous kinds. In the case of Plücker’s discoveries, we now take him to have been performing experiments on streams of electrons. So far as we know, there is just one kind of electron; it is not thought to have any further underlying structure of quarks, or whatever, unlike the proton which is thought to have a three-quark structure mediated by gluons. We might have been lucky in that the original unique kind assumption turned out to be correct; in fact the kind might be an ontological simple. But this is something that only the world will tell us as we develop our theories of the electron. The electron has not gone the way of Blondlot’s ill-fated N-rays, which turned out to be nonexistent.14 There are currently some highly speculative claims that at very low temperatures electrons can be caused to divide into two electrinos; so electrons might turn out not to be ultimate simples.15 But even if there are electrinos, and the unique kind supposition has not been correct, the term “electron” would not thereby be denotationless; it would ambiguously denote the two kinds of electrinos that have been postulated. Setting aside this possibility, the world has made Plücker’s defi nite description lucky; there is a “something” that uniquely satisfies it; moreover it satisfies it perfectly.
14 For an account of the N-rays that Blondlot postulated in 1903 see Dahl (1997, ch. 14.1). 15 The highly speculative supposition of electrinos in Maris (2000) does at least provide some explanation of otherwise puzzling and hitherto unexplained phenomena concerning very low temperature liquid helium.
From Reality to Truth
2.4
209
Finding Out More About What Can Be Identified
A burning question for physicists was: what is going on in experiment (E)? Plücker’s discovery led to a spate of reduplications of his experiment in Europe; many were performed to learn more about what was happening. Here I mention two experimental discoveries; in §§4 and 5 more is said about the theories that were proposed as the cause of the phenomena. The fi rst experimental discovery is due to Johann Hittorf, a student of Plücker, who followed Plücker’s recipe for reduplicating the experiment. He placed a solid body, for example one in the shape of a Maltese Cross, between a point-like cathode and the glow on the walls of the tube and observed that a shadow of the cross was cast on the tube’s walls. At this point, inference to the best explanation plays a role. The best explanation of the shadow effect is that the “somethings” coming from the point-cathode travel in straight lines (assuming of course that there is no magnetic field present). The paths of these “somethings” can be blocked by a body, causing a shadow to appear in the greenish-yellowish glow. This can be summed up in a further condition (H) for Hittorf: (H) The “somethings,” items x of kind K, travel in straight lines (in the absence of magnetic fields) from a cathode to the walls of the tube causing its glow; and if the cathode is more point-like, and items x are intercepted by an opaque body, then a shadow is cast on the walls of the tube. Thus the kind of “something” that Plücker identified, viz., (¶k)[Pk], is discovered to have a further set of properties; call these “the H-properties.” Collectively physicists could then say of the “somethings” already identified by (¶k)[Pk]: (¶k)[Pk]Hk. Hittorf also introduced a name for the “somethings” Plücker identified, using the description: Glimmstrahlen (glow rays) = (¶k)[Pk]. The name, like many others introduced at the time, did not really catch on. The second group of discoveries are three culled from the many made by Eugen Goldstein. (G1) Shadows are cast by an opaque object not only when the cathode is point-like but also when the cathode is a large extended surface compared with the object placed in front of it; but in this case the shadow cast is diffuse and not sharply defi ned.
210 Robert Nola (G2) The “somethings” coming from the cathode are not emitted in all directions, as in the case of light from a light source, but are emitted largely at right angles to the surface of the cathode and travel in straight lines (in the absence of a magnetic field). This shows that the “somethings” are not like light. This was important for Goldstein because he was an ether theorist about both light and the “somethings”; but, because they behaved differently in the respect mentioned, they must be different features of waves or motions or vibrations in the ether. (G3) Goldstein definitively established a result that Plücker anticipated—whatever is emitted from the cathode is the same regardless of its metal composition. Here there is an enumerative inductive inference from the many cases of metal electrodes that Goldstein investigated, such as tin, platinum, iron, etc., to the conclusion that the “somethings” given off from the cathode were independent of the metal composition of the cathode. These three discoveries can be conjoined to form the claim (G) so that the overall knowledge about the “somethings” now becomes: (¶k)[Pk][Hk & Gk]. In 1876, nearly 20 years after they were fi rst discovered, Goldstein introduced a name that did catch on. It referred to the “somethings”: Kathodenstrahlen (cathode rays) = (¶k)[Pk]. The defi nite description (¶k)[Pk] plays an important role at this early stage in identifying the “somethings.” The claim that the description denotes is equivalent to the claim that there is a (kind of) “something” such that instances of it in (E) cause effects P (under manipulation). From our later stance this merely appears to be a law-like generalization about the behavior of the “somethings” in certain circumstances. From the point of view of the physicists of the time, however, this played a more central role in enabling the identification and re-identification of what the experiments are directed upon. Up to this point, what is known about cathode rays, and the properties used to identify them, are extrinsic properties either gotten by observation, or by experiment, or by using a little theory from elsewhere in physics to infer some property (for example to infer that they travel in straight lines). Nothing has been determined, so far, about their intrinsic properties, let alone their ontological category. And nothing is said about any explanatory theory about, or theoretical models of, cathode rays. Some of the many different, rival theoretical models of cathode
From Reality to Truth
211
rays are discussed in §4. The next section concerns the Ramsey–Lewis account of theories and theoretical terms and the place of the generalized description, (¶k)[Pk], in that context.
3 THE RAMSEY–LEWIS THEORY OF DENOTATION FIXING
3.1
Ramsification and Descriptions
There is a connection between the use of defi nite descriptions in identifying cathode rays, and general procedures for the Ramsifi cation of theories. Consider a theory T, the nonlogical vocabulary of which can be divided in two. The fi rst are the O-terms, short for “observation terms.” Despite being called “O-terms,” these are rarely the observational terms supposed by some positivists. They are other or old terms that have their meaning fi xed outside the context of theory T; for that reason O-terms are best thought of as “outsider” terms. In contrast, Tterms (i.e., theoretical terms) have their meaning fi xed by the theoretical context of T in which they occur; alternatively they have a role defi ned by their theoretical context, or they are “implicitly” defi ned by the context of T in which they occur (to use an older and perhaps less satisfactory notion). T-terms are “insider” terms; they get their meaning by the role they play inside theory T. Not all theoretical terms of theory T will occur in a name position; but it is possible to syntactically transform T so that its theoretical terms occur in a name position, thereby enabling one style of quantifier to be used.16 Theories will also contain a (fi nite) number of T-terms and O-terms. To simplify matters let us suppose that T has just one T-term, “τ” and that the (fi nite number of) O-terms can be represented by the one symbol “O” (and I will assume in what follows, without further mention, that terms of logic are readily available).17 Then theory T can be written as: (1) T(τ, O). The next step is to replace the theoretical term by a variable; this yields the following open sentence containing only a variable, logical expressions and O-terms:
16 See Lewis (1970/1983, 80) for how theoretical terms, which can come in many different syntactical categories, are to be placed in name position to facilitate the use of the one style of quantification. 17 For the quite general case of n T-terms and a fi nite number of unspecified O-terms, see Lewis (1970/1983, §§III, IV).
212 Robert Nola (2) T(x, O). Denote the Ramsey sentence of (1) by “TR.” This is obtained by placing an existential quantifier in front of the open sentence (2) to form a closed sentence: (3) TR = (∃x)T(x, O). In forming the Ramsey sentence, though the sole theoretical term has been eliminated, the variable x still ranges over the theoretical entities in the world. In the example above, the Ramsey sentence says that there does exist at least one item in the domain of the world over which the existentially quantified variable ranges that satisfies (2). (This leaves open the possibility that there might be more than one.) If P is this worldly item, then P realizes (i.e., satisfies) the open sentence (2), and it also makes the Ramsey sentence (3) true. Lewis’s modification to Ramsey’s approach attempts to close off the possibility that there might be more than one realizer. This is done by replacing the existential operator in front of (3) by a defi nite description operator. If the defi nite description operator is “(¶x)” then a defi nite description of the following form results: (4) (¶x)[T(x, O)]. A name “τ” can be introduced to denote what the description denotes: (5) the denotation of “τ” is given by: τ = (¶x)[T(x, O)] (i.e., the item denoted is P). What fi xes the denotation of “τ” (if there is one) is the defi nite description which employs only “outsider” O-terms. Of primary importance is the description; of lesser semantic importance is the name that piggybacks on the description for its denotation, which is effectively shorthand for the description. This illustrates that (4), which has the form of a Russellian description, is nothing but a special case arising from the Lewis descriptions given in (4). So one can set, within the general context of Ramsification, the account of Russellian defi nite descriptions in the previous section about how one was initially able to refer to cathode rays. As Russell would put it, we have a way of denoting an unobservable item with which we are not directly acquainted by using a defi nite description about items with which we are acquainted—viz., the items denoted by the O-terms.18 The defi nite
18 It is unclear to what extent Russell, and others, were aware that the theory of descriptions is a special case of the Ramsey sentence (or to be more correct the
From Reality to Truth
213
description (¶k)[Pk] is a simple one that appeals only to the causal role that the “somethings” play in bringing about the effects Plücker observed upon manipulating the “somethings.” The idea that they are “somethings” is well captured by the defi nite description operator.
3.2
Possibilities of Realization
On Carnap’s understanding, the Ramsey sentence says that if there is an item that realizes the Ramsey sentence then the theoretical term of the theory will denote that item. This allows that there can be multiple realizations. For Carnap this sits quite happily with his brand of antirealism in which, as far as he was concerned, the difference between realism and instrumentalism was merely linguistic.19 So it is necessary to look separately at the cases in which the open sentence (2) is realized by just one item, none, or more than one. Unique realization arises if the open sentence (2) is realized by one and only one item,
Ramsey–Lewis description). Note also that in adopting Russell’s talk of knowledge by acquaintance and description we are not committed to Russell’s epistemology. However it is useful to talk of things with which we are not acquainted (they are not observable or are unobservable) but which we can get to know by means of description in terms of those things with which we are already acquainted, i.e., the denotata of the O-terms, which have their meaning and/or denotation specified outside the context of theory T. 19 For Carnap’s account of the Ramsey sentence, see Carnap (1966, ch. 26–28). For his “resolution” of the realism/antirealism dispute, see especially pp. 255–256. 20 In the general case of a theory with n theoretical terms, T(τ 1, τ 2 , . . . , τ n, O), the Ramsey sentence will be of the form TR = (∃x1)(∃x 2) . . . (∃x n)[T(x1, x 2 , . . . , x n, O)], and the associated open sentence T(x1, x 2 , . . . , x n, O) will be realized (if it is realized) by the n-tuple
214 Robert Nola there is no item that realizes them. The case for this is made in §§4 and 5, where some of the theoretical models of electrons are investigated and shown to be quite false theories, including even Thomson’s own theory. This indicates that some modification needs to be made in the use of the Ramsey–Lewis theory (see §4.1). We also need to allow for both perfect realizations and imperfect but best realization. In the case of the defi nite description used to pick out electrons, the O-terms in the description are all qualitative; and in fact <electron> is the perfect realizer of the open sentence [P-] associated with the defi nite description (¶k)[Pk]. However much later when laws were developed about the behavior of electrons, they were at best only approximately true of electrons. That is, the worldly “somethings,” electrons, are not the perfect realizers of the proposed laws governing electrons. Rather than say that there are no realizers of these laws, it can be allowed that there are imperfect but very good, and perhaps even uniquely best, realizers of these laws. The importance of this emerges in the fi nal §6, when I discuss the development of the various Hamiltonian equations for the electron. So, interestingly, although we can get by with the idea of a perfect realizer for early theories of the electron which yield qualitative denotation fi xers, later classical and quantum theories yield denotation fi xers expressed in quantitative terms, and these do not strictly denote electrons. Thus the idea of an imperfect but best realizer of an open sentence based upon equations about the behavior of electrons comes into its own. Carnap’s account of the Ramsey sentence allows that two or more items realize it. In the case of our simple electron theory the two realizers might be, say, <electron> and <schmelectron>. Lewis’s introduction of a generalized defi nite description is intended to stave off the possibility of multiple realization; but if it were to arise then the definite description is to be regarded as denotationless. Multiple realizability of theories can be resisted on a number of grounds. (i) Uniquely realized theories are more satisfactory than multiply realized theories because there is nothing in the way in which scientists propose theories that suggest that they would settle for multiple realization; instead they aim for unique realization and largely achieve it. If scientists had intended to leave open the possibility of multiple realization then they would have used existentially quantified versions of the Ramsey sentence rather than the defi nite description version proposed by Lewis. By using the defi nite description version the possibility that T is multiply realized is ruled out. (ii) Good scientific theories will be uniquely realized. It is not that multiple realization is not possible, but it is not a desideratum for being a good theory. (iii) To countenance multiple realization is, in the context in which Carnap sets matters, to concede too much to instrumentalists and other nonrealists. (iv) In the context I am considering, the interpretation of theoretical terms of a theory takes place in a context in which all the O-terms of a theory have an interpretation outside the context of
From Reality to Truth
215
the theory. The idea that all theories must be multiply realized is not established in the case in which O-terms already have a fi xed interpretation. 21 A different position can be found in later papers by Lewis when he adopted Hartry Field’s notion of partial reference (Field 1973). In this account, at some stage of theoretical development, the Ramsey sentence of the theory might be satisfied by two or more items. Rather than say that any term introduced on this basis is denotationless, Lewis says that it is ambiguous in reference. Multiple realization is to be accommodated by the fact that theoretical terms are ambiguous, it being left open whether future development of the theory exposes and removes the ambiguity. 22 To illustrate, consider just the simple case of the definite description used to pick out electrons. Suppose this is multiply realized by two different 1-tuples <electrons> and <schmelectrons>. Though they both realize the same Ramsey sentence, they are distinct items. Like most or all of our theories, the theory of electrons is not a fi nal or a complete theory; it is a provisional theory that survives for a time and then is replaced by another theory. The replacing theory can then reveal the difference between the two items. Alternatively, a new experiment displays the difference between electrons and schmelectrons. Multiple realization can arise in other ways: for example, Grover Maxwell’s claim that different items can stand in the same causal structure. Note that the denotation fi xer (¶k)[Pk] relies on causal-structural relations in which the “somethings” denoted stand; given Maxwell’s claim, multiple items could be picked out by the description. He says: the reference of the term “electron” is fi xed (ontologically) by specifying the positions that electrons occupy in causal-structural networks. Similarly the reference of “having a negative charge of 4.8 × 10–10 e.s.u.” is (ontologically) fi xed by the causal-structural role played by such charges. However the reference of such terms is not (to this date) epistemically determined. The terms do refer to intrinsic properties, but we do not know what the referents are, e.g., we do not know what a negative electric charge is. (Maxwell 1978, 397) This accords with the view that the pair <electron, negative charge> realizes our claims about the causal-structural relations in which these items
21
Some of these considerations occur in Lewis (1970/1983, §III). So far I have followed Lewis’s earlier paper, “How to Defi ne Theoretical Terms,” which appeared fi rst in 1970. For his later view see his 1984 paper “Putnam’s Paradox”; here he talks of indeterminacy of reference. See also a 1994 paper, “Reduction of Mind”; here he talks of ambiguity of reference. 22
216 Robert Nola stand which are expressed in our theories, or the claims made by simple denotation fi xers such as (¶k)[Pk] in the case of the 1-tuple <electron>. And it accords with the view expressed above that though we may be correct about the causal-structural relations in which these items stand, it does not follow that we need know anything about their intrinsic natures or their structures. From this Maxwell argues that multiple realization is possible. He provides a diagram (Maxwell 1978, 384) in which items a and b are cause or effect nodes in a large set of causal relations to other nodes in an overall network. He uses this to illustrate the following principle: it is (logically) possible for different causal networks to have the same causal structure; or, in other words, one and the same causal structure may be realized in a number of different ways. . . . Thus God could have created a causal network such that it differed from the one in the diagram only in that the positions occupied by a and b were occupied by different events. . . . This creation would have been a different causal network, but it would have been the same causal structure. (Maxwell 1978, 390) What is being claimed is that if items a and b in one causal network were to be replaced by different items, say c and d, then a different causal network would be present. But both causal networks would have the same causal structure. Multiple realization follows from this. Because (i) the very same causal structure can be instantiated by different causal networks (these being individuated by the different items standing in the same causal-structural relations); and (ii) because causal structures are the means whereby the items that stand in causal relations are picked out (on the basis of the kind of denotation fi xing described above), then it does not follow that one and only one item will be picked out—there may be several. The crucial premise here is (i). Maxwell uses it to argue for a version of structuralism that downplays the role of our knowledge of the intrinsic properties of the items that stand in those structures claiming that “physical science provides us with knowledge of structural properties but not with knowledge of intrinsic properties” (Maxwell 1978, 397). Maxwell’s structuralism is epistemic and not ontological in that intrinsic properties do exist, but we cannot know anything of them except the causal-structural roles they play. It will be argued that Maxwell’s claim (i) has a counterexample; in the particular case of cathode rays we can come to know some of their intrinsic properties (such as their negative charge). Maxwell’s account can be challenged in the following way. If theory T is realized by both electrons and schmelectrons, then if T is not a fi nal theory and is open to revision at a later time, a new theory T* can come to prevail within which electrons can be distinguished from schmelectrons; it is discovered that they stand in at least one different causal-structural relation.
From Reality to Truth
217
Or T might bifurcate into two theories T* and T**: the fi rst satisfied only by electrons and the second only by schmelectrons. Here multiple realizability is due to the incompleteness of the theories we actually propose. Multiple realizability can arise not only for T but any successor theory T*. We think that not only were our theories of the past incomplete and so limited but also that our current theories are similarly incomplete; and such incompleteness can be extended to any theory proposed in the immediate future. For such a sequence of theories multiple realizability is not a problem. It is part and parcel of the growth of science in which we discover more and more of what is in the world. In discovering more, we not only introduce new denoting terms but also refi ne the denotation of older terms as we go along. But Maxwell’s point can come to apply in the following situation. What happens when the sequence of theories runs out, and our theories are no longer provisional but fi nal (assuming there is such an end point in the sequence of theories). Can multiple realizability still prevail? That is, in the fi nal theory T of electrons will there still be at least two 1-tuples, <electrons> and <schmelectrons>, which satisfy the same Ramsified sentence of T? This possibility is envisaged in one of the last papers of Lewis (forthcoming 2008); but it raises issues that go beyond matters to be discussed here.
4 THREE EARLY THEORETICAL MODELS OF ELECTRONS
4.1 Phenomenal Experimentation and Theoretical Explanatory Models Several theories or models were constructed to explain the growing range of phenomena that were observed concerning Plücker’s “somethings” identified by (¶k)[Pk]. From the beginning, even though physicists agreed about how to identify the “somethings,” and eventually came to agree about what to call them, they disagreed widely about what they were. In fact all the rival theories of cathode rays, up to and including Thomson’s own “corpuscular” theory, were seriously false of cathode rays. The existence of rival theories poses a problem for the standard employment of the Ramsey sentence. There is a procedure, introduced in §3.1, for Ramsifying a theory that involves all the component propositions of the theory. But if any of these rival theories were to be used to fi x the denotation of a theoretical term such as “electron,” then no common denotation, if any denotation at all, would be fi xed, even on a liberal construal of what it is to be the best but imperfect satisfier of the relevant Ramsey sentence. In the previous sections a different approach has been adopted in which a defi nite description, based in what we can be acquainted with through experimentation, has been used to fi x a denotation. Even though the theory of descriptions has
218 Robert Nola been shown to be an instance of the more general Ramsey–Lewis approach, the need to employ a full theory is sidestepped in fi xing denotation; in fact the denotation can be fi xed in a quite theory-independent manner. Such a use of the Ramsey–Lewis approach is consonant with what is known colloquially as the “Canberra plan.”23 The Canberra plan approach to investigating some domain of phenomena begins with collecting what are called the generally accepted “platitudes” about some domain. Or as is alternatively said, it begins with our “folk” knowledge of the domain. Thus if the domain of investigation is the psychophysics of color, then we begin with the host of folk platitudes about color, such as “snow is white,” “grass is green,” “white is what snow looks like to normal perceivers in normal conditions,” and so on. As color is the domain of investigation, color terms will play the role of the T-terms, and all other terms are either logical vocabulary or are O-terms (the “outsider” terms whose meaning is fi xed outside the T-terms of psychophysics). Other domains of investigation, such as our ordinary “folk” psychology, or our ordinary everyday “folk” moral claims, provides a set of platitudes from which any investigation can begin in each respective domain. Exactly how the platitudes are uncovered in any of the domains mentioned is not always a straightforward matter; but it is not one I go into here. In the case of the sciences, what would correspond to the “folk” scientific platitudes? They will be the host of ordinary reports that we can make either about what we observe or about the outcomes of experiments, an example of which would be Plücker’s reports of his experiment using Geissler tubes. Such platitudes are not always ready to hand as are those of the domain of color or of psychology. Rather they are a growing set of eventually well-established claims of experimental science, such as those added by Hittorf and Goldstein (see §2.4). The “folk” scientific platitudes stand in contrast to the theories or models that are constructed to explain these platitudes. Such a distinction is not unrelated to other distinctions that have been drawn in the sciences, such as that between phenomenal physics and explanatory physics; the former is a descriptive enterprise based on ordinary observation and/or experimentation, and the latter involves theories and models for the purposes of explanation of what we come to know of the phenomenal. From this it should not be concluded that the phenomenal is free of all theory, as what goes on in experimentation can draw on theory. This is something that the T/Oterm distinction allows. This distinction is drawn according to whether a term is an “insider” or an “outsider” with respect to a given theory T;
23 For an account of the philosophical method of the Canberra Plan see Nolan (2005, 213–227), and Braddon-Mitchell and Nola (forthcoming 2009) “Introducing the Canberra Plan.”
From Reality to Truth
219
as such outsider terms can contain vocabulary that draws on any theory other than theory T. Given these preliminaries a distinction can be made between the phenomenal (i.e., the experimentally determined) properties of cathode rays and the theoretical models that were used to explain the phenomenal. In what follows, models are not to be understood along the lines of semantic models for a theory. Rather a model will be taken to be an idealization of a real system that either fits the real system well, not so well, badly, or not at all. A model is an ideal object that does not simply set aside certain aspects of a real system; rather it idealizes some aspects of what is purported to be a real system, which it attempts to represent while not modelling other aspects at all. Thus a point particle is an idealization of a real object in that it attempts to model, say, a real system’s center of gravitation, or its position in space, but not its volume. Sometimes electrons are regarded as point particles that have mass and charge; they also have spin, or fi xed intrinsic angular momentum, which in turn is sometimes modelled on the spin the Earth has as it rotates its axis. In such a model it is hard to see how a point particle with zero volume can have spin; this is a matter that will hopefully be rectifi ed with increasing concretization of the models and the changing context of theory from classical to quantum theory. 24 The next subsection briefly sets out three of the theoretical models that were constructed to explain some of the phenomena observed about cathode rays. Only a little is said about the empirical inadequacy of these models. It is evident that each model is false of the real items it attempts to model. In §5 Thomson’s corpuscular model is considered.
4.2
Three Early Models of Cathode Rays
As already noted in §2.1, Plücker recorded the following about his experiment: in (P 7) a “sputtering” effect occurs in which some of the platinum of the cathode is torn off and deposited on the inside of the glass tube, slowly blackening it. He established this by showing that the metal deposited on the glass was in fact the same as that of the cathode. He speculated on the cause of this phenomenon: “It is clearly most natural to imagine that the magnetic light to be formed by the incandescence of these platinum particles as they are torn from the negative electrode” (Plücker 1858, §51, 410).
24 That the planetary model of spin suggested above is quite inadequate and that we have no good model of what spin really is, is argued in Morrison “History and Metaphysics: On the Reality of Spin,” ch. 14 of Buchwald and Warwick (2001, 425–449).
220
Robert Nola
This may be set out in the following two-part condition S (for sputtering), of which (i) is the report of observations, and (ii) is a causal explanatory claim: (S) (i) in (E) bits of incandescent platinum are sputtered off from the cathode and deposited on the glass; and (ii) the occurrence of (i) is the cause of the “rays of magnetic light.” Clause (ii), unlike (i), introduces a speculative causal explanatory hypothesis about one of the effects of “the rays of magnetic light”; but it does not provide any model of the “rays of light” themselves. Is this causal hypothesis correct? Once Goldstein discovered the existence of what he called “canal rays” (in effect a flow in the opposite direction from cathode rays of positively charged ions of the gas) he was able to show that Plücker’s hypothesis (ii) was wrong; the gouging out of the metal of the cathode was due to the impact of the much heavier canal rays and not the cathode rays themselves. What would happen if S were to be included in any denotation fi xer? Though clause (i) reports a correct observation, clause (ii) turned out to be a false explanation of what was causally happening. So nothing satisfies S because of the false causal claim (ii). If S were to have been included in any denotation fi xer, such as “the unique kind of thing k such that instances of k satisfy both (P&S),” this would lead to a defi nite description that picks out nothing. Excluding S is in accord with the suggestions above in §4.1 about how the Ramsey–Lewis procedure is to be applied in this context. This also suggests a maxim of conservativeness in selecting what is to count as a denotation fi xer. The second model is due to Goldstein but had been adopted by many others, such as Hertz. Goldstein was an ether theorist about light and cathode rays; both were some kind of disturbance in the ether. Since cathode rays come off the cathode plate at right angles, unlike light rays which radiate in all directions, Goldstein thought that they must be different kinds of ether disturbance. All ether theorists agreed about one thing: cathode rays were not the propagation of any material particle or of an electric charge from point to point within a cathode-ray tube. However, they did not always agree about what kind of ether disturbance was involved; so a number of different variations were proposed of the basic ether model. Of interest to all theorists was what was happening at the surface of the cathode and what was happening downstream from the cathode. In Goldstein’s model, at the surface of the cathode there is an interaction with the surrounding medium, the ether. The charged cathode produces a sudden, strong impact causing a disturbance in the ether in the form of longitudinal waves propagated in the direction of the impact. When the waves strike the remaining molecules of gas in the tube, there is an interaction between the medium and molecule, which
From Reality to Truth
221
in turn gives rise to a new source of similar wave disturbance that began at the cathode. Thus cathode rays were modelled as a series of locally generated rays initiating at the surface of the cathode with further local generation at each gas particle in the tube. 25 The third kind of model is that of cathode rays as a stream of particles propagated down the length of the tube. Such a stream could, of course, impact upon the gas particles that remained in the tube. Such impacts were used to provide explanations of the phenomena that could be observed in the tube; in particular their absence was used to explain the Faraday and Crookes dark spaces. Particle theorists, however, differed over exactly what kind of matter was transmitted down the tube. Varley, who was the fi rst to suggest a particle theory (1871), spoke of “attenuated particles of matter projected from the negative pole [cathode]” (qtd. in Dahl 1997, 62). Crookes’ model was sometimes equally unspecific in referring to a “fourth state of matter”; but on other occasions he was more specific about the order of the size of the particles; namely, they were molecular and cathode rays were really a “torrent of molecules.”26 The molecular torrent model accounted quite well for some phenomena, for example the quantitative deviation of cathode rays in magnetic fields. The model also easily allowed for the idea that the molecules had momentum and carried a charge (in contrast it was hard to see how waves could carry charge). It also provided a qualitative explanation of the appearance of the Crookes dark space. The “torrent of molecules” coming off the cathode pushed back the more slowly moving molecules of the rarefied contained gas; there is, however, a conflict at the boundary between the two where collisions take place. Thus the dark space is dark because no collisions can take place in contrast to the bright glow where collisions do take place. The “torrent of molecules” model ran into difficulties, some of which were raised by advocates of the ether model.27 (I) The molecules constituting cathode rays were thought to be as massive as the smallest chemical molecules (later this proved to be about 1,800 times the proper mass to be attributed
25 For a brief account of Goldstein’s ether theory, and Helmholz’s skeptical position, see Buchwald (1994, §10.2 “Goldstein’s rays,” 135–137); for an account of Hertz’s experiments, see §§10.3, 10.4 (137–169). Some of Goldstein’s considerations are available in English in Goldstein (1880). Thomson directed several criticisms at the ether theory, one of which was that “it is impossible to predict what will happen under any given circumstances, as on this theory we are dealing with hitherto unobserved phenomena in the aether, of whose laws we are ignorant” (Thomson 1897b, 293–294). 26 For an account of Crookes’ view, see Dahl 1997, §4.2. See also Whittaker (1951, 351–352) on the early particle theories of cathode rays of Varley and Crookes. 27 Some of the problems facing both models are summarized in Whittaker (1951, 353–354).
222
Robert Nola
to cathode rays, i.e., electrons). That they were too massive led to misleading quantitative results in many areas. (II) This also led to a false explanation of Crookes’ radiometer experiment in which it was thought that the rotation of the vane placed inside a cathode ray tube was due to the transference of the momentum of the impacting cathode rays to the vane. Thomson later showed that this was false, when he argued that the cathode rays particles were so light that the rotation could not be due to momentum transference (he argued it was due to differential heating). (III) Ether theorists (Goldstein28) showed that the mean free path of the cathode rays in the tube was about 1/150th that of the prediction from the “torrent of molecules” theory. The probability that these predicted outcomes should arise was so extremely low that ether theorists took this to be a point in favor of the wave model. (IV) For the molecular torrent model there was an expected Doppler effect due to considering the moving particles as sources of light; but this remained undetected. (V) As described in the next section, cathode rays impinging on the inner side of thin metal foils placed at one end of tube were detected on the outer side of the foil. It was thought unlikely that the particles passed though the thin foil. So this result was taken to favor the ether theory because the cathode rays emerging from the outer side of the foil were simply more of the same kind of local transmission of waves to be expected in the ether theory. (VI) Hertz, an ether theorist, unsuccessfully tried to show that cathode rays were also affected by electrical fields as well as magnetic fields. He took his lack of success to be a point in favor of the ether theory (wrongly, as it transpired).29 The above list of problems facing both models shows that there was doubt about the empirical adequacy of each. Solving these problems became an important part of the development of both models. But what is clear is that both models proposed quite different accounts of the same “somethings” that they all agreed were to be identified by (¶k)[Pk]. No ontological discontinuity can be inferred from this; but there is an “ontological” incommensurability between the ideal “objects” postulated in the wave models and the particle models that make the theoretical models rivals.
5 THOMSON’S CORPUSCULAR MODEL OF THE ELECTRON AND RAMSIFICATION. In 1897 J. J. Thomson published two papers (1897a, 1897b) in which he summed up a number of crucial experimental results that he and other
28 Some of these objections are spelled out in Goldstein (1880, 239–244) in his evaluation of Crookes’ particle theory. 29 J. J. Thomson later showed why Hertz was unable to discover any effect. For an account of the changing fortunes of evidence in this case see Achinstein (2001, ch. 2).
From Reality to Truth
223
experimenters had obtained. They contained his important and novel corpuscular model of cathode rays.30 In line with the Canberra plan approach to Ramsification, I fi rst consider the large number of newly discovered experimental results that became broadly accepted; it is these, not the full corpuscular theory, that carry the burden of denotation fi xing. Of the many experimental results, I mention only six (A) to (F).
5.1
Some New Experimental Discoveries Experimental Claim (A): all cathode rays are negatively charged.
Though cathode rays were known to be charged, it was not clear what charge they had. In 1895, Perrin showed that the charge must be negative; and by 1897, Thomson improved on his experiment to overcome some objections from ether theorists. Experimental Claim (B): the m/e ratios for cathode rays converge on values between 0.3 × 10 –7 and 1.6 × 10 –7. Experiments to determine this ratio were carried out from the late 1880s up to the time of Millikan and beyond. Each of these experiments assumed some theory about the motion of a particle in a magnetic field or some thermodynamics. In fact, in his two papers, Thomson gave results about m/e ratios by drawing on both kinds of background theories. This shows that these experimental determinations are not theory free; but they did not draw on any of the specific features of Thomson’s corpuscular model that I consider later. The next three experimental results are important independence claims which help establish that cathode rays were not just things local to cathode ray tubes; they provide inductive evidence for the ubiquity of the “somethings” called “cathode rays” throughout nature. Experimental Claim (C): the amount of magnetic deflection of cathode rays (in a constant field) is independent of the gas in the tubes. Whatever gas was used in the tubes, the magnetic deflection was always the same (for a constant field). From the gases used it is a simple inductive step to conclude that for all gases the cathode ray deflection is the same.
30 Thomson’s held the theory outlined here at least until his Nobel Prize lecture of 1906; but after that his views began to change. His later views are not considered here: neither are the reasons he did not call his corpuscles “electrons” until well after this time.
224
Robert Nola Experimental Claim (D): the m/e ratio is independent of the kind of gas used in the cathode tube.
This result is a consequence of Thomson’s work on the m/e ratio of cathode rays investigated in different gases in the tubes such as air, hydrogen, carbonic acid, etc. (Thomson 1897b, 306–307). This result can be inductively generalized: for all gases the m/e ratio of cathode rays is independent of the gas. Experimental Claim (E): the m/e ratios are independent of the kind of electrode used. This is an inductively based claim that Thomson reports from the experimental work of the ether theorist Goldstein, and others, all of whom used electrodes of different metals (aluminium, iron, platinum, tin, lead, copper, etc.). The fi nal two claims about the extreme smallness of cathode rays can be taken together. Experimental Claim (F1): cathode rays are much smaller than any other known chemical element by a factor of well over 1/1000. Experimental Claim (F2): the distance the rays travel outside the tube is only dependent on the density of the surrounding atmosphere and not the chemical nature of the outside medium (whether air, hydrogen, sulphur dioxide, etc). In both of his 1897 papers, Thomson reports experiments conducted by ether theorists, such as Hertz and Lenard, in which cathode rays are investigated outside the tubes in which they originate. It would appear that they could, unlike light, “pass” through thin foils of metal at the end of cathode ray tubes and be observed in a separate chamber. (Whether anything actually passes through the foils is a disputed hypothesis. An alternative view is that the thin foils themselves pick up a charge over their inside surface and then become cathodes emitting cathode rays from their outer surface—the view Thomson advocated. Similarly, ether theorists did not assume that anything passed through the foils; rather new, local waves could be produced at the outer surface of the foil.) Once outside the tube, the cathode rays displayed the characteristic glow effect and could be deflected by magnetic fields. The distance they traveled in the atmosphere of the observation chamber depended on its pressure only and not its composition; as it was evacuated they could travel up to several meters at very low pressures. Of interest was the distance the cathode rays travelled at normal pressures, or their mean free path (the distance they travel before the intensity of the rays falls by a half). For cathode rays it is about half a centimeter; that of a molecule of air is 10 –5 cms. That is, on average cathode rays travel in air
From Reality to Truth
225
thousands of times farther than the constituents of air do. From this Thomson concludes: “Thus, from Lenard’s experiments on the absorption of the rays outside the tube, it follows on the hypothesis that the cathode rays are charged particles moving with high velocity; that the size of the carriers must be small compared with the dimensions of ordinary atoms or molecules.” (Thomson 1897a, 108).31 There are two points to note about this quotation. The first is the distinction Thomson makes between an experimental fact (about mean free path of cathode rays outside the tube and their relative size) and the hypothesis that is meant to explain this, viz., that cathode rays are charged particles. It is this distinction that is drawn upon in taking the “Canberra plan” approach to the use of the Ramsey–Lewis account of denotation fixing described in §§5.2 and 5.3. The second is that even though Thomson drew on the work of Lenard, Thomson’s corpuscular hypothesis was not endorsed by Lenard, who was at that time an ether theorist.
5.2 A New Description for Fixing the Denotation of “Cathode Rays” All of (A) to (F) are experimental discoveries about a “something” which has already been identified by the description (¶k)[Pk]; no ontological incommensurability has crept in here. So the very same thing also has the properties specified in A to F. Moreover: cathode rays = (¶k)[Pk]= (¶x)[(A)& . . . &(F1)&(F2)]x. The fi rst description identifies cathode rays in the parochial setting of cathode-ray tubes. Moreover it is couched in terms which refer only to the extrinsic relations in which the cathode rays stand. The second description contains quite new elements which arise from discoveries about cathode rays both in their parochial setting and then outside it. It is also couched in terms which refer to some of the intrinsic features of cathode rays, such as charge and mass. Moreover it contains identifying features for cathode rays which at the time obtained wide currency. One of these is (B), the distinctive m/e ratio possessed by cathode rays but not by any other particles known in the physics of the time. Another is (C), the distinctive angle of deflection of cathode rays in magnetic fields. Other particles would have different angles of deflection; this would serve to differentiate one particle from another if there were several in the same field.
31 The path of the cathode rays outside the tube depends only on the density of the medium through which they travel. This also supports Experimental Claim (F), as cathode rays must be much smaller than the atomic elements they pass through if their mean free path is greater by an order of a thousand. Being so comparatively small, they can, so to speak, get through all the gaps that there must be in atmospheres at ordinary pressures without bumping into, or being absorbed by, the atoms of the atmosphere.
226
Robert Nola
5.3
Thomson’s Corpuscular Model—the “Startling Hypothesis”
What was Thomson’s model of cathode rays, or corpuscles, as he called them? Listed below are seven of the hypotheses that make up his model in the two 1897 papers; some hypotheses are retained in subsequent theories, and others are not. Hypothesis H1: (i) there is a primordial element of matter, much smaller in mass than that of any known atomic element, and (ii) it is a constituent of all matter. This is unexceptional but hardly uniquely identifying. However, associated with it is a further claim that is clearly false and which Thomson came to reject only well after his 1897 papers: Hypothesis H 2: the primordial element is the only element out of which all matter is constituted. Thomson then develops his explanatory hypothesis: Let us trace the consequence of supposing that the atoms of the elements are aggregations of very small particles, all similar to one another; we shall call them corpuscles, so that the atoms of the ordinary elements are made up of corpuscles and holes, the holes being predominant.” (Thomson 1897a, 108) Here Thomson introduces a name, “corpuscles,” for what he takes to be a single kind and sets out a further hypothesis: Hypothesis H3: there is a predominance of holes in matter. Thomson cites no direct experimental evidence for this, though he does use it to explain why such corpuscles have a greater mean free path than any of the atoms they comprise. The next hypothesis concerns the causal mechanism which produces cathode rays: “Let us suppose that at the cathode some of the molecules of the gas get split up into these corpuscles, and that these, charged with negative electricity, and moving with high velocity form the cathode rays” (Thomson 1897a, 108–109). Two further hypotheses can be identified here. The fi rst concerns how the cathode rays arise at the cathode and the second his core identity claim: Hypothesis H4: the molecules of the residual gas get torn apart at the cathode releasing some of the corpuscles to form cathode rays. Hypothesis H5: cathode rays are nothing but streams of corpuscles.
From Reality to Truth
227
There are two further quite speculative hypotheses that round off his model of corpuscles. Hypothesis H6: as the only constituents of nature, the corpuscles are primordial atoms (not the same as chemical elements) which arrange themselves in a law-governed way to constitute all matter including chemical elements. Hypothesis H7: the primordial atoms are really vortices in the ether and obey ether-vortex laws. Over many years Thomson worked on a highly speculative theory about how aggregations of primordial corpuscles would hang together in a stable configuration to form atoms. This is something that reaches back to his work in the early 1880s on how centers of repellent forces might arrange themselves in stable patterns. 32 Both Thomson and Maxwell had earlier suggested theories of a vortex atom arising from a singularity in a uniform ether. The theory originates in Helmholz’ work on perfect fluids, in which indestructible vortices that obey certain laws of rotational and translational motion emerge. The theory has an application in hydrodynamics, but its more speculative use was as a theory of how the primordial atoms that constitute all matter in the universe emerge as vortices in a universal plenum, such as the ether. One suggestion Thomson makes is that there is a law of force of the kind envisaged by Boscovich in which at small distances the force is repulsive but at greater distances is attractive—this involves considerable mathematical complexity owing to the number of interactions involved. As an alternative he suggests a model based on experiments concerning how different numbers of floating magnets arrange themselves in patterns of equilibrium (Thomson 1897b, 313–314). This is important background to Thomson’s last two fundamental hypothesis about his corpuscles: one about how the primordial atoms arrange themselves and a second about what these atoms are. These seven hypotheses are the core of Thomson’s speculative model of his corpuscles—his “startling hypothesis.” If they are conjoined, an open sentence can be formed by deleting the theoretical term “corpuscle,”
32
In an essay on how vortex rings can form stable combinations, Thomson wrote that “the properties of bodies may be explained by supposing matter to be collections of vortex lines in a perfect fluid that fi lls the universe” (Thomson 1883, 1). Aspects of the theory of vortex rings last for quite some time in Thomson’s thinking about his corpuscles; he devotes a whole chapter to how aspects of the vortex model might work in his informal Yale lectures of 1903; see also Thomson (1911, ch. V).
228
Robert Nola [(H1)& . . . &(H7)]x.
Placing a defi nite description operator in front, a definite description can be formed: (¶x)[(H1)& . . . &(H7)]x. Or for short (¶x)((H Tx)—“H T” being Thomson’s “startling hypothesis.” The defi nite description can then be used to fi x a denotation for Thomson’s theoretical term “corpuscle.” Does anything perfectly satisfy the open sentence or even play the role of being the best but imperfect satisfier? Because there is no such thing as the ether, then H7 is false; and so the description denotes nothing. However it is possible to reject H 7 while accepting H6. This would occur if one were to adopt the view that cathode rays are really material particles but still held the view of H6 that such charged material particles constituted all elements and still have to come together in some way to form chemical elements. Such is one way of taking Thomson’s talk of corpuscles but dropping the view that there is an ether. However it is still the case that H 2 and H6 (with or without H 7), and following in their train a false H 5, ensure that nothing either perfectly or imperfectly satisfies the open sentence. So, the defi nite description denotes nothing, and the term “corpuscle” fails to denote. What kind of denotation fi xer can be obtained from the overall consideration of Thomson’s theory, including his experimental results? It is a two-part defi nite description of the form: (¶x)[(H Tx)]&[{(A)& . . . &(F)}x]. Nothing realized this whole expression, as nothing realizes the fi rst conjunct, even though something, viz., cathode rays, realizes the second conjunct. This corresponds to the quite general use of the Ramsey sentence, or the Lewis–Ramsey denotation fi xer, which employs all the elements of a theory rather than some restricted subset of claims associated with the theory. But what carried the burden of denotation fi xing is not the full set of claims that Thomson makes about theory and experiment, as set out in the above description. Rather it is the second conjunct, which concerns only the phenomenal experimentally determined properties of electrons, that carries the burden of denotation fi xing.33
33 Not all cases of denotation fi xing can eschew the total theoretical context in which the term is introduced, but many can. Terms such as “neutrino” or “black hole” are highly dependent on the total theoretical context in which they are introduced.
From Reality to Truth
229
6 IMPERFECT REALIZATION AND THE HAMILTONIAN FOR THE ELECTRON As more and more experimental facts are discovered about electrons, new ways of identifying them emerge through the determination of quantitative values of their properties, rather than qualitative properties, as in the case of the initial identifying description such as (¶k)[Pk]. One such example is the description: “the unique kind of thing such that it is negatively charged and has m/e ratio of 1.6 × 10 –7” (this being one of the values determined by Thomson, 1987a, 109). Thomson’s value is of the right order but is otherwise incorrect; so nothing, not even the electron, is a perfect realizer of that description. But perhaps the electron is the best of the acceptable set of imperfect realizers of the description. Given a different description which takes into account error factors, the electron may turn out to be the (unique) perfect realizer of the open sentence “- is negatively charged and has m/e ratio of m± δ” (where “δ” is an error factor). Either way, the description can be used to fi x a denotation via perfect or best imperfect realization. These kinds of considerations can be extended to the laws that have been proposed to govern the behavior of electrons. The electron may not be the perfect realizer of these laws, but it may be the best of the imperfect realizers. And as the laws improve, so there will be a lessening in the degree to which they are imperfect realizers. There might be a temptation to say that the prehistory of the electron indicates that the kind of thing that was investigated at those times is not the same as the thing that was investigated later, when relativity theory and quantum theory came to dominate in physics. And this is reinforced by talk of changing concepts of the electron. But change in concept (however that is to be identified), along with change in the theories in which the concept is embedded, does not entail change in the thing denoted. What has been argued here is that, despite conceptual and theoretical discontinuity, there is continuity in the item denoted, right from the beginning of that prehistory until now. Such a view is also maintained in Bain and Norton (2001) who add to the story given above in the following three ways. The fi rst kind of continuity they locate is that due to historically stable, intrinsic properties attributed to electrons (Bain and Norton 2001, 453–455). There is stability in the sense that properties of electrons that are discovered at one point in the historical development of experiment, theory, and models are kept on in later historical phases. (This is consistent with some degree of instability which can arise if a property is not kept on.) Early stable properties (intrinsic or extrinsic) include: charge; Millikan’s determination of the nonfractional character of the charge; the m/e ratio (though better values of this were obtained over time); and degree of deflection in a given magnetic field. Later properties include spin and the Pauli exclusion property. These add to the core of growing knowledge of the properties in much the same way as did the discoveries of Hittorf and
230 Robert Nola Goldstein with respect to Plücker’s initial discoveries and his identification of what was discovered via the denotation fi xer (¶k)[Pk]. A second continuity they locate is “structure,” a common feature that is (often but not always) preserved through changes in theory; this “is simply the smallest part of the latest theory that is able to explain the successes of the earlier theories” (Bain and Norton 2001, 456). The appeal here is to a version of the correspondence principle: this says that later theories are to incorporate earlier theories as limiting cases (for example, under some parameter restriction or idealization of a law or idealization of some aspect of a model, etc.). There is no guarantee that there will always be such a structure, but where there is, and where it is combined with historically stable properties as is the case with the electron, then continuity of denotation is assured. To ensure this, however, the notion of the best but imperfect satisfier needs to be employed in considering any sequence of earlier theories taken as limiting cases of later theories. This leads to the third aspect of continuity, that due to laws. Bain and Norton argue that the structure that fi lls the bill is the Hamiltonian or Lagrangian for the electron. 34 There are a number of different Hamiltonians, as the Hamiltonian schema is embedded in one or another theory. Thus, following Bain and Norton (2001, 456) the Hamiltonian for the electron is a schema of the following sort: H = (p—eA)2/2m + eφ (where p is the momentum, e is the charge, m is the mass of the electron, and A and φ are the vector and scalar electromagnetic potentials). Embedding this into classical dynamics yields the theory that Thomson needed for his account of the deflection of his corpuscles in an electric field, or Millikan needed for his oil-drop experiment. Embedding the Hamiltonian in the special relativity theory produces a new equation with the additional factors above those provided by classical theory: H = [(p—eA/c)2c2 + m 2c4]½+ eφ. This introduces no new property, but it does describe the behavior of electrons by taking into account relativistic effects through additional factors. Using these equations one can form a generalized, defi nite description (call this “the Hamiltonian description”), which says, roughly: “the unique kind of thing k such that k satisfies Hamiltonian equation H” (there being a different defi nite description for each version of the Hamiltonian). Does
34 The Hamiltonian of a system is an equation concerning the energy of the system expressed in terms of momentum and position (i.e., potential energy).
From Reality to Truth
231
the electron perfectly satisfy the above classical, and classical relativistic, Hamiltonian descriptions? Or does it only imperfectly satisfy them? The electron cannot perfectly satisfy both; in fact it perfectly satisfies neither. It is the best imperfect satisfier of both; but it satisfies less well the fi rst Hamiltonian description embedded in classical theory, and it satisfies better the second Hamiltonian description embedded in the special theory of relativity (because the latter describes the behavior of electrons better than the former). The difference lies in the added factors and the different functional relation between the expressions contained in the latter. There are further versions of the Hamiltonian. A third form takes into account the novel property of the spin of the electron; in this case a newly discovered, but subsequently stable, property of electrons is accommodated within a new theory and yields a new Hamiltonian. A fourth version is the Hamiltonian due to Dirac; and a fifth builds on this by yielding additional terms to form a quantum-field theory account of the electron within quantum electrodynamics (QED). And so on. Again, each of these Hamiltonians forms a Hamiltonian description. Does the electron perfectly realize all of these Hamiltonian descriptions? It does not, yet the electron remains the best but imperfect realizer of each of these Hamiltonian descriptions, but with increasing degree of accuracy of realization. Using the story told by Bain and Norton, the denotation “electron” can now be fi xed as the best but imperfect realizer of an open sentence associated with a description which schematically says: “the (unique) kind which (i) has historically stable properties, P1, P 2 , . . . , Pn, and (ii) obeys a form of Hamiltonian (the form differing as to which theory of motion it is embedded in).” In the historical story told here, realism about electrons precedes the presence of correct theoretical models of electrons that purport to explain their behavior. Of course there is the truth expressed in the fact that the initial identifying description (¶k)[Pk] does denote and the truths subsequently discovered experimentally about what has been identified. But uncovering these truths took place against the background of deeply rival theoretical models of electrons, all of which were false of electrons. Current relativistic and quantum theories of electrons fare somewhat better in providing truths, or approximate truths, about electrons. But for those acquainted with current theories, electrons turn out to be quite strange entities the nature of which strains the traditional ontological categories into which we place items of the external world. But despite this, it does not follow that we cannot get a descriptive “handle” on electrons and then discover more about them with the progress of science.35
35 I would like to thank Alan Musgrave for his careful reading of, and comments upon, the penultimate version of this chapter. I made some changes in response but perhaps not enough to fully satisfy him.
12 Reflections on the Gem Fred Kroon
1
THE GEM
In his acerbically witty “Idealism: A Victorian Horror-Story (Part II),” David Stove excoriated a certain style of philosophical argument that he called “the Gem” (Stove 1991). He found an ardent supporter in Alan Musgrave, whose “Conceptual Idealism and Stove’s Gem” corrals a range of other idealisms that supposedly thrive on the Gem (Musgrave 1999). To get a flavor of this style of argument, consider Berkeley’s objection to materialism. Berkeley thought that we cannot even think or conceive of objects that exist “without the mind.” The reason? When we try to think of such objects, we fi nd that “they are apprehended by [the mind],” so that we are not, after all, conceiving of objects as existing without the mind (Berkeley 1710, paragraph 23). As Stove puts it, “you cannot have trees-withoutthe-mind in mind,” for “you cannot have trees-without-the-mind in mind, without having them in mind” (1991, 139–140). At best, you can have trees-in-the-mind in mind. And that, thinks Berkeley, is precisely what we do have in mind. Trees-without-the-mind are not even conceivable. What remains is Berkeley’s immaterialist universe. Musgrave and Stove take the most famous and most influential crafter of a Gem to have been Kant (1929). In Musgrave’s words, Kant rested his philosophy entirely on a Gem. Observing that we cannot think of things without bringing them under the categories of our thought, Kant concluded that we cannot think of things-as-they-are-inthemselves. What, then, can we think of? Why, of things-as-thought-of-byus, of course. So Kant distinguished things-in-themselves (Ding-an-sich) from ‘phenomenal things’, things-as-perceived-by-humans or things-asconceived-by-humans or things-as-thought-of-by-humans. (Musgrave 1999, 26) Kant leaves us with two worlds, but, as in the case of Berkeley, Musgrave sees Kant’s conclusion as tending almost inexorably towards a more thoroughgoing
Refl ections on the Gem
233
form of idealism, one that allows us to escape from the apparently inescapable cognitive disadvantage of not being able to know about reality as such. It is thus scarcely surprising, Musgrave thinks, that according to post-Kantian German idealists we do know about reality as such but only because this reality involves objects-as-cognized-by-us. There is no other reality. Here is how Stove and Musgrave characterize the “Gem” pattern instanced by Berkeley’s and Kant’s arguments. Given a transitive verb X signifying some kind of cognitive relation (“perceive,” “have in mind,” “know,” and so on) and a term C expressing a necessary condition for Xing things, a Gem is any argument of the form: [G] You cannot X things unless C is met. Therefore, you cannot X things as they are in themselves. (Stove 1991, 152–153; Musgrave 1999, 27) Musgrave actually uses the slightly different form of wording “You cannot X things-as-they-are-in-themselves,” where he calls the hyphenated expression a “hyphenated name” and the entities they purport to name “hyphenated entities.” Nothing hinges on this difference in wording, as Musgrave and Stove clearly agree on the most important point: that the phrases in question are to be understood as genuine nominals. Because few other writers use hyphenation, I will for the most part use the terminology of “qua-names” and “qua-entities” rather than “hyphenated names” and “hyphenated entities,” taking a qua-name to be any noun-phrase that uses a restrictive qua modifier or any variant of such a noun-phrase (including such phrases as “the atoms of Bohr’s theory of the atom”). Musgrave thinks that idealism is the inevitable upshot of an instance of [G]. For standing behind the argument is the assumption that there are things that people are able to X. Not things in themselves, of course—that is ruled out by the conclusion. The things in question can therefore only be qua-entities: things as they are under condition C or things-as-they-areunder-condition-C. Stove himself thought that idealism expired around 1940, despite Gems reaching even greater heights of popularity with the advent of such intellectual trends as cultural relativism. That is because Stove took idealism to hinge on C taking some mental condition or other as value (1991, 166). Musgrave, on the other hand, allows for more generous forms of “conceptual idealism” that include Gem-inspired versions of cultural relativism. In his words, “conceptual idealism fueled by Gems, implicit and explicit, is a dominant metaphysics of our age” (1999, 28). Philosophers in the grip of such a metaphysics no longer believe that we have knowledge of a Kantian world of things-as-they-appear, because they think this rests on too imperialistic a view of what we can know. What we are able to know is more fi ne-grained: we can only know about worlds relativized to linguistic or conceptual schemes, a form of conceptual idealism:
234
Fred Kroon There is no unique world-as-conceived-of-by-humans. The world-asconceived-by-the-Aristotelian differs radically from the world-as-conceived-by-the-Newtonian. The world-of-the-Eskimo is not the same as the world-of-the-Kalahari-bushman. (Musgrave 1999, 28)
And as Musgrave takes great delight in pointing out, it doesn’t stop there: This gets really exciting if we drop human chauvinism and bring in non-human animals too. The world-of-the-chimpanzee is different from the world-of-Albert-Einstein, and both are worlds apart from the world-of-the-honeybee. (Musgrave 1999, 28) But what, precisely, is the source of Stove’s and Musgrave’s disaffection for this “despicable” and “unspeakably contemptible” form of argument (Stove 1991, 147, 161)? The answer seems clear enough. To the extent that the premise of the Gem is some kind of conceptual truth or tautology, they think that such arguments are plainly invalid, for “you cannot validly obtain a non-tautological conclusion from a tautological premise” (Musgrave 1999, 26). As to why a conclusion like “You cannot have trees-without-the-mind in mind” is not even a covert tautology, Musgrave and Stove think the conclusion is quite obviously false, and a fortiori not a tautology of any kind. It is evident that we can have such objects as “trees-without-the-mind” in mind, as the latter are just ordinary external objects, and we attend to such objects all the time. We do it, and Stove thinks even birds do it: The feat [of having trees-without-the-mind in mind] is so far from being impossible, that every bird that alights on a tree manages it easily. (Stove 1991, 140) Musgrave concurs, insisting that the idealist talk that differentiates trees from trees-without-the-mind and the-Moon-as-experienced-by-humans from the Moon need not be taken seriously. We can just see it as a fancy way of drawing attention to the diversity of experience or thought or talk about the world. . . . The Moon-as-experienced-by-humans is just the Moon— and the same goes for all other hyphenated entities, including the Kantian Moon-in-itself. (Musgrave 1999, 28)
Refl ections on the Gem
235
On this view, then, the generalized Gem’s conclusion that “you cannot X [‘perceive,’ ‘have in mind,’ etc.] things-as-they-are-in-themselves” is not tautological because it is plainly false: we X the Moon-in-itself, for example, whenever we X the Moon. But Musgrave, unlike Stove, also toys with another reason to reject the conclusion, one that adopts an important part of what he takes to be the view of purveyors of the Gem. In this second way of understanding their argument, the conclusion is not a conceptual truth because it carries ontic commitments that are at the very least implausible, hence nontrivial and hence not entailed by the tautological premise. We don’t have the slightest reason to believe either in things-in-themselves or in things-as-falling-under-C once these are taken to be distinct kinds of entities with distinct sets of properties. The tendency to think otherwise is based on a belief in a kind of word-magic: the incredible view that describing things in certain ways, or conceptualizing things using certain concepts, creates entities meeting the descriptions and falling under the concepts. We should spurn such a word-magic, along with the hyphenated entities it brings in its wake: [T]here are no hyphenated entities. Hyphenated entities are all ersatz entities spawned by bad word-magical philosophy. . . . There is no such thing as the Moon-as-it-is-in-itself (this concept is empty), and there is no such thing as the Moon-as-it-is-perceived-by-humans (this concept is empty, too). There is just the Moon, the unhyphenated Moon, and different percepts or concepts or theories about it. (Musgrave 2001, 53) In this second construal too, there is every reason to think the conclusion of a Gem is false. Combining it with the claim that we X things-as-they-areunder-C, the conclusion would imply that things-in-themselves constitute a class of entities distinct from things-as-they-are-under-C and that the notion of a thing simpliciter, essentially unresolved into one or the other kind of quaentity, is therefore confused. Since a Gem’s premise is supposed to be a trivial truth describing a necessary condition for X-ing things, the utter implausibility of what is thus implied shows a fortiori that a Gem is an invalid argument. This may seem a comfortable situtation to be in—two ways of showing that the Gem is invalid! But there is something odd about the fact that we have been given these two reasons for rejecting Gems. For the reasons are incompatible: Musgrave’s two claims (1) The Moon-as-experienced-by-humans is just the Moon and (2) Hyphenated (or qua-) entities like the Moon-as-experienced-byhumans do not exist
236
Fred Kroon
can’t both be true. My own view is that the second reason for rejecting the Gem is the more promising. Such arguments typically assume that a quaname like “O as it is in itself” in some sense carves out a special class of entity distinct from that carved out by “O as it is experienced by humans,” say. (Intuitively, what is essential to the former are the properties O has in itself; what is essential to the latter are properties that O is experienced to have by all humans.) At the same time, I agree that there is something to the idea that the Moon-as-experienced-by-humans and the Moon-as-it-isin-itself are nothing over and above the Moon.1 Clarifying these ideas, however, is going to require some work. In the meantime, I do not propose to relitigate Musgrave’s and Stove’s rejection of Gems: to the extent that they are taken as arguments from trivial premises to overwhelmingly interesting ontological conclusions invoking special qua-entities, the interpretation presupposed by (2), they don’t deserve the time of day. But they do deserve understanding, and on that point there is rather more to be said than we fi nd in either Musgrave’s or Stove’s account. In what follows, I argue that there may be something quite compelling about certain Gems on other, arguably intended, ways of interpreting them. Beginning in the next section with a quick tour of the sort of metaphysical realism that embeds Musgrave’s and Stove’s attack, I claim that some arguments of this type, including Kant’s, may well be worth taking seriously because of the way they give appropriate expression to a healthy epistemic humility rather than an unhealthy commitment to special qua-entities. Much of the remainder of the chapter is dedicated to showing how we can resist the unhealthy ascent from qua-names to qua-entities.
2
REALISM AND THE QUESTION OF KNOWLEDGE
Stove and Musgrave see their attack on the conclusion of the Gem as grounded in their unabashed realism. Musgrave is especially forthcoming. After suggesting that the Gem-inspired use of hyphenated names encourages “a kind of conceptual or epistemic idealism—once we assume that
1 I deny that there is a strict identity, however. For that would mean that knowing O has some relational dispositional property R is sufficient for knowing that O as it is in itself is R, and this sounds quite implausible, given the close link between the way a thing is in itself and the thing’s intrinsic properties (see §3 and n. 20). Stove in particular seems not to appreciate the special meaning that “in itself” has for Kant and just treats it as a modifier that indicates a deep, arcane form of objectivity; see his gibe: “[W]e know quite a lot about numbers: yea, even about numbers ‘as they are in and for themselves’—or, as we say it in English, numbers” (Stove 1991, 160) and his insistence that a Gem with the conclusion “We can’t eat oysters as they are in themselves” is “no worse” than a Gem like Kant’s (161).
Refl ections on the Gem
237
they stand for things, that they are not empty names” (2001, 52), he makes this declaration: Metaphysical realism opposes this conceptual or linguistic idealism. Metaphysical realism says that hyphenated names are all empty names, that there are no hyphenated entities. . . . There is just one Moon, the unhyphenated one. Aristotle had beliefs about it, some true, some false. Galileo had beliefs about it, some true, some false. And we have beliefs about it, some true, and likely enough, some false. (Musgrave 2001, 53) It is hard not to symphathize with the sentiment of these comments. Musgrave is here giving voice to commonsense entity-realism: insofar as the Moon-as-conceived-by-Aristotle is supposed to be different from the Moonas-conceived-by-Galileo and different again from the Moon-in-itself, there simply are no such things—there is only the Moon, a mind-independent object about which people may have radically varying beliefs. But Musgrave also makes it clear that this kind of commonsense metaphysical view is hard to separate from what seems an equally commonsense epistemological view. In perceiving, and interacting with, entities like the Moon, we acquire beliefs that are more or less reasonable. Some beliefs count as knowledge, some as justifed beliefs that are not knowledge, some as mere beliefs. Although there is much that we don’t know, many of the propositions we ordinarily take ourselves to know about the objects we interact with are ones that we do know. Of course, there are interesting questions to be asked about the nature of knowledge, but for the most part that shouldn’t affect our conviction that there is much that we know (and don’t know). Musgrave himself is a Popperian, and he adopts a fallibilist realism on which knowledge is conjectural: true belief that has withstood serious criticism. 2 Stove—no Popperian3 —believed in probabilistic justification. Both believe in knowledge, and both are scathing about skepticism as a serious epistemological position. Skepticism can have different manifestations, however. Perhaps we should reject the kind of brain-in-a-vat type skepticism that declares that we do not know, and are not even justified in believing, propositions that imply the existence of an external world.4 But this is by no means the only kind
2
See, for example, Musgrave (1993, ch. 15). This is putting it mildly. See, for example, Stove (1982). 4 Brain-in-a-vat type skepticism appeals to the fact that there are relevant possibilities, such as brain-in-the-vat scenarios in which all such propositions are false, that no evidence can possibly rule out. David Chalmers has recently argued that such scenarios may involve genuine metaphysical possibilities about the nature of 3
238
Fred Kroon
of skepticism. Perhaps we know that there are objects in an external world and know that we interact with them but are able to have only an imperfect understanding of what they are like (call such a property-oriented form of skepticism “epistemic humility,” following Langton 1998). No doubt there are versions of this kind of position that are again worthy of quick rejection. Why suppose for a moment that I may sometimes be looking at a rock, sometimes an insect, sometimes a different kind of object, whenever the best evidence available to me tells me that I am looking at a person, on the grounds that some evil demon may have arranged for my evidence to be entirely misleading as to the nature of the objects I interact with? But there are versions of such a skepticism that can’t be dismissed so readily. And the nature of one such version has a crucial bearing on the version of a Gem that most interests Stove and Musgrave. To begin with, consider the following fanciful scenario. Jones’ sense organs tell him that certain objects he interacts with have a range of salient properties (intrinsic and extrinsic) that they don’t in fact have. Instead, his inclination to ascribe these properties depends systematically on the genuine properties of the objects he interacts with. Perhaps, for example, his environment contains strange flexible balls composed, unbeknownst to him, of certain very unusual molecules whose effects on his sense organs and other property-detecting instruments make him systematically misdescribe the real properties of the balls. He and other observers construe the balls as large and oval when the balls are really tiny and spherical; as slow-moving when they are really possessed of a fast zig-zag motion that is impossible to detect; and so on. (More radical scenarios might involve a more radical gap between the balls’ real properties and how the properties register on us.) Because of the systematic dependence of the registered properties on the properties that are responsible for the registering, this kind of scenario is in some ways rather different from the example mentioned in the preceding paragraph. In the present scenario, Jones can know a great deal about such objects, not just that they exist. Among other things, he can know that they are large, oval, and slow-moving—so long, of course, as this is taken to be code for “the balls are such that, under standard conditions, they must appear to perceiving agents as large, oval, and slow-moving.” The balls don’t really have these properties, but there is a sense in which they near-enough have them— taking as our standard for proximity the fact that, given their real intrinsic and extrinsic properties, the balls must (nomically) appear to agents like Jones as having these other properties as their intrinsic and extrinsic properties (a natural perspective, given that this is the agents’ own perspective). Put another way, Jones and others know that the balls presented to them in
the external world rather than metaphysical possibilities in which there is no external world (Chalmers 2005). If so, brain-in-the-vat skepticism may be closer than it at fi rst appears to the kind of epistemic humility I discuss in §3.
Refl ections on the Gem
239
experience are large, oval, and slow-moving, although this is not true of the balls as they really are. If this can be true for certain strange balls, it is hard to see why something like this can’t also be true for objects at large. Perhaps we are all systematically wrong about the objects we experience; in particular, wrong about their intrinsic properties, which remain hidden from us even though they have a large part to play in the way the objects appear to us. Such an epistemic humility—a form of skepticism about the identity of the properties possessed by entities in our environment—seems entirely consistent with metaphysical realism, at least in its epistemologically untainted form. 5
3
FROM KANTIAN TO “KUHNIAN” HUMILITY
Even if a skeptical scenario of this type is logically possible, however, it is still not likely to count with Musgrave and Stove as providing a good argument for epistemic humility. Perhaps we can’t evidentially rule out such a possibility, but this is not yet to allow that the mere existence of this possibility matters to claims about what we can genuinely know (else we would have to allow “evil demon” scenarios to rule out knowledge of the external world tout court). Musgrave would simply reject the kind of epistemology that gives merely possible scenarios such a decisive status, and so would Stove. There is, however, another way of understanding the bearing that such possibilities have on skepticism, and it is one that yields a rather better philosophical argument for the claim that we are ignorant of (especially) the intrinsic properties of the objects we interact with. Perhaps we can’t know a thing’s intrinsic properties because there is an independent epistemic barrier to knowing such properties and not simply because we can’t evidentially rule out a possible situation in which we get these properties wrong. Here I have in mind the kind of argument Rae Langton has recently provided for a new way of developing a one-world interpretation of Kant’s phenomena–noumena distinction (Langton 1998). According to Langton’s Kant, our deep and irremediable ignorance of noumena is not ignorance of mysterious entities forever hidden from us behind a veil of appearances (which in turn are the only objects we humans can know about). What we are ignorant of are the intrinsic properties of substances, properties that do not depend on
5 “Untainted” in the sense of not requiring that agents have the capacity to gain knowledge, or even justified belief, about the properties they might most like to know about, including the intrinsic properties of things. This is still consistent with a moderate epistemological element in one’s account of realism (e.g., the kind of account preferred by Pettit 1991).
240
Fred Kroon
there being any other substances.6 Langton’s Kant holds that substances are the independently existing things that we humans interact with, but they are not things that we can know about as they are intrinsically (that is, as they are in themselves, without regard to the existence of anything else).7 We can only know about such things as bearers of causal powers, and because causal powers are irreducible for Kant, acquaintance with a thing’s causal powers will not give us acquaintance with its intrinsic properties. The properties we think of as intrinsic are at bottom relational and hence extrinsic. It follows that Langton’s Kant is not an idealist but a realist: a realist who counsels us to accept ignorance concerning the intrinsic properties of the independently existing things that we interact with when we experience the world. Langton labels this Kantian form of epistemic humility Kantian humility.8 This interpretation is, of course, in direct confl ict with Stove’s and Musgrave’s indictment of Kant as the supreme Gem-crafter. To the extent that their indictment of Kant is based on the classical two-worlds interpretation of the fi rst Critique, it is not hard to predict how they would have responded. They would have pointed to the many passages in the fi rst Critique that appear to treat the term “thing in itself” as a term for a distinctive kind of entity—passages in which things in themselves are declared to be utterly unknowable (e.g., Bxx, A30/B45, A44/B62), not located in space or time (e.g., A26/B42, A30/B45), radically distinct from objects as they appear or appearances (e.g., A491/B519 ff), and so on. Musgrave, I might add, enforces such a two-world interpretation of Kant by consistently hyphenating the phrase “thing in itself,” something
6 Although Langton develops this interpretive view in far more detail than anyone else, she is scarcely alone in holding it. See, for example, Baldner (1988). Van Cleve suggests that C. I. Lewis and even J. S. Mill maintained such an interpretation (van Cleve 1999, 150ff.) 7 Langton argues (persuasively, in my opinion) that Kant’s notion of the properties that a thing has in itself (an Sich) is best understood as the notion of the thing’s intrinsic properties, to use contemporary parlance. On the latter notion, see Langton and Lewis (1998). 8 Langton acknowledges that Kant makes claims that seem to steer him close to idealism—just think of Kant’s view that space is ideal. The way she sees Kant grappling with this issue leaves her with an account of the ordinary objects we encounter in experience that is more than passing strange but not more so than the one we get from the usual two-worlds view. As Langton describes the view in her book’s fi nal sentence:
the ideality of space does not, in [Kant’s] opinion, undermine the reality of what is presented in space: namely bodies, constituted by forces that are mind-independent properties of absolutely independent substances, things in themselves. (Langton 1998, 218)
Refl ections on the Gem
241
that Kant himself rarely does.9 But hyphenation is scarcely necessary: there are many passages that seem simply unintelligible unless we adopt a two-worlds interpretation. Remember too that Kant even introduces the separate and unstructured terms “phenomena” and “noumena,” as if to emphasize the utter separateness of these two kinds of entities.10 Not only would Stove and Musgrave almost certainly have rejected Langton’s interpretation of Kant,11 it is also clear that they reject epistemic humility concerning our knowledge of intrinsic properties as a viable epistemological thesis in its own right. For Langton’s Kant, the receptive nature of human cognition provides the deep reason for our inability to know things as they are in themselves. There is nothing in Musgrave’s or Stove’s epistemological outlook that suggests receptivity of this kind should be seen as a barrier to knowledge of the intrinsic properties of things. One imagines that for Musgrave hypotheses concerning the intrinsic nature of substances might survive as rigorous a regime of testing as that confronting any other hypthesis. For Stove, they might be as highly confi rmed as any other hypothesis. Both would think the appeal to receptivity was a covert appeal to a Gem and hence deeply fallacious. Langton disagrees. She sees Kantian humility, or something close to it, as something that hovers near the surface even in familiar contemporary manifestations of metaphysical realism. Many metaphysical realists hold that things affect other things in virtue of their intrinsic properties. But a thing’s intrinsic nature needs the collusion of its dispositional properties if it is to bring about a response: the thing is disposed under certain circumstances to bring about the response because its intrinsic nature is such that it would under these circumstances, and in accordance with contingent laws of nature, bring about the response. And if that is so, then far from denying the thesis of humility we are giving it tacit support:
9 Kant sometimes hyphenates the phrase, but this tends to occur only in the context of views he is trying to distance himself from, such as Leibniz’s version of transcendental realism (e.g., A264/B320). It is common for commentators to hyphenate the phrase when it occurs in the index of a book, presumably because in the absence of a context the role of the qualifier “in itself” is not clear (in some ordinary contexts, its role is very clearly adverbial, as when we say “The book in itself was a pleasant enough read, but knowing the author spoiled the book for me”). Hyphenating the phrase can make it clear that the phrase is used as a genuine nominal, a qua-name. Virtually all English translations of the fi rst Critique, including Kemp-Smith’s, contain “things-in-themselves” as an entry in the Index. 10 See especially A249ff and B306ff. (I have in mind the notion of a noumenon in its negative rather than its positive epistemic sense.) 11 For Musgrave’s reading of Kant, see Musgrave (1993, ch. 12).
242
Fred Kroon [If] the ground is distinct from the power, and contingently connected with it, then our orthodoxy is faced with a conclusion surprisingly similar to Kantian Humility. . . . [Our name for the ground] becomes the name for a something-we-know-not-what—ominously similar to a Kantian thing in itself. Our contemporary orthodoxy must concede that it, too, is faced with a conclusion similar to Kant’s—that there are intrinsic features of the world with which we can never become acquainted. (Langton 1998, 176)
Langton’s Kantian point can be directed against Stove and Musgrave as follows. There is a flaw in both the Popperian idea that certain hypotheses about intrinsic natures should be accepted because they survive a regime of rigorous testing and the inductivist idea that they should be accepted because they are confi rmed on the basis of inductive procedures. Such attempts at testing and confi rming are simply unable to show anything about intrinsic properties—they can only show something about relational properties grounded in intrinsic properties whose identity is kept hidden. So we should espouse epistemic humility about such properties.12 Note, once again, that Langton’s point does not rest on the kind of epistemology that focuses on logical possibilities consistent with the evidence. It is thus rather different from the kind of argument for a version of humility (Ramseyan humility) recently given by David Lewis.13 A Kantian humility rests on the
12 Note that such humility is bound to affect certain extrinsic properties as well (as we already saw in the case of Jones, where Jones’ inability to know the intrinsic properties of certain strange balls also affected his knowledge of relational properties like distance). Here is a trivial example. If P is an unknowable intrinsic property of substance X, and Q is a knowable extrinsic property, then the property of having both P and Q is an unknowable extrinsic property of X. For Kant, of course, we only know a thing’s extrinsic properties to the extent that they refl ect a relation to human cognition. 13 See Lewis (2009, circulated as a preprint since 2001). Lewis argues that being the ground of a certain disposition is a case of role-occupancy. But there are a variety of occupied roles, and more often than we realize we only know of the properties of things as the occupants of such roles. In particular, on the assumption that each theoretical term in the fi nal, true theory of the world T is defi ned on the basis of the theory’s Ramsification, much of what T says about the fundamental properties of entities could be realized by fundamental properties other than the ones featuring in T’s actual realization; we could pick any of them, and be none the wiser. And if humility applies to most or all of the fundamental properties, then the fact that they provide a minimal base on which all else supervenes suggests that Humility applies to a great range of less than fundamental properties, both intrinsic and extrinsic. Indeed, thinks Lewis, we can’t even identify the various properties that
Refl ections on the Gem
243
putative existence of independent barriers to knowledge and justification, not on the mere existence of logical possibilities. There are other skeptical doctrines that acknowledge that the world contains independently existing objects many of whose properties we don’t know. If you think that there is one world and one true theory of this world (perhaps formulable in a variety of ways), but if you also think that there are other equally well-supported and equally plausible alternative theories of what this world is like, compossible with all possible evidence but positing different (kinds of) objects and/or attributing different properties, then you too believe in a form of humility; for in this view there can be no neutral standpoint from which to privilege one such account of the world above the others. Those who believe that scientists may well work with different paradigms that do not permit the epistemic resolution of cross-paradigm differences (somewhat in the manner of Kuhn, say) might then be said to subscribe to such a humility. I’ll describe this as “Kuhnian” humility, although the scare quotes around the name “Kuhn” are there to remind us that the early Kuhn seems to have maintained a rather different doctrine: the radical view that scientists somehow create their own worlds (Kuhn 1970, 150, passim). The latter is an ontological rather than an epistemic doctrine and, as we saw earlier, one that Musgrave takes to be a form of conceptual idealism. It claims that scientists inhabit theory-dependent worlds containing (as Musgrave sees it) not atoms and stars simpliciter but such strange entities as objects-as-they-are-in-Newtonian-Mechanics and atoms-as-they-arein-Bohr’s-theory-of-the-atom.
4
HUMILITY AND THE ROAD TO QUA-ENTITIES
It is time to face some questions that have already surfaced (or are close to the surface). The fi rst one involves Kant-interpretation. Langton makes a good case for the thesis that Kant espouses humility and a one-world metaphysics, but how is this interpretive thesis consistent with the clear evidence (as Musgrave and many other Kant commentators would see it) that Kant uses the phrases “thing in itself” and “thing as it appears” as genuine referring expressions that apparently mark out two very different kinds of entities? The second question involves a puzzle concerning the kind of humility
we are thus unable to distinguish; for we don’t, and can’t, know these properties other than as the possible occupants of various roles implicitly described by our fi nal true theory T. But as Langton (2004) notes, Lewis’s version of humility thus seems to rest on our inability to rule out a certain skeptical scenario on evidential grounds—an inability that Lewis’s own well-known brand of contextualism has its own way of disarming (and rearming).
244 Fred Kroon we pretended to ascribe to Kuhn. Humility is a radical epistemic doctrine, and Kuhn and many of his followers are instead quick to use phrases like “objects as they are in Newtonian Mechanics” and “the atoms in Bohr’s theory of the atom”: genuine referring expressions, so it seems, that apparently mark out different kinds of entities, thereby suggesting a radical ontological doctrine rather than a radical epistemic doctrine. Given the option of humility, what could possibly have recommended this alternative way of presenting their insights? The answers to these questions are, I believe, linked. In my view, declarations of humility are occasions that make it especially tempting to adopt a mode of speech incorporating noun-phrases that purport to refer to a plurality of special entities and worlds—despite the fact that humility itself officially declares that there is no such plurality, only ignorance. To understand the attractions of this mode of speech, suppose we wanted to describe the epistemic situation faced by Jones in our earlier example from the perspective of Jones himself. It would then be odd to declare that the balls that Jones encounters seem large and oval to him, even though they are really small and round. From Jones’ perspective, the use of words like “seem” is simply inappropriate (it reminds one of the bad old idea that the right language for describing one’s experience of the world is a phenomenalist language). Given the robust disposition of the balls, the balls are large and oval for Jones. One feels like saying “What he sees are large, oval balls. That is how they are in his world.” Alternatively, one might say something like “The balls as they appear to Jones [the balls of Jones’ world] are large and oval, and very different from the balls as they really are,” now talking as if we are contrasting certain intrinsic properties of one set of balls with the intrinsic properties of a different set of balls. Claims of this kind are an attempt to describe things and their properties from two perspectives. The fi rst one focuses on the manner in which the real intrinsic and extrinsic properties of things register on agents like Jones; it redescribes what are, at bottom, relational response-dependent properties (insofar as they involve the responses of agents like Jones) as the response-independent properties of things “as they appear.”14 The second one focuses on properties independently of the way they register on agents. Such a Janus-faced way of talking, so I have suggested, is almost inevitable if one wants to combine epistemic humility with a phenomenologically honest way of characterizing agents’ interaction with the world. If so, it is scarcely surprising to fi nd Kant using the phrases “thing as it appears” and “thing in itself” as genuine nominals when emphasizing
14 Thus, being an x such that the intrinsic nature of x results in x appearing to have P as an intrinsic property gets counted as (intrinsic) property P, now taken as an intrinsic property of a thing as it appears.
Refl ections on the Gem
245
humility. Far from being incompatible with the interpretive suggestion of Kantian humility, such a use is just what we should expect. Turn now to the question I raised about “Kuhnian” humility. Suppose that “Kuhnian” humility applies to a community of scientists C in their quest to fi nd the correct theory of the world: they do not, and cannot, know which of various equally plausible fi nal theories of the world is true. But that shouldn’t prevent them from articulating one particular theory as the true theory. Perhaps there is some “paradigm” T (to use Kuhn’s term), comprising methodologies, heuristic rules, and principles, that they prefer to others because they prefer its mix of plausible-making features to that of others; and suppose they rationally pursue this paradigm in their attempt to discover the true fi nal theory of the world. We might then talk of certain claims being justified or (for Popperians) corroborated for this community C, where justification and corroboration are now understood as relativized to the methodologies prescribed by T. But putting things this way is hard to square with a belief in humility. Pressed to articulate which claims have been justified or corroborated in this way, against a background of humility concerning C’s ability to identify the actual true fi nal theory of the world, it is now tempting to say the following kind of thing. What has been justified or corroborated, and so what we, following the work of C, have good reason to believe, is not, for example, that objects simpliciter have some property P, but only that objects as they are in T (more generally, Φ’s as they are in T) have property P. We might even be said to know that objects as they are in T have P, roughly on the grounds that we know that if T were true (which is something we cannot know, given humility) objects would have property P. By contrast, suppose another community of “knowledge” holds fast to a belief in T´ rather than T, and, following their work, suppose we know that if T´ were true objects would have property Q, not P. We might then be said to know that objects as they are in T´ have property Q, not P. And, to honor humility, we might even add that no one can know whether objects as they really are have property P or Q. Such formulations strike me as both natural and appropriate if we want to combine “Kuhnian” humility with a phenomenologically honest way of characterizing a community’s epistemic practices (in particular, the sense its members have that the beliefs they acquire when engaging with the world on the basis of T may be both hard-won and eminently reasonable and not an appropriate locus of skepticism). In short, it is natural to use qua-names like “thing as it appears” / “thing in itself” and “object as it is on theory T” / “object as it really is” when we want to combine a thesis of humility with descriptions that capture the phenomenology of an agent’s epistemic engagement with the world. But now, of course, we have all the ingredients for trouble. My emphasis on the way a thesis of humility encourages their use suggests that this use does not intentionally signal a commitment to classes of special
246 Fred Kroon qua-entities. Intentional or not, however, how do we respond to the claim that such a commitment is exactly what we do get and that the semantics of qua-names leaves us no other choice? My suspicion is that many of those who accept Kuhnian relativism sense intuitively that this is indeed the case—that the natural way in which we talk of agents’ interaction with “their” world (including a willingness to admit to their having knowledge of this world but not of the world as it really is) does indeed carry these striking ontological consequences. I also think this is at least part of the reason why philosophers jump so readily to a two-world interpretation of Kant, and why devotees of Kuhn jump so readily to Kuhnian relativism as the right interpretation of what might have begun as an epistemic relativism (“Kuhnian” humility). If so, it may also be part of the reason why certain Gems have proved so attractive to so many philosophers.15
5
BACK FROM THE BRINK
Musgrave, of course, shows scant patience with the use of such phrases, especially in hyphenated form and especially as used in Gems. He thinks that so used they bear witness to a belief in word magic. But there are also indications that he thinks they often enshrine a misleading (and to that extent inappropriate) way of saying something that metaphysical realists also accept: that they are “philosopher’s gobbledy-gook” for something harmless. So far, I have argued that their use is appropriate rather than inappropriate, given the attractions of a linguistic vehicle that captures an epistemic agent’s felt epistemic engagement with the world. But I also think—here agreeing with Musgrave—that this use harbors no genuine commitment to forms of relativism or conceptual idealism. Metaphysical realists could accept such talk; they could even accept it in hyphenated form, so long as they saw the use of hyphens as a device that simply underscored the name-like nature of the phrases. They could accept it if they saw it as merely useful pretense. Here is why. We often pretend that the world is a certain way. In Ken Walton’s influential development of this idea, children as well as adults play games of make-believe on the basis of props that mandate that they imagine certain things (Walton 1990). Thus a children’s game may require its participants to imagine that a certain oddly shaped stump is a bear, that actions done to the stump are acts against the bear, and so on. Another sort of game, indulged in by children as well as adults, involves reading or listening to a
15 I certainly don’t see it as the whole reason, and perhaps it doesn’t apply to all Gems. Like Stove, however, I fi nd it hard to think of reasons that are even remotely compelling, and to that extent this one seems to me to have the edge. (For Stove’s attempt to understand the popularity of Gems, see Stove 1991, 161ff.)
Refl ections on the Gem
247
story, a game that requires its participants to imagine that certain events really happened (that there really was a famous detective called “Holmes” who lived in Baker Street, for example). In Walton’s terminology, something is fictional or true in a game of make-believe when it should be imagined, given the rules regulating the pretense (Walton 1990, 39ff.). Children playing games and readers of novels are typically caught up in internally oriented pretense (what Walton 1993 calls “content oriented” pretense). As participants, they are interested in the world of the fiction and its goings-on. But not all pretense is like this. In externally oriented pretense we pretend that the world is a certain way, not in order to pursue the imaginative thought that the world is that way but to say something about the real world that provides the occasion for our pretense. The realworld circumstance that involves Johnny’s putting a rope around the stump makes it fictional that a bear has been lassoed. The latter fictional truth is what interests him in this piece of internally oriented make-believe. But it is not what interests his mother, for example, as she tells him that “this hunter gets no dinner if he doesn’t leave the bear alone for a spell.” Mother is involved in externally oriented make-believe, affi rming the real world circumstance that makes her claim true in the make-believe—the fact that Johnny will get no dinner if he doesn’t stop playing with the stump. The bear in the children’s game and Holmes in Conan Doyle’s novels don’t exist. In our pretense, we “do as if” there is something answering to the terms “the bear” and “Holmes,” even though we fi rmly believe there isn’t. But a lot of pretense involves objects in the real world. Such is the case, I claim, with qua-names.16 In their case the pretense is again externally oriented. To begin motivating this account, consider first a singular instance of such a name: a particular homely example, to show, yet again, how common locutions of this kind are, and how little we should fear them. Suppose Anne claims that (3) Jones as he was ten years ago was trustworthy, unlike the Jones of today. According to the view I shall advance, Anne’s use of the qua-name “Jones as he was ten years ago” indicates her involvement in a piece of what we might call constitutive pretense. Anne assumes a way of counting people that is different from the way we standardly count them. Whereas people
16 Although I don’t have space to defend this view of qua-names in detail, I believe it has important advantages over some of the other proposals in the literature, for example, the intensionalist account of Fine (1982) or the counterpartoriented account of Lewis (2003). For one thing, it extends to expressions that don’t have the usual form of qua-names, such as “Langton’s Kant,” “the later Wittgenstein,” “the atoms of Bohr’s theory of the atom,” and so on (Kroon 2001).
248
Fred Kroon
are organisms that exist over time and whose qualities may vary radically over time, we sometimes adopt the graphic pretense that they are “cut up” differently. We assume different ways in which they may be constituted with respect to the passage of time, thereby letting us assume a plurality of people where there is in fact just one. We don’t really accept such a view of persons, of course (that is why we keep the real world in focus by talking of Jones, as he was ten years ago), but in the scope of our pretense we do (that is why we keep the real world at bay by talking substantivally of Jones-as-he-was-ten-years-ago or the-Jones-of-ten-years-ago). According to the position I am defending, we sometimes make claims about persons, their qualities, and people’s attitudes towards them from within the scope of this kind of pretense.17 Suppose, then, that we accept the role of pretense in the utterance of sentences involving singular qua-names. Now ask: what makes (3) fictional or true in the speaker’s pretense—an imaginative set-up in which Jones is constituted first by the properties he had ten years ago and next by the properties he has now? The answer seems straightforward. What makes (3) true in this pretense is simply the fact that Jones was trustworthy ten years ago but isn’t so now. That is the perfectly ordinary fact that gets reinterpreted by (3) in terms of the different qualities possessed by two different time-bound versions of Jones. If we now take the pretense in question as externally oriented in the sense encountered earlier, we can say that Anne’s utterance of (3) is used to affirm the circumstance that makes (3) fictional and hence that Jones was trustworthy ten years ago but isn’t so now. The same kind of account is available to deal with claims like “I admire Jones as he is portrayed in the movie but can’t stand the real Jones [Jones as he really is].”18 So there is a way of counting sentences like (3) true or false (not just true or false in the pretense, but really true or false), namely by taking the speaker to be affi rming the circumstance that would make the sentence true in the pretense. That way we get truth conditions for sentences containing (singular) qua-names that involve us in no special ontological
17 Of course, the claim can’t be that speakers consciously understand such a sentence as make-believe or pretense or that it involves speakers in the kind of richly imaginative pretense characteristic of reading novels, say. Imposing such requirements would be misplaced. First, the “conscious articulation” requirement is implausibly strong. It is enough that a pretense account offers the best overall account of our linguistic behavior (including inferences we make from, and to, such a sentence). Second, we shouldn’t think that pretense is always richly imaginative. Pretense may be no more than a thin “making as if” the world is a way that it need not be, whatever our purpose. There is even less reason to expect conscious articulation and richly imaginative pretense when the pretense is externally oriented. 18 See Kroon (2001), which presents an earlier version of these ideas, especially as they apply to Kant.
Refl ections on the Gem
249
commitments to things like temporal parts, fictional counterparts, and so on, even though in uttering such sentences we choose terms that are genuine nominals and that genuinely pick out strange time-bound or fictionbound entities—but do so only from the point of view of an expressively useful piece of pretense. But constitutive pretense needn’t only involve singular terms. For a start, stories we treat as told fact may include not only singular fictional terms like “Holmes,” but also general fictional terms like “[slithy] tove.” In their case, treating the stories as props in games of make-believe involves pretending that the terms denote genuine kinds. Qua-names too may be general rather than singular, and in their case the pretense approach again advocates using the idea of constitutive pretense. Turn fi rst to Kant and his talk of “things in themselves / things as they appear.” The pretense view offers the following perspective on such talk. In talking of “things as they appear,” we are involved in constitutive pretense to the effect that things are as they appear to be; more precisely, that the properties we judge them to have on the basis of the proper exercise of our human cognitive capacities really are their properties.19 Kant, as we have seen, believes we have no right to think that the real intrinsic properties of things include any of the properties we thus ordinarily ascribe to things, as he takes our ordinary ascription of properties to be grounded in certain (relational) response-dependent properties of things that give us no insight into their real intrinsic properties. But he also thinks that there is virtue in pretending otherwise. It provides a way of mapping the philosophically unreflective, unrelativized way in which we usually describe the world onto locutions that remain unrelativized but that at the same time affi rm the response-dependence of the properties thus ascribed to things—and all without generating a commitment to special qua-entities, because these locutions are all uttered from within the scope of the pretense that things are as they appear to be. Note, furthermore, how very natural a piece of pretense this is, mirroring as it does the way we naturally, if unreflectively, understand our engagement with the world. Consider, for example, the claim that things as they appear are (necessarily) extended in space and time, fall under physical laws, have the ability to interact causally with other things, and so on. And assume that Kant is right in thinking that we are constrained to experience the world in terms of such properties. From the perspective of the pretense that things are as they appear to be, it now turns out that these properties
19 Note that “as they appear” refers to the stable product of general human cognitive capacities. The present way of describing Kant’s views does not compromise the primary–secondary quality distinction (which Kant accepts), as he thinks that secondary qualities lack intersubjective stability and so can’t be classed as genuine properties of things as they appear (A29/B45).
250 Fred Kroon are indeed genuine intrinsic and extrinsic properties of things, possessed in a way that owes nothing to the cognitive powers of humans. Just how it also succeeds in affi rming the essential cognition- or response-dependence of such an ascription of properties is a function of the externally oriented nature of the pretense. To say that “things as they appear are (necessarily) such-and-so” is to affi rm the circumstance that makes this statement true in the pretense. But this is just the circumstance of our being constrained to experience things as being such-and-so. So what we manage to assert is that humans are constrained to experience things as being such-and-so, thereby affi rming the response-dependence of the property of things that underlies our ascription (in the pretense) of the property of being such-and-so. Kant also claims that things in themselves, unlike things as they appear, are utterly unknowable. On the pretense account, to say that “things in themselves are unknowable” is to revert to an equally graphic piece of constitutive pretense and then to reason from its implicit commitments—the pretense that the only properties a thing can truly be said to have are its intrinsic properties, properties that depend on nothing “outside” of those things. To pretend in this way equally abstracts from the facts; it is to assume a particularly sterile account of property possession. (A thing’s extrinsic properties, including dispositional properties that reflect the way it affects humans, are no less genuine properties of the thing.) But as a piece of externally oriented constitutive pretense there is nothing wrong with it. In saying that “things in themselves are unknowable,” speakers participating in this pretense in effect affi rm that things are unknowable in respect of their intrinsic properties and hence that nothing is known about the intrinsic nature of things. 20 Turn now to the other case that interested us: the use of qua-names like “physical objects as they are according to Newtonian mechanics,” “the Moon as conceptualized by Aristotelians,” “atoms as they are in Bohr’s theory,” and so on (as well as such variants as “Newtonian objects,” “the atoms of Bohr’s theory,” etc.). Relativist–constructivists, strong sociologists of knowledge, and the like, think it is clearly correct to make claims like the following:
20 Or, at any rate, that nothing intrinsic is known about the intrinsic nature of things. It is, of course, the intrinsic nature of intrinsic properties that Kant took to be unknowable. We do know some of the extrinsic properties of intrinsic properties, as we know the way they impact on us. Continuing the externally oriented pretense of two worlds that are differently constituted, we might even formulate this latter clarification in substantival terms. As Kant put it, “nothing that belongs to a thing in itself can be found [in mere appearance]” (A45/B62), thereby affirming that the nature of the intrinsic properties underlying a thing’s cognition- or response-dependent properties given in experience is not recognizable through these properties.
Refl ections on the Gem
251
(4) The objects of Newtonian mechanics are radically different from objects as conceptualized by Aristotelians, and the instruments of Newtonian experimentalists were designed to determine facts about the former, not the latter. They think that these are clearly different kinds of entities: the result of different productive forces that leaves the fi rst kind of objects with radically different properties from the second. 21 What the pretense view claims is that this confuses pretense with genuine truth. According to the pretense account, we should simply acquiesce in the “clear truth” of (4), while resisting the reifying move that tolerates literal talk of different theorists living in different worlds. Properly construed, talk about the difference between the objects of Newtonian mechanics and objects as conceptualized by Aristotelians is in itself no more ontologically threatening than the claim that Jones as he was ten years ago is different from the Jones of today. Motivated, as I see it, by the attractions of a mode of speech that captures the phenomenology of the theory-mediated engagement of theorists with the world, speakers adopt the graphic pretense that the entities they are talking about are theory-bound kinds constituted by the intrinsic and extrinsic properties these entities possess according to one or another theory. As in our earlier cases, this is externally oriented pretense, with (4) being used to affi rm the perfectly ordinary proposition that Newton’s theory attributes radically different properties to objects from those attributed by Aristotelians, and that, in attempting to determine particular properties of objects, Newtonian experimentalists used instruments relying on Newtonian theory. 22 Properly understood, then, the “clear truth” of (4) gives no succor to constructivists.
6
THE GEM RECONSTITUTED
I have suggested that a philosophical willingness to use the language of “things in themselves,” “things as they appear,” “X as it is in theory T,” and so on, in the context of a Gem, is not per se a sign of deep philosophi-
21 See, for example, the work of the sociologist Karin Knorr-Cetina (e.g., KnorrCetina 1993). For a persuasive critique of constructivism in its various manifestations, see Devitt (1997). 22 It may be thought that this move is hard to reconcile with the thought that theoretical terms are defi ned in terms of their place in a theory, say in the Ramsifying manner defended by David Lewis (1970). But even Ramsifiers allow that terms may have overlapping meanings, given in terms of shared postulates. That is all I need to give content to the idea of pretending that atoms are as Dalton took them to be, for example, or that they are as Bohr took them to be.
252 Fred Kroon cal confusion (or “idiocy,” to use one of Stove’s descriptors) and is almost certainly not in the case of perhaps the greatest crafter of a Gem, namely Kant. I have also argued that, contrary to Musgrave, it is not the case that once you indulge in such talk, “impeccable philosophical argument will lead you to a multiplicity of (putative) entities” (2001, 50). 23 Such talk reveals muddleheadedness only when those who use it misconstrue it as serious talk about such a multiplicity of entities—perhaps because they take the initial air of naturalness that attaches to such talk as showing that they have reasonably committed themselves to such entities. It shouldn’t be so construed. It is through-and-through pretense-talk—such entities exist only from the perspective of the pretense. At the same time, however, this engagement in pretense allows speakers to affi rm serious claims about the real, nonpretend world. Edgy tone aside, Musgrave’s claim that “[A] profundity such as ‘The Moon-as-conceived-of-by-the-Aristotelians was perfectly spherical’ is just philosopher’s gobbledy-gook for ‘Aristotelians thought that the Moon is perfectly spherical’” (1999, 28) may thus not be so far off the mark. Such an appeal to pretense even yields a sense in which Musgrave’s (1) “The Moon-as-experienced-by-humans is just the Moon” and (2) “Hyphenated entities like the Moon-as-experienced-by-humans do not exist” are both true. When pretending that there is such a thing as the Moon-as-experienced-by-humans, we are pretending something about the Moon; no other entity is involved. That is the sense in which (1) is true. When we step outside of the pretense, there is nothing that corresponds to our use of a qua-name like “the Moon-as-experienced-by-humans.” That is the sense in which (2) is true. Most importantly, the pretense account yields an interpretation on which at least some Gems deserve to be taken seriously. (Not all. Not Berkeley’s, for example.) The Gems I have in mind are the ones that focus on epistemic humility, and, so I have argued, are naturally prone to feature qua-names. Take the following variant of the Gem-schema [G] introduced in §1, and let X range over epistemic relations X such that X-ing something is able to yield knowledge or at least justified belief about that thing:
23 Musgrave is actually talking of phrases like “the table of common sense” and “the table of physics,” but he clearly intends what he says to cover qua-names in general. I am here assuming that “lead[ing] . . . to a multiplicity of [putative] entities” means leading to a serious, nonpretended commitment to such entities. I accept, of course, that in the scope of the pretense such phrases do stand for “a multiplicity of . . . entities,” but nothing philosophically pernicious follows from this. (Almost as if he recognizes this involvement of pretense, Musgrave suggests that the way out of a commitment to special hyphenated entities “is to insist that the Kantian hyphenated names are empty names, that the hyphenated entities they pretend to be names of do not exist” (Musgrave 2001, 51; emphasis mine).
Refl ections on the Gem
253
[G´] You cannot X things unless these things have the power, given their intrinsic properties, to generate appropriate X-style responses in you on the basis of which you form beliefs about them. This being a necessary condition for X-ing things, and because such powers are contingent, you therefore cannot X things in themselves—you can only X things as they appear. Then the kind of consideration pointed to by Langton in her comments on Kantian humility suggests that, suitably revised, [G´] may well harbor a compelling argument—so long as we combine it with something like the deflationary pretense interpretation of the qua-names “things in themselves” and “things as they appear” suggested in the previous paragraph. Even a Kuhnian Gem that has as its conclusion the claim that we can’t X things as they really are, but only things as they are conceptualized by some paradigm or other, has more going for it than Musgrave and Stove allow. Suppose, for the sake of argument, that there is good reason to accept the thesis of “Kuhnian” humility. Then, so long as the qua-names “things in themselves” and “things as they are conceptualized by paradigm T” are given their pretense interpretation, it is no less reasonable to accept the constructivist-sounding conclusion. What is not reasonable, either here or in the case of [G´], is to accept the serious qua-entities-importing interpretation of the conclusion in terms of which Musgrave and Stove formulate their rejection of Gems.
13 On the Sortal Dependency of Individuation Thesis Paul Snowdon
In Sameness and Substance Renewed David Wiggins aims to present the main strands of his strongly interconnected metaphysical ideas in as clear and direct a way as possible, having progressively refi ned and refocused them in the course of three major books. At the center of this web is a thesis which Wiggins calls the thesis of the sortal dependency of individuation (2001, 55), though at times he seems prepared to call it the sortal dependency of identity (2001, 23), and which he usually refers to as D.1 My aim here is to determine, insofar as I can, what this thesis actually claims and whether it is a thesis (or, perhaps, if there is any ambiguity in its content, a group of theses) that we should accept. The spirit of my engagement with the results of Wiggins’ reflections is one of considerable admiration tinged with a certain degree of skepticism, and what has emerged here will, I hope, at least be of interest to others who are, like myself, moved to think seriously about his body of work.
1
THE D-THESIS
When Wiggins states D for the fi rst time in a relatively full way, in Chapter 2 of Sameness and Substance Renewed, it is formulated as follows: a = b if and only if there exists a sortal concept f such that (1) a and b fall under f;
1 My attribution to Wiggins of a preparedness to employ the name “sortal dependency of identity” may be reading into a particular sentence more than is meant to be (hence, is) there. Nothing, of course, depends on this attribution. I wish myself to rename Wiggins’ central thesis “D” and henceforth refer to it as that, rather than as “D.”
On the Sortal Dependency of Individuation Thesis
255
(2) to say that x falls under f or that x is an f is to say what x is (in the sense Aristotle isolated); (3) a is the same f as b, that is coincides with b under f in the manner of coincidence required for members of f, hence congruently . . . (Wiggins 2001, 56) Taking this formulation of D as the guide to its general character as a claim, it is natural to comment on its significance as follows. D advances a claim about the necessary and sufficient conditions for the truth of any identity proposition or claim. As it stands, D is, strictly speaking, an unmodalized biconditional, but we can, surely, take it that Wiggins is really advancing it as a necessary truth about identity. Now one thing that stands out about D as formulated is that it says nothing about our procedures for establishing identities, nor does it say anything about our ability to grasp or understand identity propositions. Hence, if “individuation” stands for something cognitive or psychological that we do, or, maybe, achieve, as I claim that it in fact does, then it seems that D actually says nothing about individuation at all. So there is some inclination to register at least mild surprise about the name Wiggins assigns to it. Rather, I think, D says something about the conditions for the obtaining of an identity and so qualifies as what I would be inclined to call a “metaphysical thesis” about identity. At this point, there is a question that merits expression about the scope of the metaphysical thesis. Is it meant to be a thesis about identity in general or a thesis about identity in the domain of, as Wiggins puts it, “continuants, including living substances and other substances” (2001, 1)? As Wiggins repeatedly emphasizes that he is concerned with identity as a highly general and irreducible category, there is considerable pressure to read the thesis as a completely general one. However, he does direct the reader’s attention to continuants, and the third clause of the D thesis introduces the idea of coincidence, which naturally attracts a spatial reading. I leave this undecided but, to avoid potential unfairness, I restrict discussion to the domain of continuants. I want to add two things here, the fi rst reinforcing the idea of the D-thesis as metaphysical and not psychological or cognitive. (i) The central idea of D as formulated is that for the identity to obtain, a and b must fall under a sortal concept. How are we to understand the notion of a sortal concept? In Section 5 of his preamble, Wiggins makes clear that by a “concept” he does not mean something that is, as it were, in someone’s mind or an ingredient in their thinking or categorizing; but it is a way that things are or can be. In Wiggins’ way of speaking concepts exist, or are there, so long as there are things of a relevant type, and their existence does not depend on there being thinkers thinking about them or latching onto the type in question. Thus, Wiggins says: “The concept horse is not then an
256 Paul Snowdon abstraction such as horse-hood or horse-ness (whatever these are). It is something general, or better, universal; and to that extent it is philosophically contentious” (2001, 10). Later he says: “ . . . the reference of a sortal predicate such as horse i.e. what ‘horse’ stands for, is the concept horse, the general thing horse” (2001, 79). The concept of a horse is then something general which exists, or is present, in the world, and to which we can refer, and to which it is the role of certain predicates to refer. As Wiggins also says: “But horse or mammal or carnivore surely are things that we need to speak of or quantify over, in metaphysics and in science” (2001, 10). Perhaps it fits Wiggins’ notion of a concept to describe a concept as a sort or type there is in the world. A sortal concept is then one special type of existent types. It does not matter at this point how the term “sortal” is more precisely explained because it is now clear that what the sortal dependency thesis is claiming is that for an identity judgment to be true the objects in question must belong to a certain sort of sort. The object does that (or is that) by simply being a certain way, quite independently of how it is thought of, or conceived of, by anyone. In this sense of concept, no one possesses them, and no one lacks them; they are not acquired or applied; and no mind is necessary for them to be there. Wiggins calls his use of “concept” Fregean. The cognitive correlate of a concept Wiggins calls a conception. That is, so to speak, a mental focusing on, or perhaps a means of mentally focusing on, a concept. Conceptions are said to be Kantian concepts. Now, it is far from clear that these two categories, world-located Fregean concepts and mind-located Kantian concepts (or conceptions), are all we need to pick out in order to make sense of what we need to make sense of, because, after all, a conception seems to be a complex mental construction itself consisting of, or involving, more basic mental elements, but for our purposes, the crucial point is that concepts on Wiggins’ understanding are denizens of the world owing nothing to thought for their existence. (ii) The second thing that I want to add relates to the absence from Wiggins’ formulation of D of any special modal requirement, and my remark that we should read it as a suggested necessary truth. Wiggins is rightly keen to draw, or acknowledge, a sharp distinction between theses about the link between identity and sortals, which is what D concerns, and whatever de re essentialist claims are true in relation to objects and the sortals they fall under. Chapter 4 of Sameness and Substance Renewed attempts to derive a defensible essentialism from D plus some supplementary assumptions. We should, surely, follow Wiggins in acknowledging, and being very careful about, this important distinction. However, I conjecture that Wiggins’ commitment to the separation between these aspects of his overall account might explain the absence of modality in D as formulated. But, we could, surely, preserve the same distinction between the essentialist claims that Wiggins thinks are correct and D itself, even if D is itself advanced as a necessary truth. So I do not see any reasons to understand the commitments of D as being to something less than a necessary truth.
On the Sortal Dependency of Individuation Thesis
257
Next, however, if we stare critically at the formula, we might be further moved to comment that there is an element of redundancy in the formulation of D. Clause (2) seems to spell out what, or a part of what, Wiggins understands by the term “sortal,” and so it merely fills out what the introductory words have already committed him to, and does not, therefore, add to that commitment. In assessing the expanded claim up to the end of (2) the chief interpretative problem is to understand what is meant by “sortal concept,” which task we might divide into elucidating “sortal” and “concept.” Earlier I looked at what Wiggins says about the notion of a concept, and, as just remarked, clause (2) in effect says something in response to the worry about “sortal.” As I see it, though, given a grasp on “sortal concept,” the D-thesis up to the end of (2), and viewing it therefore as simply a necessary condition for a moment, merely says that if a is b then there is some sortal f which a and b fall under or belong to. Clause (3) is the joker in this metaphysical pack, for it is conceptually very complex, introducing the notions of coincidence, manner of coincidence, and congruence. However, if we concentrate on the initial condition in (3), that is the condition before the words “that is,” which is that “a is the same f as b,” then it seems to add very little to (1) and (2) combined (or perhaps, as one might say, simply (1)). Thus, if it is a requirement for a and b to be identical that they fall under the same sortal concept f then it will fairly obviously be a requirement that they are the selfsame f. Formulated this way the beginning of (3) adds very little, if anything, to (1) (together with (2)). Further, as it would seem that the word “same” in the phrase “same f” introduces what “=” introduces, we would have a necessary and sufficient condition for identity where one of those conditions itself involves an identity, albeit an identity presented in the construction “is the same f as” rather than simply in “is the same as.” I want to suggest, therefore, that there is what I shall call a minimal metaphysical reading of Wiggins’ thesis D. That says that; necessarily, a = b if and only if there is some sortal concept f such that a and b are f(s) and are the same f. Confronted by this minimal metaphysical reading, it seems to me that the obvious comment to make is that if it is true that necessarily everything that there is falls under a sortal concept, that is, belongs to one of those types which count as a sortal type, then the specified condition will be necessary and sufficient for any true identity, whereas if there can be entities not falling under sortals then there could be true identities which do not satisfy the proposed condition. So the central critical question that the minimal thesis raises is whether everything must fall under a sortal concept. 2
2 In a hurry to highlight the centrality in what Wiggins is committed to of a thesis linking existence and sortals, I am neglecting a distinction that Wiggins himself draws our attention to, between a thesis he calls D(i), which says that for any object and for any time at which it exists that object falls under a sortal at that time, and
258 Paul Snowdon However, supposing that that thesis is true, then, although the minimal metaphysical thesis version of D would be true, we can further voice the question: why not simply focus on and place at the center of the discussion the crucial proposition that each entity must fall under a sortal? Would it not be clearer to say that the crucial thesis is that existence depends on, or requires, falling under a sortal? That would more appropriately be called the sortal dependency of existence thesis. In fact, and at this point I am expressing a hypothesis that I cannot properly support in the confi nes of this chapter, it seems to me that Wiggins has fallen prey to a fairly common error which I shall call the error of identity fi xation. Identity fi xation stands for the error of formulating a thesis as one about identity when it is more accurately a thesis about something else (in the present case, of course, existence). Another instance of this, it seems to me, is the tendency to describe the so-called problem of personal identity as a problem about identity. The types of facts that this problem aims to elucidate need not be expressed in identity judgments. The type of judgments it is the aim of that problem to analyze can be expressed in tensed claims, which do not contain the identity predicate, for example, “John will be in London in two days,” and moreover, not all identity judgments about persons express the type of fact that is being analyzed, for example, “The person next to the tree is the person playing the drum.” There is, therefore, no obvious reason to think of it as a problem about identity. It is, of course, not to be expected that those who put claims about identity as central to these issues, including of course Professor Wiggins in relation to the present issue, will accept my claim that this represents a distortion of the real problem. What defense of Wiggins’ emphasis can be provided? His emphasis reveals, I suggest, that he believes the D-thesis is, in a sense, the fundamental claim. Now, I cannot see that it is explanatorily more fundamental than the claim that I have called the sortal dependency of existence claim. However, what might be defensible to suggest is that although there is no explanatory priority between the two metaphysical claims the D claim is more easily provable and so has what one might call a dialectical priority. If this could be maintained then something like Wiggins’ practice might not be criticizable. However, the suggestion that it is somehow easier to prove the D-thesis than the sortal dependency of existence claim seems the opposite of the truth. There seems to be nothing about identity propositions which makes it clearer that, for them to be true, objects must fall under sortals, than would become available simply by
the stronger claim, which he calls D(ii), which says that for any object there is a sortal under which it falls at all times that it exists. Wiggins is, and acknowledges he is, committed to the second, stronger, claim. My own formulation of the existence claim, despite an absence of dotting and crossing, is meant to express this second thesis.
On the Sortal Dependency of Individuation Thesis
259
asking and strenuously trying to determine whether objects must fall under sortals. Indeed, it seems reasonable to suggest that the extra complexity in D would make it harder to determine its truth-value.
2
AS MUCH ABOUT SORTALS AS IS NECESSARY
Having set out in a schematic way a minimal interpretation of D, and criticized the way that Wiggins makes D metaphysically central, I need, in order to confer (or perhaps to assure ourselves that we have conferred) enough determinacy on its content, to say a little about what we are to understand by “sortal.” Now, one way to pin down what sortals are sufficient to give a defi nite enough sense to the thesis is by providing a list and adding that we are talking about concepts (or sorts) of a similar sort to those on the list. It could be done like this: examples of sortals are animal, mammal, cat, Persian cat, artifact, table, garden table, etc. It is noticeable that we use nouns to capture or express sortals, and that sortals belong in structures such that they range from the more (for example, animal) to the less, general (for example, cat). These structures are present in each different case. So, perhaps, animal, plant, artifact are three very general sortals, and under them will range more specific ones, and, under those, more specific ones still. Given this structure, and allowing that all the concepts on a single branch from the general to the more specific are sortals, it is obvious that any item itself falls under a range of sortals. We cannot, therefore, talk of the sortal under which an entity falls. None of this is news for Wiggins, of course. Now, as far as I can see, these remarks, based on fi xing the notion of a sortal by examples, suffice to confer enough determinacy on D to understand it and to consider it. The question can at least be raised whether a more perspicuous explanation of the notion of a sortal is available. Judging by his practice, Wiggins’ own attitude to this question is that one (but not necessarily the only) perspicuous way to convey what sortals are is to characterize them as (possible) answers to the “What is it?” question, as he adds, in the sense of that question captured or isolated by Aristotle. If, for the moment, we ignore the important allusion to Aristotle, there is a strong suspicion that the question—what is it?—admits of perfectly satisfactory answers that do not amount to mentioning what are intuitively sortals. After all, we should expect, I suggest, that the question “What is it?” resembles other w-questions—such as “When did the battle of Hastings occur?” or “Where does she live?” Such questions seem to permit of a large range of answers. Thus to the question “When did the Battle of Hastings occur?,” I can say “1066,” or “The spring of 1066,” or “Sometime before 1070,” etc. Similarly one would expect the “What is it?” question to have many permissible answers, which is what seems to be the case. For example, I enter a room carrying an object and someone asks “What is it?” I can reply “It’s a
260 Paul Snowdon shell,” but equally I can say “It’s a present,” or “It’s a gift,” or “It’s a little surprise for Sid,” and so on. None of these last three answers mentions a sortal, I assume, but each is admissible. Again, I can be accompanied by someone, and it can be asked “What is he?” and I can say “My assistant,” and again this is not a sortal. Certainly, too, I can answer the question “What is that?” by saying “It is a baby.” Now “baby” is presumably a phased sortal, not a true (full) sortal. Further, we can imagine a context in which it is quite appropriate to answer the question “What is that?” with the answer “a thing.” Here is such a context. I am training my child to use the word “thing” with appropriate generality. I point to a number of objects and call them things, and then I point to, say, an apple and ask “What is it?” and the answer “A thing” is quite appropriate. Now “thing” is not a sortal, it is simply too general. So the question “What is that?” admits of many answers, and we certainly could not pin down the category of sortals as what is required to answer that question. If Wiggins adds that this is to ignore the qualification “In Aristotle’s sense” or “As Aristotle interpreted it,” then the response that seems appropriate is to wonder both what the way is that Aristotle has to capture the particular interpretation of the question, and why a direct expression of that way (whatever it is) cannot enable us to home in on the category of sortals directly, without reference to the question “What is it?” However, although not particularly attracted to the clarification of “sortal” via allusion to the “What is it?” question, I think that the list method conveys well enough what “sortal concepts” are for our purposes.
3 A NONMINIMAL METAPHYSICAL INTERPRETATION OF D It would not be right to regard thesis D as completely or satisfactorily captured in the minimal formulation. The reason is that in expanding, or providing a gloss on, clause (3), Wiggins introduces some concepts and ideas that spell out his particular understanding of what (3) says and why it is important. I am not at this point going to try to explain in a detailed way what those ideas and concepts are on my reading. What, however, seems to be involved are three things: fi rst, the introduction of a new relation between a and b expressed by the word “coincides”; second, that coincidence can be in different manners required by different sortals; and third, where there is coincidence between a and b in the manner as required by the appropriate sortal f, they coincide congruently, which is a technical term meaning, roughly, that the coincidence yields or entails or sustains an identity between a and b. So what I want to say is that there is a nonminimal metaphysical reading of D, which adds to the minimal reading of the claim (roughly) these further categories, coincidence, coincidence in the manner required by a sortal f, and congruent coincidence and proposes a
On the Sortal Dependency of Individuation Thesis
261
thesis involving them, spelling out what being the same f amounts to, and why, from the point of view of identity, that is important. Although I do not wish to investigate these notions with any concentration because there is a third interpretation of Wiggins’ thought that I wish to set out and concentrate on instead, I shall indicate what it seems to me Wiggins means by those terms and to express a worry. The way that I think we can interpret these notions is, first, that it is a condition for a and b to be identical that they coincide spatially and temporally so that wherever and whenever a is so is b and vice versa. However, spatiotemporal coincidence between a and b is not enough for identity; they must further fall under the same sortals, in which case the coincidence is “congruent,” meaning that it is sufficient for identity (with all its implications and requirements). In Wiggins’ view, the extra requirement of what we might call sortal coincidence as well as spatiotemporal coincidence reflects the response, in the first chapter, to cases which have been cited in favor of the relativity of identity. However, there are some worries about Wiggins’ approach. The fi rst is that my reading of “coincidence” does not seem to allow room to talk, as Wiggins does, of coincidence “under f in the manner of coincidence required for members of f.” If “coincidence” means spatiotemporal coincidence then it makes no sense to talk of it being under f; a and b are simply in the same places at the same times, and in fulfilling that coincidence the sortals they fall under have no relevance or role. Second, Wiggins himself does not explain how the fact of coincidence is one that is prior to and a ground for identity. He faces the question, which I am not meaning to say is unanswerable, of how we can make sense of the spatiotemporal history of a itself, without employing identities about a, such as a was the thing occupying that space at that time. Third, even if Wiggins has shown that, in the cases he was considering in Chapter 1, they require that identity judgments must (in some sense) incorporate reference to sortals, it would not follow from that that they must always do so, unless Wiggins has already shown, at least, that all material continuants have to fall under sortals. As far as I understand it, by the beginning of Chapter 2 that has not been shown at all. I have tried to explain how to understand the nonminimal metaphysical claim and to raise some questions about the thesis so interpreted.
4
A COGNITIVE READING OF D
I have so far distinguished two metaphysical interpretations of D, and expressed some questions that they seem to raise or face. However, I think that we have at least to put on the table a quite different type of thesis. Part of the reason for supposing it is needed is that Wiggins calls D a thesis about individuation. It is, after all, individuation which, according to D as named, is supposed to depend on sortals. In his preamble, Wiggins is careful to specify what he understands by “individuation.” Wiggins explains
262
Paul Snowdon
that he wishes to advance a characterization of “what it amounts to, practically and cognitively, for a thinker to single a thing out at a time” (2001, 1). Wiggins calls this activity of singling out “individuation.” He says that “to single x out is to isolate x in experience; to determine or fi x upon x in particular by drawing its spatiotemporal boundaries and distinguishing it in its environment from other things of like and unlike kinds” (2001, 6). It is clear that this activity or occurrence is a psychological or cognitive one. Wiggins further summarizes, in the Preamble, his account of how this is done by suggesting that singling out x requires something which “approximates to this; the thinker’s singling x out as x and as a thing of a kind f such that membership in f entails some correct answer to the question ‘What is x?’” (2001, 7). Clearly, the introduction of “f” coupled to the “answer to the ‘What is x?’ question” formula means that the thinker is being required to classify x under a certain sortal in order to single it out, or at least, as Wiggins puts it, to approximate to classifying it under a sortal. Now, this is, it seems, a psychological dependence thesis, completely appropriately called the sortal dependency of individuation. I think, therefore, that Wiggins must be credited with acceptance of such a thesis, and, moreover, he seems to think, at times, of it as what D stands for. We should remember, also, that in Chapter 5, entitled “Conceptualism and Realism,” Wiggins’ general purpose is to argue, amongst some other things, that his approach, which he is prepared to call conceptualism, is as realist as we should wish to be, and in the course of arguing that he employs the famous net analogy. He says, “Our claim was only that what sortal concepts we bring to bear upon experience determines what we can fi nd there—just as the size and mesh of a net determine, not what fish are in the sea, but the ones which we shall catch” (2001, 152). Now, the net analogy clearly is relevant to the sortal dependency of individuation but has absolutely no link to the more metaphysical claims. In the Fregean notion of a concept, concepts are rather general things on the far side of the net, waiting to be caught, not devices of catching. So, at this stage, Wiggins is in effect treating his central claim as the psychological one. Moreover, by this stage the nonpsychological interpretation of “concept” talk, so carefully established earlier, seems to have been forgotten or left behind. Concepts are what we have and bring to bear on experience. I am therefore inclined to say that as Wiggins uses the name D he sometimes lets it stand for a psychological claim about what is required for singling out objects in experience, that is to say, for individuation. This then leaves us with three theses that Wiggins is proposing—the two metaphysical ones and what I am calling the psychological one. My critical interest here is in the psychological thesis. But before focusing on it we can ask: what is the relation between the psychological thesis and the metaphysical ones, in particular the minimal one? It seems clear that the minimal metaphysical principle could be true without the psychological one being true. The fact that objects around you
On the Sortal Dependency of Individuation Thesis
263
all do fall under sortals does not mean that in order to pick them out you need to know what sortals things fall under. The link the other way is more interesting. Obviously, when Wiggins talks of bringing an experienced object under a sortal as something that is involved in thinking about it, or singling it out, he does not suppose that bringing it under just any sortal somehow helps. He means that bringing it under the (or a) sortal it does fall under is what is involved. That is to say, the notion of bringing it under a sortal is truth involving. So if the psychological thesis is true it will follow that any object that we can single out, or perhaps that can be singled out by any creature, will fall under a sortal. And so for that range of objects the minimal metaphysical thesis will hold. However, it has no implications for objects that cannot be singled out or individuated in Wiggins’ sense. There may be continuants that we cannot single out, and there may be other kinds of things. So the minimal thesis in its full generality will not be entailed. It can also be noted that the closest metaphysical thesis will be the one I called the sortal dependency of existence thesis and not the one about identity.
5
D AS A PSYCHOLOGICAL THESIS
In order to assess the psychological thesis I need to make four preliminary remarks. The fi rst two pick out aspects of Wiggins’ views that I ignore. The second two address interpretative problems about two categories employed in Wiggins’ theories. (I) When Wiggins states what is necessary for individuation he qualifies the condition as that of “approximating” to singling x out as an f (2001, 7). If we stress that element it becomes very hard to know what is being claimed. Thus, if someone can approximate to singling an object out as an f without actually doing so, what is involved in approximating to it? Just what degree or type of information is required? I propose, therefore, to ignore the notion of approximation. There may be some unfairness here, but we need a relatively defi nite thesis to focus on. (II) The second element in Wiggins’ exposition of singling out is that he talks of the subject’s “practical sensibility” in relation to x as a way of satisfying the conditions he is maintaining (2001, 7). However, (i) Wiggins does not specify quite what he has in mind by “practical sensibility”; (ii) he does nothing to justify the attitude he takes here, which is that the distinctive practical sensibility is a way of fulfilling the sortal dependency of individuation thesis, rather than the alternative view, which is that the distinctive practical sensibility is a different way to single out objects without categorizing them under sortals; (iii) we should not exaggerate the role of practical sensibility. One reason is that we can single out items to which we cannot take a practical stance because they are too distant, hence outside our range for engagement with, say the sun or the moon, or an extremely distant mountain range. Another
264
Paul Snowdon
reason is that we can single out objects, which are within our range to engage with, but the behavior and nature of which are so unpredictable that all we can do is observe them and wonder. I therefore also ignore this strand in Wiggins’ view, again running the risk of unfairness. (III) I have so far relied on Wiggins’ language to pin down what the psychological thesis is about. It is about “singling out” or “picking out” or “individuating” something done by a thinker to an object at a time (2001, 6). But the time has come for us to ask, what exactly are we talking about here? I think that this is not an entirely easy question to answer. What seems to me to best fit the way Wiggins speaks is that picking out is what a thinker does when he engages in perception-based object-directed thought. Thus I might think to myself, while focusing on a glove, “That was not here earlier,” or I might wonder, staring at an object in the distance, “What is that?” This is to understand picking out, hence individuating, as what has been called thinking about “in presence,” as opposed to thinking about “in absence.” I can do the latter by employing proper names or descriptions. But although I am in a perfectly good sense picking out an object in thinking about it in these latter ways it is not what I think Wiggins means by picking out. In engaging in his sort of occurrent thought process I am, then, individuating an object. But there is one remark that he makes which is, perhaps, hard to reconcile with this understanding of picking out. He says; “Singling out is the sheetanchor for information about particulars” (2001, 6). This does not seem quite right. We need not engage in occurrent thinking about objects in our environment to gain information about them. We gain information by perceiving the environment, without individuating any thing in it. If Wiggins wishes to maintain the last characterization, it becomes very hard to know quite what individuation is. I prefer then to abandon his claim about picking out being the sheet-anchor of information. There is however, consistent with my interpretation, a way to respect something like Wiggins’ claim. One must accept, surely, that ordinary perception gives us information about objects, and it is the basic way that we acquire it. However, to advance beyond elementary knowledge of the world and of objects it is plausible to say that singling out is necessary. To develop a conception of, say, cats or spiders, we do have to single out particular examples and study them. So the sense in which Wiggins is correct is that singling out is the sheet-anchor for advanced, systematic, and thorough knowledge of objects and of the types of objects there are. But we should notice a consequence of this defense of Wiggins’ remark. I think that we can say that we develop our understanding of a notion like that of a cat by building up a body of knowledge about such creatures. Now, what I have just said is that picking out individual cats (and other objects) is part of the process of building up this knowledge. This, however, implies, I suggest, that we cannot think that we need sortals of this kind, of the moderately specific kind, in order to single out objects. The truth is that we can only do what amounts to acquiring such sortals as a response to observing them in virtue of already picking them out.
On the Sortal Dependency of Individuation Thesis
265
(IV) If that is what Wiggins means, then I think that we have a defi nite enough understanding of the topic of the thesis. We can ask whether some of the things that Wiggins says about individuation, aside from its dependence on sortals, are correct. But initially I focus on the relation of picking out, understood as explained above, to be the employment of sortals. However, what does Wiggins have in mind when he talks of singling the object out as of kind f, that is to say as falling under a sortal? Now, it seems to me that when he advances the dependency thesis Wiggins tends to think of the sortals which we need to use to single objects out as ones which belong, roughly, in the middle range of sortals and not the ones that belong at the more general level or the ones that belong to the more specific level. For example, of the three sortals under which a certain cat falls—animal, cat, Maine coon—there is no plausibility at all in supposing that singling it out requires employment of the most specific sortal, as we obviously can quite happily single out particular cats (or, say, trees) without knowing what basic sort they are, but it seems to me that Wiggins tends to imagine it is the middle sort that we need to employ. Be that as it may, I now argue that it is not necessary to single out under any sortal.
6
SOME APPARENT COUNTEREXAMPLES
Having, I hope, moved the debate along somewhat I want to look at some simple and obvious examples. There seem to be lots of examples where people can think of an object in ignorance of its sortal type. These can come about in different ways. First, the thinker might be familiar with the basic sorts in the environment but be ignorant in relation to a particular thought about an object as to what sort of object it is. For example, a child can point to a robin and ask “What sort of bird is that?” The child clearly picks out a particular robin without thinking of it as a robin. The child, does of course, think of it as a bird. That, though, is not necessary. Thus someone could point at a creature flying in the air and ask “What is that?,” being unsure whether it is a bird or a very large flying insect. Perhaps this time the sortal is “animal.” However, even that is not necessary, because someone could point at a flying thing and wonder of it what it is, without knowing whether it is an animal, or a working model of an animal, or a kite. This sort of thinking about an object is hardly recherché or colossally artificial. Another sort of case is where the thinker is not familiar with the sorts present in the environment. Science fiction fi lms provide examples where an object falls to earth, and the scientists ask “What is it?” Sometimes it turns out to be a lump of an unknown metal or a spacecraft, and sometimes events reveal that it is a creature, usually very dangerous. The quite normal idea is that as we encounter objects from locations we have not explored, we shall not immediately know what type or sort of thing they are.
266 Paul Snowdon Another imaginary example of sortal ignorance is what we might call the rapidly evolving world scenario. Evolution, of course, does occur slowly, but we can imagine that it might have happened much quicker. In such a world a short sleep would result in the disappearance of previous zoological and biological sorts and the emergence of entirely new ones. Each day the subjects would have to classify things again, but, I submit, singling out would not be difficult. There are other sorts of cases. I have in mind examples where we have perceptual contact with an object which gives us very little indication what sort of object it is. Thus, I hear a rustle and wonder “What is that?,” where the focus of the thought is on the noise maker. This kind of perceptual contact does not enable me to determine what basic sort the item belongs to, but it does enable me to think about it. Another interesting case is this. I wake up in a darkened room in need of a drink, so want to locate my glass of water. In the darkened condition I can, as we say, make out objects close by, and wonder of one of them whether it is my glass. Now, here my thought is not directed onto the object by any sortal conviction. The targeting, rather, enables the sortal question to be raised about a particular item. Or think of picking out objects in the far distance. Our perception of them might be so hazy as to make classification impossible, but we have succeeded in picking out the item, which may reveal itself to be a battleship. Artifact sortal kinds provide interesting examples too. One reason they are interesting is that new artifact sorts are continually emerging, so even a thorough knowledge of your environment gives no guarantee that you can know what sort a new artifact belongs to. Even well-informed thinkers can fi nd themselves asking of an artifact—what is that? With nonartifact sorts new cases tend to be rarer. There may be other reasons for uncertainty. I pick up an object made from twigs and seaweed and shells and held together by tar and fraying rope and then put it on a base and have it on my desk. Someone might say “What is that?” and it is clear what object they have in mind. I might be inclined to respond by saying “What indeed?” We can add to these counterexamples by reminding ourselves that it is through engagement with particular objects in our environment that we are able to acquire the sortal notions that we possess. At a certain stage on encountering an object, we might ask “What is that?” and be told that it is a giraffe. This question and answer begin our acquisition of that sortal notion. Furthermore, it is in such encounters with objects that the naturalist develops his or her knowledge, in the first place, of the type of objects in their environment and builds their taxonomies, to which we are privy secondhand. The examples I have appealed to do not seem to me strange or recherché. They may not represent the usual case of the well-informed thinker, knowledgeable about his or her environment, but they seem to establish that we can pick out and think about an object without knowing what sort of object it is (say a robin or a golden retriever) and even without knowing
On the Sortal Dependency of Individuation Thesis
267
what general sort it is (say a model or an animal). Moreover, the commonsensical reflection that such contact must be possible in order to acquire the sortals we employ also seems perfectly sound.
7
THE AMBIGUITY ARGUMENT
In the light of these examples it seems to me that the D-thesis in its psychological version looks incorrect. In fact a philosophical argument is what makes it attractive, despite the contrary evidence. The argument that I think holds the view in place is not presented in a direct way by Wiggins, and it may be that he would not accept it, but its influence, I believe, is pervasive in the thought of those who accept his dependency thesis. I want to set it out and discuss it. I argue that it does not work, but I also think that the argument sets problems which are not easy to solve. How might acceptance of the thesis that I have been attacking be motivated? The chief argument for it (which I shall call the ambiguity argument) rests on three premises. (1) The multiple-occupancy thesis (hereafter the M-thesis) is true. What is that? It is a proposition that concerns the relation between objects and space. The orthodox view over the last forty years, the fullest defense of which has been provided by Wiggins himself, has been that two objects (which are not identical to each other) can occupy the same space at the same time. There is a way to take this claim which makes it uncontroversial. Thus, there is a space that my left hand occupies at a time which is also a space I occupy at the same time, and I am not identical to my left hand. The point is that the total space that I occupy is not the same as the total space my left hand occupies. The theoretically important claim is, therefore, that it is possible for two nonidentical objects to coincide exactly in respect of the space they occupy at a single time. This is the M-thesis. The standard way to support the M-thesis is by citing supposed examples where object a and object b are spatially coincident at a time but where they cannot be supposed to be the selfsame thing because they have different life histories. The most-cited example is, of course, where a is a statue and b is the lump of clay out of which it is made. Because a only started to exist when the sculptor made it which is much later than when b, the lump, began to exist, they cannot be identical. In the orthodox view two things would be added. First, this type of situation is possible only if a and b are different sorts of thing, for example in our example a is a work of art, and b is a lump of matter. Second, it is possible for a and b to be perfectly spatiotemporally coincident throughout their histories and yet be nonidentical, so long as they fall under different sortals. (2) If the M-thesis is true then in order for a thinker to be thinking of one rather than the other object occupying the same space at a single time, there must be some way the thinker is thinking of the thing he has in mind, some concept he is applying, which targets his thought onto that object rather than the other one (or others). (3) The only way to do this is by thinking of the item under the
268
Paul Snowdon
sortal category into which it falls and the other object (or objects) there does not. This argument, then, alleges that the only ground for disambiguating the direction of thought onto an object is by the thinker’s employment of a sortal concept to pick it out. To this argument, some people would simply reply that premise (1) is false. However, to those of us less sure about that, the crucial question is whether the other premises can be questioned. Premise (3) is, I think, doubtful because, for one reason, there remains the possibility admitted by many supporters of the M-thesis that there might be a property, other than its sortal nature, possessed by only one of the co-located entities, and it could be singled out via that property. This is not merely an abstract possibility. Rather, the actuality of such descriptions is relied on by people who argue for the fi rst premise in the ambiguity argument, at least in the normal way to argue for it. To appreciate this, all that is necessary is to ask: how is it established that, for example, the relation between the statue and the lump is not identity? It is not established simply on the basis of the fact that the lump is a lump and the statue is a statue, as that sheds no light on the question of their relation. To establish their nonidentity a certain property is cited, usually a modal property, which it is claimed applies to one but not to the other. Now it is, more or less, implicit in this procedure that one can discriminate the two objects present in virtue of this property difference. It is not, therefore, necessary to do so by employing the sortals to pick them out.3 Again, consistent with M is the idea that there are, for example, only two things co-located, say an F-thing and a G-thing. The thinker can then pick out the F-item as the thing over there which is not a G, without thinking of it as an F-item. So, premise (3) overlooks these possibilities, and for all I know, others. There is, then, no demonstration of its truth, even given (2). Equally, and now considering premise (2), it is not obvious that there are not other features of thinkers than their application of a concept of being, say, F, which makes it true that what they are thinking of is that F-thing. One candidate for such a feature is that they have a primitive disposition to trace the thing they are thinking of in a way which matches its being the Fthing and not the G-thing. Another possibility, to my mind quite plausible, might be that our perceptual experiences have a character that directs our object-directed thought onto one rather then the other object present. For example, there may be in a single space at a single time a certain quantity of matter and, say, the animal that it makes up. These two things are not identical. However, it may be claimed that our experience traces and makes salient the animal and not the matter. What is important here is simply that there are alternatives to (3) that have not been ruled out.
3 I am very grateful to Josh Parsons for clarifying this point for me and for other ideas too.
On the Sortal Dependency of Individuation Thesis
269
Given, then, that the argument rests on at least two premises that can be questioned, it should not, I feel, tempt us to ignore the earlier examples. My aim is not to solve all the problems that the argument throws up, which I cannot do, but to identify ways of resisting the argument and rendering its conclusions nonobligatory. The motive for resisting is that the conclusion simply does not seem true.
8 ANOTHER PROBLEM FOR THE AMBIGUITY ARGUMENT I want to raise another query about the ambiguity argument. There is, surely, something rather artificial about the argument. When encountering an object and wondering what it is, I have no sense that an ambiguity of attention or of picking out threatens and so needs to be avoided by conceptual action. However, leaving that artificiality aside, it is important to ask how the invocation of sortals actually helps to solve the problem. To apply it to an example for purposes of illustrating the general idea: presumably, it is a necessary truth that, for example, a cat ceases to exist at death, but the cat’s body does not cease to exist then. So thinking of what is in front of you as a cat and bringing it under that sortal singles out the animal rather than the other co-occupant of the space, the cat’s body, by homing in on the item with the animal’s persistence condition rather than the cat’s body. This answer, however, faces a question. Cannot people who single out the cat disagree over the persistence conditions of cats? The answer is that they can. Then the sortal explanation seems to face a dilemma. Either it is still possible to be picking out the cat despite assigning it the wrong persistence conditions, in which case it is difficult to see how that focusing is disambiguated by the thinker’s sortal concept, or such a thinker is not picking out the cat, which however, is manifestly absurd. It does not help to say that the thinker is still applying the concept cat, because in the absence of acknowledgment of the correct persistence requirements nothing about that thinker’s focusing has been explained.
9
OTHER ASPECTS OF SINGLING OUT
I have so far argued that the psychological version of the sortal dependency thesis faces obvious counterexamples, does not seem consistent with an obvious and commonsensical account of how we form and acquire our sortal concepts, is not properly supported by the ambiguity argument, and provides no clear explanation of picking out anyway. What, though, of some of the other things that Wiggins says about picking out? He characterizes it thus; “ . . . to single x out is to isolate x in experience, to determine or fix upon x in particular by drawing its spatio-temporal boundaries . . .” (2001, 6). What, though, is meant by “drawing its spatiotemporal boundaries”? Thinking first
270 Paul Snowdon about spatial boundaries, consider the following cases. I pick out a tree, but I have no idea as to the spatial boundaries of those parts underground. I stand on the shore of Lake Michigan and pick out that lake. I have no idea as to its distant spatial boundaries. Again, I stand in front of Mont Blanc and pick out the mountain. I have only the haziest idea of its boundaries on the side away from me; moreover I would be hard pressed to determine the spatial boundaries of the base of the mountain. Finally, I am placed in a vast forest. I can pick out the forest I am in but cannot determine its spatial boundaries. Although it is clearly right to link picking out with locating aspects (or parts) of the thing, it does not seem correct to make determining its boundaries necessary. Talk of determining temporal boundaries is also, indeed more, problematic. Obviously I can single out a cat even though I have no idea when it was born or when it will die. I need have no clear idea at all of its actual history or its future. It is very difficult to accurately specify the spatiotemporal understanding required for singling out. Here again there seem to be counterexamples to requiring anything too defi nite and extensive. To be more constructive, in analyzing the phenomenon of singling out, at least three things need acknowledging. First, given the way concept-talk is introduced, we can, I think, ascribe to someone not only standing concepts of general features, such as the concept of red or man, but also the concept of a particular thing, the concept, say, of that thing, possessed by the thinker for a time in a certain attentive state in a certain perceptual situation. Of course, we need to ascribe to cognizers of our kind the general cognitive capacity to acquire such concepts in certain perceptual-cum-cognitive situations. Second, we should, it seems to me, acknowledge that in the situation where such singular concepts are elicited, the thinker’s understanding of the item will include some general characteristics of the item, but the relevant characteristics need not be particularly rich and may totally depend on the appearance of the item in question, however limited or restricted that appearance may be. Third, the proper understanding of these object-directed thoughts requires relating them to an account of the sort of perceptual contact with objects that our perceptual system gives us. We should say, that is, that our perceptual system puts us in contact with objects, and our conceptual system generates appropriate object-directed concepts. How this system works and how it develops is an empirical question.
10 THE METAPHYSICAL SIGNIFICANCE OF SORTAL DEPENDENCY It has been thought by some critics of Wiggins that if the psychological dependency thesis is correct then it implies the truth of some sort of idealism. Because the desire to avoid idealism is very strong, this motivated their opposition to the dependency thesis. Wiggins’ response was the already
On the Sortal Dependency of Individuation Thesis
271
mentioned famous fish net analogy. The dependency thesis merely implies that we will pick out those items around us which fall under the sortal concepts we apply, not that what exists around us depends on our concepts. The net analogy reveals that the idea has no idealist implications, because the net determines what fish amongst those which are there we shall catch and does not determine what is actually there. I think that the analogy does work to disarm the supposed idealist implication. Nothing is implied by Wiggins’ theory about the mode of existence of the external objects that we can single out. My own reluctance to accept his thesis does not spring at all from anti-idealist worries, but rather from its apparent inadequacy as an account of object-directed thought. However, it is important not to be dazzled by the net analogy in a more general way. If the net represents what might be thought of as our structure of sortal concepts, we should recognize that our detection of the objects around us is not limited in the way the fisherman is limited in what he can catch by the shape of his net. In particular, we can become aware in our explanatory theoretical enterprises that there is at a certain point in space an entity with certain causal powers. We need not know what sort of object it is as yet. That remains to be determined. That is to say, we can catch objects which are not ensnared in our existing sortal network. The net analogy misconceives our access to objects in the world.
11
CONCLUSION
I have tried to argue against the D-thesis on its psychological interpretation. I have also set out without attempting to undermine the metaphysical theses that Wiggins advances, although I have suggested that there are problems of focus in their formulation. The general attitude that these reflections suggest is that we should preserve a sense of the distinction between the two sorts of claims and abandon the psychological claims.4
4 I want to thank Bill Brewer, David Charles, Dorothy Edgington, Oliver Pooley, and Tim Williamson for comments when the chapter was presented in Oxford. I used Sameness and Substance Renewed as a text for a graduate seminar at UCL and learned a lot from the discussion and questions; only some of those lessons have I drawn on here. I also benefited from the discussion when the chapter was read at the Dunedin conference, including reactions by Hugh Mellor, Frank Jackson, Jack Smart, and Michael Devitt, and especially helpful comments by Josh Parsons. I have no real sense of what Alan Musgrave thinks of these issues, but I intend this chapter as an inadequate token of my esteem and friendship for him. I am also very grateful to Heather Dyke for the invitation to participate in the conference in Dunedin and for support in writing this.
14 Truthmakers for What?
1
D. H. Mellor
1
INTRODUCTION
Despite the arguments of Simons and others (Mulligan, Simons, et al. 1984, §5; Fox 1987; Simons 2000), we do not all agree that truths need truthmakers (Beebee and Dodd 2005), and those of us who do agree differ about which truths have them (Forrest and Khlentzos 2000). Many follow Martin (1996) and Armstrong (2003) in taking the maximal truthmaker view that all truths have truthmakers. I share with Heil (2000) and others the moderate view that only some truths, the primary ones, have truthmakers, while other truths and falsehoods are derivable from the primary truths by means of truth conditional semantics. (Forrest and Khlentzos 2000, 3) In this chapter I restate the case for moderate truthmaking, adding what I think determines which truths are primary but starting by saying why I think at least some are.
1 The immediate ancestor of this chapter was discussed at a conference on Truth and Reality held in January 2007 at the University of Otago, which met my considerable travel expenses. Remoter ancestors have been discussed from 2001 on at the Universities of Nottingham, Durham, London, Stirling, Melbourne, and Cambridge, the Australian National University, and the Universität Zürich. Repeated revisions of the chapter owe much to these and other discussions. I am also indebted to the British Academy for a research grant to meet my expenses in travelling to and from the Australian National University in 2002 to work on truthmaker theory. A German translation of the chapter’s Zürich ancestor is published in Mellor (2004).
Truthmakers for What? 2
273
TRUTHMAKING AND ENTAILMENT
For present purposes we may take the principal bearers of truth to be propositions, which I write 〈 . . . 〉, as opposed to sentences that express them, which I write “ . . . . ” Besides not being sentences, it will not matter what (within reason) propositions are: whether, for example, they are sets of possible worlds, true if they contain our actual world and false if not. My main assumption, following Bigelow (1988, 132) and others, is that, as Bigelow puts it, “truth is supervenient on being.” This I take to mean that only if two possible worlds differ in being can a proposition be true in one and false in the other, where “differences in being come in two sorts . . . differences in whether something is . . . and differences in how something is” (Lewis 2003, 25). This implies, for example, that all worlds where the proposition 〈The earth is round〉 (〈ER〉 for short) is false either lack our earth 2 (a difference in what there is), or their earth is not round (a difference in how it is). I assume moreover that all such worlds (¬ER-worlds for short) must also differ from ours in the truthmakers they contain (Rodriguez-Pereyra 2005, 17). This, however, does not require our world to contain a specific truthmaker for 〈ER〉 that all ¬ER-worlds lack, as moderate truthmaking does not require 〈ER〉 to be a primary proposition, that is, one that needs a truthmaker to make it true. But if 〈ER〉 is not primary, its truth can only supervene on being if its truth-value depends on whether some primary propositions with truth-values of which, in any given ¬ER world, at least one differs from what it is in our world, are true or false. That is why I take the supervenience of truth on being to require some propositions to be primary. I also take truth’s supervenience on being to imply that the truthmaker S of a primary truth 〈P〉 will generally not be another proposition.3 In other words, S’s truth-making relation to 〈P〉 will generally be what Armstrong calls cross-categorial (2003, 13). If this seems obvious, it has a less obvious but important consequence: because entailment relations between propositions are not cross-categorial, the truth-making relation cannot be identified with the entailment of 〈P〉 by the proposition 〈S exists〉. To say this is not to denigrate the role of entailment in transmitting truth from one proposition to another, merely to observe that being entailed by other truths cannot be what makes primary truths true.
2 Or one or more counterparts of it (Lewis 1973, ch. 1.9), an alternative reading that for brevity I hereafter take as read. 3 Sometimes it will be, as when a proposition 〈A〉 being primary makes true the proposition 〈B〉 that 〈A〉 is primary. Such exceptions we may ignore for present purposes, noting merely that in none of them will 〈B〉 be identical to the proposition 〈A〉 that is, or is a constituent of, what makes 〈B〉 true.
274 3
D. H. Mellor
TRUTHMAKING AND TRUTH
Taking truthmaking to relate a true proposition 〈P〉 to a generally nonpropositional entity S may make theories of truthmaking look like correspondence theories of truth. But they are not. Theories of truth tell us, rightly or wrongly, what it is for a proposition 〈P〉 to be true, which need no more tell us what makes 〈P〉 true than saying what it is to be a prime minister (to head a government) tells us what makes someone a prime minister (commanding a parliamentary majority). And as with prime ministers, so with truth; and what truthmaker theories aim to tell us is what makes propositions true, not what it is for them to be true. Truthmaker theorists therefore still need a theory of truth. For to say what makes 〈P〉 true is to say what gives 〈P〉 whatever it takes for 〈P〉 to be true, and what that is depends on which theory of truth is the right one. But the right theory need not be a correspondence theory, and for moderate truthmaking it will not be, because nonprimary truths that have no nonpropositional truthmakers need not correspond to any worldly fact. Nor, for any truthmaker theorist, can the right theory of truth be an identity theory which says that every apparently worldly fact P just is the corresponding truth 〈P〉 (Candlish 2005). This thesis, that no true proposition owes its truth to anything nonpropositional, is incompatible with the whole idea of truthmaking: that true propositions generally owe their truth to something else. Fortunately, as correspondence and identity theories are not the only tenable theories of truth, truthmaker theorists need accept neither. In fact all they need assume about truth is the relatively uncontentious equivalence principle (EP) that, for all nonparadoxical 〈P〉, (EP) 〈P〉 is true if and only if P.4 But although (EP) may tell us what it is for 〈P〉 to be true, it cannot tell us what makes 〈P〉 true. To see this, consider apparent truths like 〈Murder is wrong〉 or 〈Neutrinos have no charge〉, which seem to be about values or theoretical entities. Those who believe in values and neutrinos may claim that these two propositions are made true by the substantial facts, respectively, that murder is, as a matter of objective fact, wrong, and that all neutrinos are uncharged. But as Musgrave (1993, 266), Dyke (2007, 5),
4 (EP) is not completely uncontentious. It fails for nonclassical logics in which, for example, “if ‘P’ is indeterminate (as it is if it is a future contingent), ‘It is true that P’ is false (rather than being itself indeterminate)” (Bourne 2006, 94). Here, however, I take (EP) for granted and leave others to say how truthmaking would fare if it failed.
Truthmakers for What?
275
and others have observed, such claims can hardly be entailed by the corresponding instances of an uncontentious (EP): 〈Neutrinos have no charge〉 is true if and only if neutrinos have no charge; and 〈Murder is wrong〉 is true if and only if murder is wrong. For an (EP) that did entail these claims would at once refute (i) antirealists who say that these apparent propositions have no truth-values and so need no truthmakers (Ayer 1946); (ii) ethical naturalists who think that 〈Murder is wrong〉 is made true by natural facts (Foot 1978); (iii) empiricists who think that 〈Neutrinos have no charge〉 is made true by observable facts (Ramsey 1929b); and (iv) physicalists who take psychological truths to be made true by physical facts (Armstrong 1993). And no one thinks (EP) can do all that: no one expects a theory of truth to settle the ontologies of value, scientific theories, or the mind. (EP)’s inability to identify truthmakers is also evident in M. L. Wilson’s (1994, 540) rainbow example. To vary this slightly, suppose 〈P〉 says truly that there is a rainbow east of a car c driving north. Then (EP) tells us that 〈P〉 is true if and only if there is a rainbow east of c. So for (EP) to tell us what makes 〈P〉 true, the world would have to contain this rainbow—call it r—and that poses a problem. For even though r is clearly where some rain is falling east of c, r also keeps pace with c as c moves north, which implies either that c causes r to move with it, or that r is as far east of c as the sun is west. Both alternatives are incredible, the fi rst causally and the second spatially. But an incredible entity can hardly be part of what makes 〈P〉 true; and it need not be, as 〈P〉 need not be a primary proposition. 〈P〉 can simply be entailed by there being many raindrops east of c reflecting and refracting sunlight back to c at angles that vary with its frequency. A moderate ontology need include no rainbows at all.
4 ONTOLOGICAL COMMITMENT Yet how, if there really are no rainbows, can (EP) let 〈P〉 be true, when (EP) says that if 〈P〉 is true then there is a rainbow east of c? The short answer is that (EP) is a biconditional that can just as well be written as (EP) P if and only if 〈P〉 is true. In other words, for there to be a rainbow east of c is just for 〈P〉 to be true, which it is. But what makes 〈P〉 true is another matter and need no more include a rainbow than whatever makes 〈There is nobody in the house〉 true
276
D. H. Mellor
need include an entity (nobody) that is both larger than the universe and smaller than a point, both of which nobody is. This, however, still leaves the semantic question of how to reconcile the nonexistence of rainbows and of nobody with the truth of instances of 〈There is a rainbow . . . 〉 and 〈There is nobody . . . 〉. We can do it for the latter by saying that their “logical form” is really 〈There is not a person . . . 〉, that is, by taking these propositions to be negative existentials. But this will not work for 〈There is a rainbow . . . 〉, as rainbows are only causally or spatially, not logically, incredible: the logical form of 〈There is a rainbow . . . 〉, if there is such a thing, is undeniably that of a positive existential. Although this may only show that “inferences grounded in ‘logical form’ can . . . lead us astray” (M. L. Wilson 1994, 519), we still need to see when and how they do so. To do this, we must fi rst recall that truthmaker theorists need not take any true proposition 〈P〉 to entail the existence of a specific truthmaker S. The entailment here, if any, goes the other way, from 〈S exists〉 to 〈P〉. So the problem here is not for truthmaking but for Quine’s criterion of ontological commitment: . . . we are convicted of a particular ontological presupposition if, and only if, the alleged presuppositum has to be reckoned among the entities over which our variables range in order to render one of our affi rmations true. (Quine 1948, 13) The question then is whether rainbows have “to be reckoned among the entities over which our variables range in order to render” some instances of 〈There is a rainbow . . . 〉 true. For Quine, this depends on whether (to adapt his example) we can devise some way of so paraphrasing the statement [about rainbows] as to show that the seeming reference to [rainbows] on the part of our bound variable was an avoidable way of speaking. (Quine 1948, 13) Can we, for example, restate all true instances of 〈There is a rainbow . . . 〉 in sentences that refer only to raindrops and sunlight? If we can, the “seeming reference to rainbows” in a true 〈There is a rainbow . . . 〉 may well be just “an avoidable way of speaking.” Whether such paraphrases of “There is a rainbow . . .” can really rid our ontology of rainbows is a very moot point (Mellor and Oliver 1997a, §§5–6; Dyke 2007, ch. 1); and even if they can, we still need a criterion for when sentences stating true propositions tell us what there is and when they do not. And (EP) can certainly not be that, because all the sentences, “P1,” “P2,” . . . that we take to state a proposition 〈P〉 generate sentential counterparts of (EP)—
Truthmakers for What?
277
“P1” is true if and only if P1; “P2” is true if and only if P2; . . . —between which (EP) itself can evidently not discriminate. This is why parties to ontological disputes need more than (EP) to link what there is, and how it is, to what is true. They need theories of truthmakers—not of what truthmakers are, but of what truthmakers exist—theories which, to revert to the examples of §3, may or may not postulate values, theoretical entities, or nonphysical mental states. Those are the theories about which, in any field, realists about that field will argue with their opponents. But then why not take a theory of truthmakers to be a theory of truth itself: why not admit that truthmaker theory is really a correspondence theory of truth under another name? The reason is that a theory of truthmakers that is also a theory of truth will have to give all truths truthmakers; and this begs the question against moderate truthmaking by assuming that every truth is made true by something other than its entailment by another truth. It also begs the question of what makes propositions true, as a theory of truth can only give truthmakers for all propositions 〈P〉 by rewriting (EP) as 〈P〉 is true if and only if it is made true by P, where P is defi ned by its correspondence with 〈P〉. But this, as we have seen, begs the question against empiricists who think theoretical truths like 〈Neutrinos have no charge〉 are made true, not by corresponding theoretical facts but by quite different observable ones: and similarly in the other examples given in §3. It also invites a familiar vacuity objection to correspondence theories, as it cannot specify the fact P, which it says is what makes 〈P〉 true, independently of 〈P〉. It is precisely to escape this objection that truthmaker theorists need to supplement (EP) with independent theories of what truthmakers there are. But then these theories will only be theories of truthmakers, not of truth; and the only theory of truth they will need is (EP).
5
REALISM
There is, however, a vacuity objection to correspondence theories that does apply to truthmaker theories (Hornsby 2005). Consider the propositions that we use to say what make other propositions true: for example, propositions like 〈Fred is in brain state C〉 which, according to physicalists, tell us what make propositions like 〈Fred has toothache〉 true (Armstrong 1968a). If we now ask what makes 〈Fred is in brain state C〉 true, any nontrivial
278 D. H. Mellor answer will itself be a proposition that invites the same question—what makes it true?—thus starting an endless and arguably vicious regress. This objection has a hidden assumption, made explicit in Putnam’s attacks on what he calls “metaphysical realism,” the thesis that even our best scientific theories might be false because the world is not as they say it is. The assumption is that “we interpret our languages or nothing does” (Putnam 1980, 482), meaning that nothing in the world outside us constrains what our best theories are about. The argument for this is that any statement of an external constraint on what a theory is about, and in particular of what would make that theory true, merely extends the theory, whose extended form we can then always interpret so as to make it come out true. The right response to this argument is to reject the assumption that nothing other than ourselves can constrain what our scientific theories are about and hence what makes them true or false. In other words, realism needs realism. That is: the realism that recognises a nontrivial enterprise of discovering the truth about the world needs the traditional realism that recognises objective sameness and difference, joints in the world, discriminatory classifications not of our own making. (Lewis 1984, 228) This is the realism I am here taking for granted because without it the whole idea of nonpropositional truthmakers would make no sense nor, therefore, would the question of which truths have such truthmakers and which do not. Note, by the way, that this “traditional realism” does not require properties to be universals. The natural properties (temperatures, masses, charges, etc.) that embody objective samenesses and differences in our world could equally well be sets of exactly resembling particulars (Rodriguez-Pereyra 2002) or tropes (Williams 1953). For truthmaker theorists, any theory of natural properties (hereafter properties for short) will do, provided it does not reduce them to shadows of predicates privileged on merely linguistic grounds—as, for example, Goodman’s (1965, ch. IV) “projectibility” criterion of inductive respectability does. But that criterion is back to front anyway. Our using “blue” in inductive predictions is not what makes blue a property. It is the other way round: what makes “blue” a better predicate for this purpose than Goodman’s “grue” is that it corresponds, as “grue” does not, if not to a single property then at least to a disjunction of conjunctions of them (Mellor 1991).
6
TRUTHMAKERS AND TRUTH CONDITIONS
This distinction, between properties on the one hand, and predicates and hence concepts on the other, implies that, as Dyke (2007, ch. 2) and others
Truthmakers for What?
279
observe, a theory of truthmakers is no more a theory of meaning than it is of truth. In particular, therefore, the truthmakers of propositions must not be identified with their metalinguistic truth conditions. Take the physicalist thesis that all psychological propositions—about what we feel, see, think, want, intend, etc.—that have truthmakers have physical truthmakers. Neither this thesis nor its negation follows from the meanings of sentences saying that you see this, think that, or want or intend the other. For statements of the truth conditions of English sentences, like “Fred has toothache” is true if and only if Fred has toothache, are simply applications of the equivalence principle (EP) to sentences instead of propositions. And as such, they can no more tell us what would make these sentences true than (EP) itself can tell us what would make true the propositions they express. Physicalism about the mind can hardly be disproved by giving the truth conditions of sentences like “Fred has toothache” using the very same predicates that occur in those sentences. The idea that truth conditions link meaning and ontology derives from an ambiguity in the expression “giving a sentence’s truth conditions.” The ambiguity is between saying what would make the sentence true and using a Tarskian metalanguage to say when it is true, which need tell us nothing about what, if anything, makes it true. To see this, recall that it is only to protect socalled object languages from the Liar and other paradoxes that Tarski deports their semantic predicates, like “true” and “false,” into metalanguages that we can then safely use to say when object-language sentences are true (Tarski 1944, §9). And doing this does not require a metalanguage’s nonsemantic predicates to differ from those of its object language. Yet unless some of them do differ, a metalinguistic statement of a sentence’s truth conditions, like “Fred has toothache” is true if and only if Fred has toothache, will tell us no more than (EP) does about what makes “Fred has toothache” true. For a metalanguage to tell us that, its nonsemantic predicates must differ from those of its object language and do so in a way that gives ontological authority to its statements of truth conditions. But then it takes a nonlinguistic argument, like that from the so-called “causal closure” of physics (Papineau 2007, §1.6), to justify granting this authority to a sentence like “Fred has toothache” is true if and only if Fred is in a brain state (of type) C, which physicalists think does tell us what makes “Fred has toothache” true. Some philosophers, however, deny that any metalanguage can have such ontological authority. Carnap, for example, distinguishes two kinds
280 D. H. Mellor of ontological questions: those internal to a given language, or “linguistic framework,” and those external to it. Internal questions, raised within (say) a “thing language,” such as “Is there a white piece of paper on my desk?”; “Did King Arthur actually live?”; “Are unicorns real or merely imaginary?”; and the like . . . are to be answered by empirical investigations: From these questions we must distinguish the external question of the reality of the thing world itself . . . [This question] cannot be solved because it is framed in the wrong way. To be real in the scientific sense means to be an element of the system; hence this concept cannot be meaningfully applied to the system itself. (Carnap 1950, 242–243) In other words, although we can use a metalanguage of things to answer questions about whether certain pieces of paper, King Arthur, or unicorns exist, we cannot question the metalanguage’s ontology. To do so, by using a “meta-metalanguage,” only starts a hopeless regress: for we must stop somewhere if we are to give truth conditions at all; and wherever we stop, we cannot then question the ontology of the metalanguage we stop at. For Carnap, therefore, the justification for using a “thing language” lies in its utility, not in the existence of the entities it postulates. This doctrine seems to me both false and a false dichotomy. The dichotomy is false because the utility of our “thing language” depends on there being entities of most of the kinds it recognizes. What, after all, if not the existence, effects, and microstructure of water, makes the predicates “is water” and “is H 2O” more useful parts of our “linguistic framework” than “gruified” alternatives like “is watgin” (true of water in daylight and of gin at night)? And the doctrine is false because “to be real in the scientific sense” does not mean “to be an element of the system,” and Carnap no more shows that this is all it can mean than we saw in §4 that Putnam does. Indeed, given the traditional realism I am taking for granted, it is the other way around: it is the fi nding of new kinds of physical entity that requires physicists to add “elements” to their theoretical systems, not their additions that entitle us to call those elements real. Truth supervenes on being, not being on truth.
7
DIRECT AND INDIRECT TRUTHMAKING
Having defended truth’s supervenience on being, and argued against taking theories of truthmakers to be theories of meaning or of truth, I return to the fi rst of my two advertised questions: do all truths have truthmakers? Before saying why I think some do not, I propose to introduce some terminology. I showed in §2 how truth can supervene on being even if some
Truthmakers for What?
281
truths lack truthmakers. Yet to say that a truth that has no truthmaker is made true by “what there is and how it is” may still appear paradoxical. To mitigate that appearance I propose to say that whereas truths with truthmakers are made true directly, those that supervene on being without having truthmakers are made true indirectly. I take this terminology from that used to distinguish beliefs that perception gives us directly—as when I “believe what I see”—from beliefs that it gives us indirectly, by inference from those we get directly. And though this is indeed a different distinction, it is analogous enough to that between direct and indirect truthmaking to make the same terminology apt. One analogy can be illustrated by the fact that experienced microscopists can see directly what novices can only infer from the colored patches that are all they can see directly. Similarly in everyday life: like novices in microscopy, we infer truths about what we do not see directly from what we do see directly. And similarly with truthmaking: nonprimary truths may be (or seem to be) about entities like rainbows to which, as we saw in §4, no primary truths refer.5 The other analogy between perception and truthmaking is more important. To see it, note that we may draw the line between perception and inference in at least two ways. We may draw it in response to empirical facts about what people (say they) can or cannot perceive directly with or without instruments like microscopes. Or we may draw it on the basis of philosophical theories of perception, as in the theory that we only ever perceive sense data, from which all our other beliefs must therefore be inferred. But whichever way we draw the line, it can hardly imply that a belief’s perceptual justification depends on which side of the line our criterion makes it fall. The dependence here, if any, goes the other way: as in the argument that, because only beliefs about sense data can be fully justified, they are all that perception can give us directly, with all our other beliefs being inferred from them. Yet even if the distinction between perception and inference does have epistemological implications, that between primary and other truths has no analogous implications for truth, if only because truth, unlike justification, cannot come by degrees: nonprimary truths, that is, truths without truthmakers, are no less true than primary ones. So if, as assumed in §2, nonprimary propositions are complete truth functions of primary ones, nonprimary truths will also supervene on being, just as primary truths do. That is all I mean by saying that they too are made true, albeit indirectly, by what there is and how it is.
5 This analogy must not be pressed too far. Our inability to see microscopic entities may make the empiricists mentioned in §3 doubt their reality but not for the reasons mooted in §4 for declining to quantify over rainbows: it is not being unobservable that makes rainbows ontologically suspect.
282 D. H. Mellor 8
THE MERITS OF MODERATION
We have seen that taking truth to supervene on being no more forces us to credit all truths with truthmakers than deriving truth conditions from the equivalence principle (EP) does (§5). But nor does it require any truths to lack truthmakers. Why should we think that some do? Here are five reasons.
I
Ontological Economy
I remarked in §2 that truth’s supervenience on being requires truthmaking to be a generally cross-categorial relation, between a true proposition 〈P〉 and a nonpropositional entity S. From this it follows, as noted in §2, that there must be more to truthmaking than the fact, if it is a fact, that 〈S exists〉 entails 〈P〉. Yet as being entailed by another truth does ensure 〈P〉’s truth, and as many truths are entailed by others that are not of the form 〈S exists〉, it is not obvious that all such truths need nonpropositional truthmakers. And if they do not, then Ockham’s razor tells us not to credit them with truthmakers they do not need.
II
Negative Truths
Let 〈P〉 be a primary proposition, that is, one that does, if true, have a truthmaker, S. If 〈P〉 is false, then in classical logic its negation 〈¬P〉 will be true and therefore, if all truths have truthmakers, will have a different truthmaker, S´. Then the classical laws of noncontradiction and excluded middle, which tell us that 〈P〉 and 〈¬P〉 cannot both be true but that one of them must be, tell us that S´ can exist if and only if S does not exist. Yet why is this so, if S and S´ are distinct entities? To stipulate the incompatibility of S´ with S, without an independent reason to do so, is merely to restate in the language of ontology two laws of logic whose validity a theory of truthmakers should be able to explain. And so it can, provided it takes 〈¬P〉’s truth to follow, not from the existence of its own truthmaker S´ but from the nonexistence of 〈P〉’s truthmaker S. For the fact that 〈¬P〉 is true if and only if 〈P〉 is false follows from applying classical logic, not to all pairs of contradictory propositions but only to those where one member has the form 〈S exists〉, a form that all truthmaker theorists take to account, one way or another, for all truth-values. This being so, we can take the laws of noncontradiction and excluded middle to hold, where they do, as a consequence of two ontological facts: all entities must either exist or not and cannot do both; and the truthmaking relation between any S and any primary proposition 〈P〉, like all relations, can only relate entities that exist. This is why the truth-making relation can only make a primary proposition 〈P〉 true if its truthmaker S exists and why, if 〈¬P〉 is true if and only if S does not exist, 〈P〉 and 〈¬P〉
Truthmakers for What?
283
will satisfy the laws of noncontradiction and excluded middle. As will all complete truth functions of primary propositions, and hence, in the present view, all nonprimary propositions.6
III
Disjunctive Truths
If crediting true negations of primary propositions with truthmakers they do not need limits the explanatory power of truthmaker theory, crediting true disjunctions with them has more dire consequences. Take a true disjunction 〈P Q〉 of two primary propositions 〈P〉 and 〈Q〉 made true, if they are, by truthmakers S and T. What makes 〈P Q〉 true? Presumably, if 〈P〉 is true and 〈Q〉 is not, S does; and if 〈Q〉 is true and 〈P〉 is not, T does. In short, the truthmaker, if any, of a disjunction with only one true disjunct will also be what makes that disjunct true. Now suppose that 〈P〉 is true, 〈Q〉 is 〈¬P〉, and the disjunction 〈P ¬P〉 is made true by 〈P〉’s truthmaker S. But this disjunction is a necessary truth and therefore, in classical logic, is entailed by every proposition, including every truth that asserts the existence of any truthmaker. So if truthmakers make true every proposition their existence entails, then all of them are truthmakers for every necessary truth, including 〈P ¬P〉. But then, if the truthmaker of a disjunction with only one true disjunct also makes that disjunct true, it follows that all truths have the same truthmakers, namely all truthmakers. This result, which Restall (1996, 334) calls “truthmaker monism,” is, as he says, “not acceptable for any philosophically discriminating account of truthmakers.” Restall himself blocks the result with a restricted concept of “real entailment” that lets 〈P〉 really entail a proposition 〈R〉 only if, in every world, every truthmaker for 〈P〉 is a truthmaker for 〈R〉. But this presupposes the maximal truthmaker thesis that is here in question. For unless true negations have truthmakers, Restall will not let them “really entail” anything, which is absurd. This offers us a choice: to restrict classical entailment or to deny that true disjunctions have truthmakers and, in particular, that 〈P ¬P〉 has one. I think the reasons given in II for denying that true negations have truthmakers make this by far the better option.
6 Taking nonprimary propositions to be complete truth functions of primary ones may seem to presuppose the law of excluded middle by requiring all such propositions to be either true or false. And so it would if we excluded “neither true nor false” as a possible value of a complete truth function. But we need not do this. For if, as I argue here, all primary propositions are either true or false, no credible truth function of any number of them will ever be neither true nor false.
284 D. H. Mellor
IV
Conjunctive Truths
What makes the conjunction, 〈P Q〉, of any two primary propositions 〈P〉 and 〈Q〉 true? Because 〈P Q〉 is true if and only if 〈P〉 and 〈Q〉 are, its truthmaker needs to exist if and only if both their truthmakers, S and T, exist. The only entity that fits this bill is their so-called mereological sum, S+T, which, by definition, S and T compose. So 〈P Q〉 can only have a truthmaker if S+T exists; and all conjunctions of primary truths can only have truthmakers if the truthmakers of any number of these truths always have a mereological sum. But this principle, of unrestricted mereological composition (Lewis 1986, 211–213) is highly contentious and, I have argued, false (Mellor 2006). I see no independent reason to suppose that every two entities, however disparate—like me and the Second World War—compose a third. And without some such reason, it is gratuitous to postulate S+T just to provide a truthmaker for 〈P Q〉 when that conjunction is already entailed by 〈P〉 and 〈Q〉.
V
Necessary Truths
Although the case made in III for denying that true disjunctions have truthmakers applies to necessary truths like 〈P ¬P〉, it does not apply to all necessary truths and, in particular, not to truths of identity like 〈S = S〉. And that may seem right, as S seems as natural a truthmaker of 〈S = S〉 as it is of 〈S exists〉. Nevertheless, for the reason given in III, I propose to deny truthmakers to these truths too. For the fact that in classical logic 〈S = S〉, like all necessary truths, is entailed by every existential truth makes it as hard to say why S alone makes 〈S = S〉 true as it is to say why S alone makes true 〈S exists S does not exist〉.7 The only necessary truths that I think may have truthmakers are existential ones like “There are prime numbers” if, unlike “There is nobody in the house,” they cannot be paraphrased as negative existentials. For only by granting these truths some truthmakers can we avoid concluding that only contingent entities can exist at all. But that is not an issue I need to settle here. For as every necessary proposition is true in all possible worlds, and is therefore a complete truth function of every primary proposition, its truth will supervene on being anyway, albeit trivially: because its truth-value is the same in all possible worlds, it cannot have different truth-values in any two worlds that do not differ in being.
7 S might still have a unique ontological relation to 〈S = S〉 if that proposition, although true in all possible worlds where it exists, only exists in S-worlds. S could then be the truthmaker, if not for 〈S = S〉 itself, then at least for the related but contingent proposition 〈〈S = S〉 exists〉. But as whether this is possible depends on what propositions are—they could not, for example, be sets of possible worlds—I do not pursue the idea.
Truthmakers for What? 9
285
PRIMARY PROPOSITIONS: ATOMIC
Whatever truthmaker theorists say about necessary truths, their main job is to account for contingent ones, to which the rest of this chapter will therefore be confi ned. And then, setting aside necessary truths, my second advertised question is this: if all and only primary propositions are made true, if they are, by truthmakers, what determines which propositions these are? As Molnar (2000, 72) and others have argued, it cannot be the linguistic form of sentences in any everyday language. Take a language in which, by defi nition, the weather is fine if and only if it is not dull: the ontology of meteorology can hardly depend on which of the terms “fi ne” and “dull,” as applied to the weather, is used to defi ne the other. So this cannot be what determines which, if either, of the propositions 〈The weather is fi ne〉 and 〈The weather is dull〉 has, if true, a truthmaker, that is, is a primary proposition, of which the other is a nonprimary negation. What then does determine which propositions are primary? The answer to that question is implicit in the traditional realism, endorsed in §5, that “recognizes objective sameness and difference” and hence the contingent properties, such as masses, temperatures, charges, and durations, that embody this sameness in and difference between different things and events. If, then, a proposition 〈P〉 credits a thing or event a with some such property F,8 then 〈P〉 is a primary proposition which, if true, is made so by a’s being F; and similarly for contingent relations, such as the space–time separations of special relativity. What truthmakers like a being F are—tropes, combinations of particulars and universals, or something else again—depends on what contingent particulars, properties, relations, times, etc., are. But that is not our business, which is only to say what determines which entities there are, not of what kinds those entities are. This is why, as noted in §5, properties need not be universals: belief in truthmakers is compatible with any view of them “that does not reduce them to shadows of predicates privileged on merely linguistic grounds.” Equally, we can take any view of particulars that (i) is consistent with our view of properties and (ii) does not reduce them to shadows of terms—in this case singular terms—that are privileged on merely linguistic grounds. Within these limits, we can take particulars to be “bare,” haecceities, bundles of properties, tropes, or aggregates (Bigelow 1998). This is not to deny that what particulars there are may depend on what kinds of entities there are. If, for example, events are particulars (Davidson 1970a), there will be more particulars than if events are facts about
8 Or, if F is changeable, with being F at a time t, a qualification I hereafter take as read.
286
D. H. Mellor
changeable things; as there will if space–time or its points or regions are particulars (Nerlich 1994) or if temporally extended things have temporal parts (Hawley 2004). Conversely, if reality is limited to what is present (or present and past), there will be fewer particulars than if past, present, and future ones are equally real (Dyke 2005). These are contentious issues but not ones I need to settle here. Here I need only say how they affect which propositions are primary, which I say they do by affecting the output of a test that I call Quine’s test, derived from his (1948) criterion of “ontological commitment” (§4): the particulars that exist are those over which our first-order quantifi ers must range for any truth to be statable without using names or other singular terms (Mellor 1995, ch. 15.7). This, I maintain, is what determines which particulars, and of which kinds, are (constituents of) truthmakers and, therefore, part of what determines which propositions are primary. Similarly for properties if, as I and others hold (Shoemaker 1980; Mellor 1995, ch. 15), there is no more to them than the causation or the laws of nature they figure in; there is no more to masses than the laws of motion, gravity, etc. that contain them; no more to temperatures than the laws of thermodynamics, statistical mechanics, etc.; no more to beliefs, desires and other intentional mental states than the laws, if any, of intentional psychology; and so on. In this view of properties and laws, what properties there are can be determined by what I call Ramsey’s test, analogous to Quine’s test for particulars: the properties that exist are those over which our higher-order existential quantifiers must range for any law of nature to be statable without using predicates. In other words, the properties that exist are those that the existential quantifiers of the Ramsey sentence ∑ of all laws would have to range over for ∑ to be true (Mellor 1995, ch. 15.4–6).9 There is much more to be said about Quine’s and Ramsey’s tests, but only one of their implications is immediately relevant. This is that no complex (i.e., negative, disjunctive, or conjunctive) particulars pass Quine’s test, and no complex properties pass Ramsey’s test (Mellor 1995, §15.7). In Quine’s test, this is because quantifiers that range over particulars a and b need not also range over ¬a, a b or a b to enable us to state negative, disjunctive, or conjunctive truths about a or b without using names or other singular terms. Similarly for Ramsey’s test: existential quantifiers that range over properties F and G need not also range over ¬F, F G or F G to enable us to state laws involving negations, disjunctions, or conjunctions of F or G without using predicates. Both these tests could be modified to admit complexes of the particulars and properties that pass the unmodified tests, but the latter are better, not
9 ∑, unlike Ramsey’s own sentences (1929b), results from substituting existentially bound variables for all predicates, not just theoretical ones, in the conjunction of all true law statements.
Truthmakers for What?
287
only because they are more credible and economical but because their economy lets us extend the arguments of §8, for denying truthmakers to negative, disjunctive, and conjunctive truths, from propositions to sentences. Suppose “a is F” and “b is F” are sentences expressing primary propositions 〈Fa〉 and 〈Fb〉 which, if true, are made so by a being F and by b being F. Now suppose the sentence “a or b is F” is true, and we agree with §8 III that the disjunctive truth 〈Fa Fb〉 has no truthmaker. “a or b is F” could still have a truthmaker if we took it to say, not that a is F or that b is, but that what is F is the disjunctive particular a b. Quine’s test prevents this by excluding disjunctive particulars like a b, just as, by excluding conjunctive and negative ones, it prevents “a and b are F” being made true by a b being F and “a is not F” being made true by ¬a being F. Similarly for complex properties, whose common confounding with complex predicates like “is F or G” makes them more credible initially than complex particulars. Thus suppose the sentences “a is F” and “a is G” express primary propositions 〈Fa〉 and 〈Ga〉 which, if true, are made so by a being F and G respectively. Now suppose the sentence “a is F or G” is true, and we agree, as before, that the disjunctive truth 〈Fa Ga〉 has no truthmaker. “a is F or G” could still have a truthmaker if we took it to say, not that a is F or a is G, but that a has the disjunctive property F G. But this Ramsey’s test prevents, by excluding disjunctive properties like F G, just as, by excluding conjunctive and negative ones, it prevents “a is F and G” being made true by a’s being F G and “a is not F” being made true by its being ¬F.
10
PRIMARY PROPOSITIONS: MOLECULAR
The primary propositions proposed in §9 are those ascribing properties (including relations) Fx, Rxy, . . . that pass Ramsey’s test to particulars a, b, . . . that pass Quine’s test. That makes these propositions—〈Fa〉, 〈Rab〉, . . .—atomic: by which I mean that, unlike the molecular propositions— 〈¬Fa〉, 〈Fa Fb〉, 〈Fa Rab〉, . . .—whose truth-values theirs determine, they contain no other propositions. But not all primary propositions are atomic: many propositions contain others of which they are not complete truth functions. Take ascriptions of mental states like 〈X believes P〉, propositions like 〈Probably P〉, or counterfactuals like 〈¬P ⇒ ¬Q〉. Few if any of these molecular propositions are complete truth functions of their constituent propositions if those constituents are contingent: for generally, whether a contingent 〈P〉 is true or false, we may or may not believe it, its truth may or may not be probable and, if 〈P〉 and 〈Q〉 are true, 〈Q〉 might or might not have been true had 〈P〉 been false. So in the present view, as the truth-values of these and many other molecular propositions are not fixed by those of their constituents, they too are primary: they too will need truthmakers to make them true.
288
D. H. Mellor
Truthmaker theorists may of course deny that non-truth-functional molecular propositions need truthmakers by denying them truth-values at all. The rest of us, who think some of them do have truth-values, and are true, must say what in the world makes them so. Doing this is a task for specific theories of the mind, of probability, of conditionals, and so on, but I should at least mention some serious candidates. One such candidate has of course been mentioned already: physicalism, debates about which are debates about whether psychological truths, such as true instances of 〈X believes P〉, only ever have physical truthmakers. Similarly with metaphysical debates about probability. Suppose, for example, “Probably P” credits 〈P〉 with a greater chance of being true than 〈¬P〉, as in 〈The coin toss’s chance of landing heads >1/2〉, and suppose that this proposition is true. What makes it true will still depend on what chances are: for example, frequencies, actual or hypothetical, or propensities (Mellor 2005, ch. 3–4). That is what rival theories of chance aim to tell us: what makes propositions like 〈〈P〉’s chance of being true = p〉 true for a given p. And similarly for theories of subjective and of objective epistemic probability.10 Similarly again for non-truth-functional counterfactuals like 〈¬P ⇒ ¬Q〉. For Lewis (1973), what makes this counterfactual true, if it is, when 〈P〉 is false, is that 〈¬Q〉 is true in all the possible worlds most like ours where 〈¬P〉 is true. For those who think that only our world exists, these conditionals need actual truthmakers to make them true: for example, an object a having a mass M, which makes true all instances of 〈if a force F were applied to a that did not alter M, a would accelerate at F/M in the direction of F〉 (Mellor 2000, §5).
11
LAWS OF NATURE
Primary propositions may not all be atomic or molecular. Statements of the laws of nature, in particular, may be neither. What makes them true will depend, for a start, on whether laws are necessary or contingent. If they are necessary, that is, hold in all possible worlds, then in the present view, statements of them, like other necessary truths, will need no truthmakers. This admittedly contentious reading of law statements does therefore forestall the question of what does make them true, that is, of what laws are. But Ramsey’s test for properties gives this reading a less attractive implication, by making all properties figure in laws and thus exist, if at all, in all possible worlds, which puts them in the same ontological boat as numbers and other necessary entities, whose existence the present view gives us no truth-making reason to believe, as I noted in §8.
10 This assumes that objective epistemic probabilities exist and are contingent. In the present view, true ascriptions of them will need no truthmakers if they are necessary, as many of their advocates believe (Mellor 2005, ch. 6).
Truthmakers for What?
289
Truthmaker theorists may react to this result in several ways. It may be most palatable to resemblance nominalists like Rodriguez-Pereyra (2002), who in any case take properties to be sets of possible particulars and therefore to exist necessarily if at all. But even those who take properties to be universals may accept it, given that the necessity of 〈F exists〉 need not make 〈Fa〉, 〈Fb〉, . . . , necessary: particulars that are F in some worlds can still be (or have counterparts that are) ¬F in others. A simpler reaction is to deny that laws are necessary, thereby making the existence of most if not all properties unproblematic but reviving the question of what does make law statements true. One answer to that question follows from Mumford’s (2004, ch. 10) view of properties as embodying the laws that contain them. This lets laws be contingent, but only on properties, as for Mumford law statements not only entail but are entailed by the existence of the properties they contain. That at least gives a clear answer to the question of what makes law statements true: properties do. This answer will not, however, satisfy those who want properties to be as able to occur in different laws as particulars are to have different properties: to allow, for example, that the melting point of ice in other possible worlds could be slightly more or less than 0°C. They are better served by Ramsey (1928) and Lewis (1973, ch. 3.3), for whom what makes contingent truths like 〈All Fs are Gs〉 state laws is their being among the general “consequences of those propositions which we should take as axioms if we knew everything and organized it as simply as possible in a deductive system” (Ramsey 1929a, 150). As this only tells us which truths actually state laws, it does not prevent slightly different laws—for example, of the melting point of ice—containing the very same properties. The Ramsey–Lewis theory may also let law statements lack truthmakers altogether. For in this theory, all it takes to make 〈All Fs are Gs〉 true, whether it states a law or not, is that all actual Fs are Gs, that is, that every actual particular, a, b, . . . is either ¬F or G. But this generalization is arguably equivalent to the possibly infi nite conjunction (¬Fa Ga) (¬Fb Gb) . . . and thus a truth function of the primary propositions 〈Fa〉, 〈Ga〉, 〈Fb〉, 〈Gb〉, . . . (Ramsey 1927, 48–49). And if 〈All Fs are Gs〉 is such a truth function, then it will need no truthmaker to make it true.
12
GENERALIZATIONS
Whatever makes law statements true, other general truths raise a question that moderate truthmaker theorists can answer far more easily than their maximalist rivals. Suppose then that 〈All Fs are Gs〉 is a merely accidental truth, like 〈All members of the band have perfect pitch〉, where neither “F”
290 D. H. Mellor (“is a band member”) nor “G” (“has perfect pitch”) need correspond to natural properties. To see why this poses a problem for the maximalist thesis that all truths have truthmakers, suppose the band has just two members, a and b, both of whom are G. Now suppose the propositions 〈Ga〉 and 〈Gb〉 are both made true either directly, if “G” does correspond to a property, or indirectly, if 〈Ga〉 and 〈Gb〉 are nonprimary propositions entailed by other truths. What then makes 〈All Fs are Gs〉 true? The answer mooted at the end of §11, that this generalization is a complete truth function of its instances, and therefore needs no truthmaker, faces the objection that its instances, 〈Ga〉 and 〈Gb〉, do not entail it, because they do not entail that a and b are all the Fs there are. To get something that does entail it, we must conjoin to 〈Ga〉 and 〈Gb〉 the true proposition 〈There are no Fs except a and b〉. But as this truth is contingent—the band could have had more members—and is entailed by no other truths, it seems to need its own truthmaker. Yet how can a negative existential truth like 〈There are no Fs except a and b〉 have a truthmaker: what entity could there be whose existence entails that other entities do not exist? Maximalists have proposed various candidates: Armstrong (1997, ch. 13) and others follow Russell (1918, ch. 5) in postulating “totality facts” as truthmakers for true generalizations; Rosen and Lewis (2003) argue that what they call “the world qua-just-as-it-is” (i.e., the mereological sum of everything) can make it true that, for example, there are no unicorns. To all these proposed truthmakers for negative existential truths there are equally various objections, which any theory that credits all contingent truths with truthmakers needs to meet. In moderate truthmaker theories, on the other hand, negative existential truths need no truthmakers, because no negative truths need them (Heil 2000, §2). In these theories, all it takes to make 〈There are no Fs except a and b〉 true is that 〈Fa〉 and 〈Fb〉 are indeed the only true instances of 〈F . . . 〉. And then whatever makes 〈Ga〉 and 〈Gb〉 true, directly or indirectly, will also make 〈All Fs are Gs〉 true, even if 〈Ga〉 and 〈Gb〉 do not entail that generalization. If this exception to the rule that nonprimary truths are entailed by other truths seems surprising, it should not be, as it is an immediate consequence of denying truthmakers to negative truths. And once that is done, for example, for the reasons given in §8 II, the simple view it entails of what makes general truths true (and hence, on the Ramsey–Lewis view of laws, of what makes law statements true)—namely nothing but the truth of their instances—is yet another reason for preferring moderate truthmaker theories to maximalist ones.
References
Achinstein, P. (2001) The Book of Evidence, Oxford: Oxford University Press. Arabatzis, T. (2006) Representing Electrons, Chicago: University of Chicago Press. Armstrong, D. M. (1968a) A Materialist Theory of the Mind, London: Routledge and Kegan Paul. . (1968b) ‘The Headless Woman Illusion and the Defence of Materialism’, Analysis, 29: 48–49. . (1978) A Theory of Universals: Universals and Scientific Realism, vol. 2, New York: Cambridge University Press. . (1983) What Is a Law of Nature?, Cambridge: Cambridge University Press. . (1993) A Materialist Theory of the Mind, 2nd ed., New York: Routledge. . (1997) A World of States of Affairs, Cambridge: Cambridge University Press. . (2003) ‘Truthmakers for Modal Truths’, in Lillehammer and RodriguezPereyra (eds) (2003): 12–24. Ayer, A. J. (1935) ‘The Criterion of Truth’, Analysis, 3: 28–32. . (1936) Language, Truth and Logic, London: Dover. . (1946) Language, Truth and Logic, 2nd ed., Oxford: Oxford University Press. . (1963) The Concept of a Person and Other Essays, London: Macmillan. Azzouni, J. (1991) Metaphysical Myths, Mathematical Practice, Cambridge: Cambridge University Press. Bain, J. and Norton. J. (2001) ‘What Should Philosophers of Science Learn from the History of the Electron?’, in J. Buchwald and A. Warwick (eds) Histories of the Electron: The Birth of Microphysics, Cambridge MA: MIT Press, 2001, 451–465. Baker, A. (1975) Transcendental Number Theory, London: Cambridge University Press. Baldner, K. (1988) ‘Causality and Things in Themselves’, Synthèse, 77: 353–373. Barber, A. (1998) ‘The Pleonasticity of Talk About Concepts’, Philosophical Studies, 89: 53–86. Beall, JC (2001) ‘Is Yablo’s Paradox Non-Circular?’, Analysis, 61: 176–187. . (Forthcoming) Spandrels of Truth, Oxford: Oxford University Press. Beall, JC and Colyvan, M. (2001) ‘Heaps of Gluts and Hyde-ing the Sorites’, Mind, 110: 401–408. Beebee, H. and Dodd, J. (eds) (2005) Truthmakers: The Contemporary Debate, Oxford: Clarendon Press. Berkeley, G. (1710) A Treatise Concerning the Principles of Human Knowledge. Part I. Dublin. Bigelow, J. (1988) The Reality of Numbers: A Physicalist’s Philosophy of Mathematics, Oxford: Clarendon Press. . (1998) ‘Particulars’, in E. J. Craig (ed.) Routledge Encyclopedia of Philosophy, London: Routledge. Available online at
292 References Blackburn, S. (1984) Spreading the Word, Oxford: Oxford University Press. . (1998) ‘Wittgenstein, Wright, Rorty and Minimalism’, Mind, 107: 157–181. Boghossian, P. (1990) ‘The Status of Content’, Philosophical Review, 99: 157–184. . (1996) ‘Analyticity Reconsidered’, Noûs, 30: 360–391. . (2003) ‘Epistemic Analyticity: A Defense’, Grazer Philosophische Studien, 66: 15–35. Borg, E. (2007) ‘Minimalism versus Contextualism in Semantics’, in G. Preyer and G. Peter (eds) Context Sensitivity and Semantic Minimalism: New Essays on Semantics and Pragmatics, Oxford: Oxford University Press, 2007, 546–571. Bourne, C. (2006) A Future for Presentism, Oxford: Clarendon Press. Boyd, R. (1999) ‘Homeostasis, Species, and Higher Taxa’, in R. A. Wilson (ed.) (1999b): 141–185. Braddon-Mitchell, D. and Nola, R. (eds) (2009) Conceptual Analysis and Philosophical Naturalism, Cambridge MA: MIT Press. Brogaard, B. and Salerno, J. (2005) ‘Anti-realism, Theism and the Conditional Fallacy’, Noûs, 39: 123–139. Buchwald, J. Z. (1994) The Creation of Scientific Effects: Heinrich Hertz and Electric Waves, Chicago: University of Chicago Press. Buchwald, J. and Warwick, A. (eds) (2001) Histories of the Electron: The Birth of Microphysics, Cambridge MA: MIT Press. Bueno, O. and Colyvan, M. (2003) ‘Paradox Without Satisfaction’, Analysis, 63: 152–156. Burgess, J. A. (1997) ‘What Is Minimalism about Truth?’, Analysis, 54: 259–267. Candlish, S. (2005) ‘Truth, Identity Theory of’, in E. J. Craig (ed.) Routledge Encyclopedia of Philosophy, London: Routledge. Available online at
References 293 Da Costa, N. C. A. and Alves, E. H. (1977) ‘A Semantical Analysis of the Calculi C n’, Notre Dame Journal of Formal Logic, 18: 621–630. Dahl, P. F. (1997) Flash of the Cathode Rays: A History of J. J. Thompson’s Electron, Bristol and Philadelphia: Institute of Physics Publishing. Daly, C. (1998) ‘What Are Physical Properties?’, Pacific Philosophical Quarterly, 79 (3): 196–217. Darwin, C. (1859) The Origin of Species by Means of Natural Selection, New York: Barnes & Noble Classics, 2004. Davidson, D. (1968) ‘On Saying That’, Synthèse, 19: 130–146. Reprinted in D. Davidson, Inquiries into Truth and Interpretation, 2nd ed., Oxford: Clarendon Press, 2001, 93–108. . (1970a) ‘Events as Particulars’, Noûs, 4 (1): 25–32. Reprinted in D. Davidson, Essays on Actions and Events, Oxford: Clarendon Press, 1980, 181–187. . (1970b) ‘Mental Events’, in L. Foster and J. Swanson, (eds) Experience and Theory, Amherst: University of Massachusetts Press, 1970, 79–101. Reprinted in D. Davidson, Essays on Actions and Events, Oxford: Clarendon Press, 1980, 207–225. [Citations are to Davidson 1980.] . (1980) Essays on Actions and Events, Oxford: Clarendon Press. . (1993) ‘Thinking Causes’, in Heil and Mele (eds) (1993): 3–17. . (2001) Inquiries into Truth and Interpretation, 2nd ed., Oxford: Clarendon Press. De Queiroz, K. (1999) ‘The General Lineage Concept of Species and the Defi ning Properties of the Species Category’, in R. A. Wilson (ed.) (1999b): 49–89. Devitt, M. (1984) Realism and Truth, Oxford: Basil Blackwell. . (1991) ‘Minimalist Truth—Review Article on Paul Horwich, Truth, Oxford: Basil Blackwell, 1990’, Mind and Language, 6: 273–283. . (1997) Realism and Truth, 2nd ed. with ‘Afterword’, Princeton: Princeton University Press. . (1999) ‘A Naturalistic Defense of Realism’, in S. D. Hales (ed.) Metaphysics: Contemporary Readings, Belmont, CA: Wadsworth Publishing Company, 1999, 90–103. . (2001) ‘Incommensurability and the Priority of Metaphysics’, in P. Hoyningen-Huene and H. Sankey (eds) Incommensurability and Related Matters, Dordrecht: Kluwer Academic Publishers, 2001, 143–157. . (2002) ‘Underdetermination and Realism’, in E. Sosa and E. Villanueva (eds) Realism and Relativism: Philosophical Issues 12, Cambridge, MA: Blackwell Publishers, 2002, 26–50. . (2007) ‘Resurrecting Biological Essentialism’, Conference in Honour of Hilary Putnam’s 80th Birthday, University College Dublin. Devitt, M. and Sterelny, K. (1999) Language and Reality: An Introduction to the Philosophy of Language, 2nd ed. (1st ed. 1987), Cambridge, MA: MIT Press. Divers, J. and Miller, A. (1994) ‘Why Expressivists about Value Should Not Love Minimalism about Truth’, Analysis, 54: 12–19. . (1995a) ‘Minimalism and the Unbearable Lightness of Being’, Philosophical Papers, 24: 127–139. . (1995b) ‘Platitudes and Attitudes: A Minimalist Conception of Belief’, Analysis, 55: 37–44. Dodd, J. (2002) ‘Recent Work on Truth’, Philosophical Books, 43: 279–291. Dowell, J. L. (2006) ‘The Physical: Empirical, Not Metaphysical’, Philosophical Studies, 131: 25–60. Dreier, J. (2004) ‘Meta-Ethics and the Problem of Creeping Minimalism’, in J. E. Tomberlin (ed.) Philosophical Perspectives, 18, Metaethics, Oxford: Blackwell, 2004, 23–44. Dummett, M. (1959) ‘Truth’, Proceedings of the Aristotelian Society, 59: 141–162.
294
References
. (1991) The Logical Basis of Metaphysics, Cambridge, MA: Harvard University Press. Dupré, J. (1981) ‘Natural Kinds and Biological Taxa’, Philosophical Review, 90: 66–90. . (1993) The Disorder of Things, Cambridge, MA: Harvard University Press. . (1999) ‘On the Impossibility of a Monistic Account of Species’, in R. A. Wilson (ed.) (1999b): 3–22. Dyke, H. (2003) ‘Tensed Meaning: A Tenseless Account’, Journal of Philosophical Research, 27: 67–83. . (2005) ‘Time, Metaphysics of’, in E. J. Craig (ed.) Routledge Encyclopedia of Philosophy, London: Routledge. Available online at
References 295 . (1968) Die Grundlagen der Arithmetik, J. L. Austin (trans.) The Foundations of Arithmetic: A Logico-Mathematical Enquiry, 2nd Rev. ed. Evanston IL: Northwestern University Press, 1968. French, S. and Krause, D. (2006) Identity in Physics: A Historical, Philosophical, and Formal Analysis, New York: Oxford University Press. Geach, P. T. (1972) ‘Ascriptivism’, in his Logic Matters, Oxford: Blackwell, 1972. Ghiselin, M. T. (1974) ‘A Radical Solution to the Species Problem’, Systematic Zoology, 47: 350–383. Reprinted in M. Ereshefsky (ed.) (1992): 279–291. [Citations are to Ereshefsky.] Giaquinto, M. (2002) The Search for Certainty: A Philosophical Account of Foundations of Mathematics, Oxford: Clarendon Press. Gibbard, A. (2003) Thinking How to Live, Cambridge, MA: Harvard University Press. Goldstein, E. (1880) ‘On the Electric Discharge in Rarefied Gases’, Philosophical Magazine 10, Part I: 173–190, Part II: 234–247. Goodman, N. (1965) Fact, Fiction, and Forecast, New York: Bobbs-Merrill. . (1979) Ways of Worldmaking, Cambridge, MA: Harvard University Press. Grattan-Guinness, I. (2001) ‘The Interest of G. H. Hardy, F.R.S., in the Philosophy and the History of Mathematics’, Notes and Records of the Royal Society of London, 55 (3): 411–424. Griffiths, P. (1999) ‘Squaring the Circle: Natural Kinds with Historical Essences’, in R. A. Wilson (ed.) (1999b): 209–228. Grover, D. L., Camp, J. L., and Belnap, N. D. (1975) ‘The Prosentential Theory of Truth’, Philosophical Studies, 27: 73–125. Gupta, A. (2001) ‘A Critique of Deflationism’, in M. P. Lynch (ed.) (2001): 527–557. Hacking, I. (1983) Representing and Intervening, Cambridge: Cambridge University Press. Hale, B. and Wright, C. (2001) The Reason’s Proper Study: Essays towards a NeoFregean Philosophy of Mathematics, Oxford: Clarendon Press. Hardy, G. H. (1911) ‘Review of Principia Mathematica’, Times Literary Supplement, 321–322. . (1929a) ‘Hilbert’s Mathematical Logic’, Lecture at LeHigh University, January 11, 1929. Reported in Bulletin of the American Mathematical Society, 35: 280. . (1929b) ‘Mathematical Proof’, Mind, 38 (149): 1–25. . (1940) A Mathematician’s Apology, Cambridge: Cambridge University Press. Harman, G. (1977) The Nature of Morality, Oxford: Oxford University Press. Hawley, K. (2004) ‘Temporal Parts’, in E. N. Zalta (ed.) Stanford Encyclopedia of Philosophy, Fall 2004 edition. Available online at
296
References
. (1980) ‘Comments on Goodman’s Ways of Worldmaking’, Synthèse, 45: 139–199. Hill, C. S. (1991) Sensations: A Defence of Type Materialism, Cambridge: Cambridge University Press. Hofweber, T. (2005) ‘A Puzzle about Ontology’, Noûs, 39: 256–283. . (2006) ‘Inexpressible Properties and Propositions’, in D. Zimmerman (ed.) Oxford Studies in Metaphysics, vol. 2. Oxford: Oxford University Press, 2006, 155–206. . (2007) ‘Innocent Statements and Their Metaphysically Loaded Counterparts’, Philosophers’ Imprint, 7: 1–33. Available online at
References 297 Kant, I. (1929) Critique of Pure Reason, N. Kemp-Smith (trans), London: MacMillan. Khlentzos, D. (2004) Naturalistic Realism and the Antirealist Challenge, Cambridge, MA: MIT Press. Kim, J. (1973) ‘Causation, Nomic Subsumption and the Concept of Event’, Journal of Philosophy, 70: 217–236. Reprinted in J. Kim (1993b), 3–21. . (1984a) ‘Concepts of Supervenience’, Philosophy and Phenomenological Research, 45: 153–176. Reprinted in J. Kim (1993b), 53–78. . (1984b) ‘Epiphenomenal and Supervenient Causation’, Midwest Studies in Philosophy 9: 257–270. Reprinted in J. Kim (1993b), 92–108. . (1985) ‘Psychophysical Laws’, in E. LePore and B. McLaughlin (eds) Actions and Events, Oxford: Blackwell, 1985, 369–386. Reprinted in J. Kim (1993b), 194–215. . (1989) ‘The Myth of Nonreductive Materialism’, Proceedings and Addresses of the American Philosophical Association 63: 31–47. Reprinted in J. Kim (1993b), 265–284. . (1993a) ‘Can Supervenience and “Non-Strict Laws” Save Anomalous Monism?’ in Heil and Mele (eds) (1993), 19–26. . (1993b) Supervenience and Mind: Selected Philosophical Essays, Cambridge: Cambridge University Press. . (1993c) ‘The Nonreductivist’s Troubles with Mental Causation’, in Heil and Mele (eds) (1993), 189–210. Reprinted in J. Kim (1993b), 336–357. Kirkham, R. (1997) Theories of Truth, Cambridge, MA: MIT Press. Kitcher, P. (1984) ‘Species’, Philosophy of Science, 51: 308–333. Reprinted in P. Kitcher (ed.) In Mendel’s Mirror: Philosophical Refl ections on Biology, New York: Oxford University Press, 2003, 113–134. [Citations are to Kitcher 2003.] . (1993) The Advancement of Science: Science without Legend, Objectivity without Illusions, New York: Oxford University Press. . (2003) In Mendel’s Mirror: Philosophical Refl ections on Biology, New York: Oxford University Press. Knorr-Cetina, K. (1993) ‘Strong Constructivism—From a Sociologist’s Point of View’, Social Studies of Science, 23: 555–563. Kroon, F. (2001) ‘The Semantics of “Things in Themselves”: A Defl ationary Account’, Philosophical Quarterly, 51: 165–181. Kuhn, T. (1970) The Structure of Scientific Revolutions, 2nd ed. (1st ed. 1962), Chicago: Chicago University Press. Langton, R. (1998) Kantian Humility: Our Ignorance of Things in Themselves, Oxford: Clarendon Press. . (2004) ‘Elusive Knowledge of Things in Themselves’, Australasian Journal of Philosophy, 82: 129–136. Langton, R. and Lewis, D. (1998) ‘Defi ning “Intrinsic”’, Philosophy and Phenomenological Research, 58: 333–345. Lewis, D. (1966) ‘An Argument for the Identity Theory’, Journal of Philosophy, 63: 17–25. Reprinted in D. M. Rosenthal (ed.) Materialism and the Mind-Body Problem, Englewood Cliffs, N.J.: Prentice-Hall, 1971, 162–171. . (1970) ‘How to Defi ne Theoretical Terms’, Journal of Philosophy, 67: 427–446. Reprinted in his Philosophical Papers, vol. 1, New York: Oxford University Press, 1983, 78–95. . (1973) Counterfactuals, Oxford: Blackwell. . (1979) ‘Prisoners’ Dilemma Is a Newcomb Problem’, Philosophy and Public Affairs, 8 (3): 235–240. Reprinted in R. Campbell and L. Sowden (eds) Paradoxes of Rationality and Cooperation, Vancouver: University of British Columbia Press, 1985, 251–274.
298
References
. (1984) ‘Putnam’s Paradox’, Australasian Journal of Philosophy, 62: 221– 236. Reprinted in his Papers in Metaphysics and Epistemology, Cambridge, Cambridge University Press, 1999, 56–77. . (1986) On the Plurality of Worlds, Oxford: Blackwell. . (1991) Parts of Classes, Oxford: Basil Blackwell. . (1994) ‘Reduction of Mind’, in S. Guttenplan (ed.) A Companion to Philosophy of Mind, Oxford: Blackwell, 1994, 412–431. Reprinted in his Papers in Metaphysics and Epistemology, Cambridge, Cambridge University Press, 1999, 291–324. . (2003) ‘Things qua Truthmakers’, in H. Lillehammer and G. RodriguezPereyra (eds) (2003): 25–38. . (2009) ‘Ramseyan Humility’, in D. Braddon-Mitchell and R. Nola (eds), Conceptual Analysis and Philosophical Naturalism, Cambridge, MA: MIT Press. Lillehammer, H. and Rodriguez-Pereyra, G. (eds) (2003) Real Metaphysics: Essays in Honour of D. H. Mellor, New York: Routledge. Lynch M. P. (ed.) (2001) The Nature of Truth, Cambridge, MA: MIT Press. . (2004) ‘Truth and Multiple Realizability’, Australasian Journal of Philosophy, 82: 348–408. Macarthur, D. and Price, H. (2007) ‘Pragmatism, Quasi-realism and the Global Challenge’, in C. Misak (ed.) The New Pragmatists, Oxford: Oxford University Press. Mackie, J. L. (1977) Ethics: Inventing Right and Wrong, Harmondsworth: Penguin. Maddy, P. (1990) Realism in Mathematics, Oxford: Clarendon Press. . (1997) Naturalism in Mathematics, Oxford: Clarendon Press. Mallet, J. (1995) ‘A Species Defi nition for the Modern Synthesis’, Trends in Ecology and Evolution, 10: 294–299. Maris, H. J. (2000) ‘On the Fission of Elementary Particles and the Evidence for Fractional Electrons in Liquid Helium’, Journal of Low Temperature Physics, 120 (3 & 4): 173–204. Martin, C. B. (1996) ‘How It Is: Entities, Absences and Voids’, Australasian Journal of Philosophy, 74: 57–65. Maxwell, G. (1978) ‘Rigid Designators and Mind-Brain Identity’, in C. Wade Savage (ed.) Perception and Cognition: Issues in the Foundations of Psychology, Minnesota Studies in the Philosophy of Science, vol. 9, Minneapolis: University of Minnesota Press, 365–404. Mayr, E. (1969) Principles of Systematic Zoology, New York: McGraw Hill. . (1982) The Growth of Biological Thought, Cambridge, MA: Harvard University Press. McCloskey, M. (1983) ‘Naïve Theories of Motion’, in D. Gentner and A. L. Stevens (eds) Mental Models, Hillsdale, NJ: Lawrence Erlbaum Associates, 299–324. McGee, V. (1991) Truth, Vagueness, and Paradox, Indianapolis: Hackett. McLaughlin, B. P. (1989) ‘Type Epiphenomenalism, Type Dualism, and the Causal Priority of the Physical’, Philosophical Perspectives, 3: 109–135. McMullin, E. (2003) ‘van Fraassen’s Unappreciated Realism’, Philosophy of Science, 70 (3): 455–478. Mellor, D. H. (1973) ‘Materialism and Phenomenal Qualities’, Proceedings of the Aristotelian Society, supp. vol., 47: 107–119. . (ed.) (1990) Ramsey: Philosophical Papers, Cambridge: Cambridge University Press. . (1991) ‘Properties and Predicates’, in his Matters of Metaphysics, Cambridge: Cambridge University Press, 1991, 170–182. Reprinted in D. H. Mellor and A. Oliver (eds) (1997b): 255–267.
References 299 . (1995) The Facts of Causation, London: Routledge. . (2000) ‘The Semantics and Ontology of Dispositions’, Mind, 109: 757– 780. . (2004) ‘Wahrmacher’, in I. U. Dalferth and P. Stoellger (eds) Wahrheit in Perspektiven: Probleme einer offenen Konstellation, Tübingen: Mohr Siebeck, 2004, 103–118. . (2005) Probability: A Philosophical Introduction, London: Routledge. . (2006) ‘Wholes and Parts: The Limits of Composition’, South African Journal of Philosophy, 25: 138–145. Mellor, D. H. and Oliver, A. (1997a) ‘Introduction’, in D. H. Mellor and A.Oliver (eds) (1997b): 1–33. . (eds) (1997b) Properties, Oxford: Oxford University Press. Melnyk A. (1997) ‘How to Keep the “Physical” in Physicalism’, Journal of Philosophy, 94: 622–637. . (2003) A Physicalist Manifesto: Thoroughly Modern Materialism, Cambridge: Cambridge University Press. Miller, A. (2005) ‘Realism’, in E. N. Zalta (ed.) The Stanford Encyclopedia of Philosophy, Fall 2005 Edition. Available online at
300
References
Nover, H. and Hájek, A. (2004) ‘Vexing Expectations’, Mind, 113: 237–249. O’Hara, R. J. (1993) ‘Systematic Generalization, Historical Fate, and the Species Problem’, Systematic Biology, 42: 231–246. Okasha, S. (2002) ‘Darwinian Metaphysics: Species and the Question of Essentialism’, Synthèse, 131: 191–213. O’Leary-Hawthorne, J. and Oppy, G. (1997) ‘Minimalism and Truth’, Noûs, 31: 170–196. O’Leary-Hawthorne, J. and Price, H. (1996) ‘How to Stand Up for Non-Cognitivists’, Australasian Journal of Philosophy, 74: 275–293. Papineau, D. (2002) Thinking about Consciousness, Oxford: Clarendon Press. . (2007) ‘Naturalism’, in E. N. Zalta (ed.) The Stanford Encyclopedia of Philosophy, Spring 2007 Edition. Available online at
References 301 . (1929b) ‘Theories’, in R. B. Braithwaite (ed.) The Foundations of Mathematics and other Logical Essays, London: Routledge and Kegan Paul, 1931, 212–236. Reprinted in D. H Mellor (ed.) (1990), 112–136. Ravenscroft, I. (1997) ‘Physical Properties’, The Southern Journal of Philosophy, 35: 419–431. Resnick, M. (1997) Mathematics as a Science of Patterns, Oxford: Clarendon Press. Restall, G. (1996) ‘Truthmakers, Entailment and Necessity’, Australasian Journal of Philosophy, 74: 331–340. Rodriguez-Pereyra, G. (2002) Resemblance Nominalism, Oxford: Clarendon Press. . (2005) ‘Why Truthmakers’, in H. Beebee and J. Dodd (eds) (2005), 17–31. Rose, N. (1988) Mathematical Maxims and Minims, Raleigh NC: Rome Press Inc. Rosen, G. and Lewis, D. (2003) ‘Postscript to “Things qua Truthmakers”: Negative Existentials’, in H. Lillehammer and G. Rodriguez-Pereyra (eds) (2003), 39–42. Russell, B. (1903) The Principles of Mathematics, London: George Allen and Unwin. . (1918) The Philosophy of Logical Atomism, La Salle, Illinois: Open Court. Sainsbury, M. (2005) ‘Pleonastic Explanations’, Mind, 114: 97–111. Schiffer, S. (1994) ‘A Paradox of Meaning’, Noûs, 28: 279–324. . (1996) ‘Language-Created Language-Independent Entities’, Philosophical Topics, 24 (1): 149–167. . (2003) The Things We Mean, Oxford: Oxford University Press. Searle, J. (1969) Speech Acts, Cambridge: Cambridge University Press. Shoemaker, S. (1980) ‘Causality and Properties’, in P. van Inwagen (ed.) Time and Cause, Dordrecht: D. Reidel, 1980, 109–135. Reprinted in D. H Mellor and A. Oliver (eds) (1997b): 228–254. Simons, P. (2000) ‘Truth-maker Optimalism’, Logique et Analyse, 43 (169–170): 17–41. Smart, J. J. C. (1978) ‘The Content of Physicalism’, Philosophical Quarterly, 28: 339–341. . (1986) ‘How to Turn the Tractatus Wittgenstein into (almost) Donald Davidson’, in E. LePore (ed.) Truth and Interpretation: Perspectives on the Philosophy of Donald Davidson, Oxford: Basil Blackwell, 1986, 92–100. Smith, M. (1994a) ‘Minimalism, Truth-Aptitude and Belief’, Analysis, 54: 21–26. . (1994b) ‘Why Expressivists About Value Should Love Minimalism About Truth’, Analysis, 54: 1–12. Smith, N. J. J. (2000) ‘The Principle of Uniform Solution (of the Paradoxes of SelfReference)’, Mind, 109: 117–122. Soames, S. (1999) Understanding Truth, New York: Oxford University Press. Sober, E. (1980) ‘Evolution, Population Thinking and Essentialism’, Philosophy of Science, 47: 350–383. Reprinted in E. Sober, From a Biological Point of View: Essays in Evolutionary Philosophy, Cambridge: Cambridge University Press, 1994, 201–232. [Citations are to Sober 1994.] Sokal, R. and Crovello, T. (1970) ‘The Biological Species Concept: A Critical Evaluation’, American Naturalist, 104: 127–153. Reprinted in M. Ereshefsky (ed.) (1992): 27–55. [Citations are to Ereshefsky.] Sorensen, R. A. (1998) ‘Yablo’s Paradox and Kindred Infi nite Liars’, Mind, 107: 137–155. . (2001) Vagueness and Contradiction, Oxford: Clarendon Press.
302
References
Stalnaker, R. (1996) ‘On What Possible Worlds Could Not Be’, in A. Morton and S. Stich (eds) Benacerraf and His Critics, Oxford: Basil Blackwell, 1996, 103–119. Stanford, P. K. (1995) ‘For Pluralism and Against Realism about Species’, Philosophy of Science, 62: 70–91. Sterelny, K. (1999) ‘Species as Ecological Mosaics’, in R. A. Wilson (ed.) (1999b), 119–138. Sterelny, K. and Griffiths, P. (1999) Sex and Death, Chicago: University of Chicago Press. Stoljar, D. (1993) ‘Emotivism and Truth-Conditions’, Philosophical Studies, 70: 81–101. . (2001) ‘Physicalism’, in E. N. Zalta (ed.) Stanford Encyclopedia of Philosophy, Winter 2005 edition. Available online at
References 303 van Valen, L. (1976) ‘Ecological Species, Multi-Species and Oaks’, Taxon, 25: 233–239. Reprinted in M. Ereshefsky (ed.) (1992): 69–77. [Citations are to Ereshefsky.] Varzi, A. (1999) An Essay in Universal Semantics, vol. 1 of Topoi Library, Boston: Kluwer Academic Publishers. Vinueza, A. (2001) ‘Realism and Mind Independence’, Pacific Philosophical Quarterly, 82: 51–70. Walton, K. (1990) Mimesis as Make-Believe, Cambridge, MA: Harvard University Press. . (1993) ‘Metaphor and Prop Oriented Make-Believe’, European Journal of Philosophy, 1: 39–57. Weatherson, B. (2005) ‘True, Truer, Truest’, Philosophical Studies, 123: 47–70. Whittaker, E. (1951) A History of Theories of Aether and Electricity, vol. 1, London: Thomas Nelson. Wiggins, D. (1987) Needs, Values, Truth, Oxford: Basil Blackwell. . (2001) Sameness and Substance Renewed, Cambridge: Cambridge University Press. Williams, D. C. (1953) ‘On the Elements of Being: I’, Review of Metaphysics, 7: 3–18. Reprinted in D. H. Mellor and A. Oliver (eds) (1997b): 112–124. Williamson, T. (1994) Vagueness. London: Routledge. Wilson, J. (2006) ‘On Characterizing the Physical’, Philosophical Studies, 131: 61–99. Wilson, M. L. (1994) ‘Can We Trust Logical Form?’, Journal of Philosophy, 91: 519–544. Wilson, R. A. (1999a) ‘Realism, Essence, and Kind: Resuscitating Species Essentialism?’, in R. A. Wilson (ed.) (1999b), 187–207. . (ed.) (1999b) Species: New Interdisciplinary Essays, Cambridge, MA: MIT Press, 1999. Wittgenstein, L. (1953⁄1968) Philosophical Investigations, G. E. M. Anscombe (trans.), Oxford: Basil Blackwell. Woodward, J. (2003) Making Things Happen: A Theory of Causal Explanation, Oxford: Oxford University Press. Wright, C. (1983) Frege’s Conception of Numbers as Objects, Aberdeen: Aberdeen University Press. . (1992) Truth and Objectivity, Cambridge, MA: Harvard University Press. . (1994) ‘Realism, Pure and Simple? A Reply to Timothy Williamson’, International Journal of Philosophical Studies, 2: 327–341. . (1998) ‘Comrades Against Quietism: Reply to Simon Blackburn on Truth and Objectivity’, Mind, 107: 183–203. . ( 2001) ‘Minimalism, Deflationism, Pragmatism, Pluralism’, in Lynch (ed.) (2001): 751–787. Yablo, S. (1993) ‘Paradox Without Self-Reference’, Analysis, 53: 251–252.
Index
A absolute minimal entities, 75; see also minimal entities abstraction, 29, 35, 79–82, 256 ambiguity argument, 267–9 analytical propositions, 169, 174 anomalous monism, 85–98 antirealism, 3, 4, 44, 47, 51–65, 68, 81, 124, 126, 135, 158, 160, 162–8, 175, 178–80, 199 antisatisfaction, 27 Aquinas, Thomas, 133 Aristotle, 69, 133, 255, 259, 260 Armstrong, David, 53n., 117 A-theory, 13 atomic contraries, 27, 28 atomic predicates, 21, 23, 28 atomic propositions, 285–7 Austin, J. L., 85 Ayer, A. J., 127
B Baker, Alan, 119 “bald man” paradox 29, 30 Beall, JC, 2 Benaceraff, Paul, 76 Berkeley, George, 82, 232, 233, 252 Berry’s paradox, 33 biological realism, 43–65 biological species concepts, 44–7 biology, 4–5, 43–65 Bishop, Errett, 112 bivalence, 2, 3, 18–20, 22, 26, 27 Bohr, Niels, 10, 233, 243, 244, 247, 250, 251 borderline cases, 2, 30, 38, 39 Boyd, Richard, 50n., 52n., 62n. “brain-in-a-vat” skepticism, 237n.– 238n.
Broad, C. D., 195 B-theory, 13, 67 Burali-Forti paradox, 31, 33 Buridan, Jean, 187n.
C Canberra plan, 218, 225 Carnap, Rudolf, 85, 213, 214, 279, 280 cathode rays, 10, 201–4, 209–12, 216–17, 219–26, 228 causal closure, 279 causal interaction, 90 causal relevance, 91 Chalmers, David, 88, 237n. Cheyne, Colin, 120 circularity, 34 cladistics, 61, 64 coincidence, 255, 257, 260, 261 color, 99–111, 218 common-ground approach, 138 commonsense, 64, 164, 165, 167, 175, 200, 237; realism, 165n. concepts, 7, 104, 117, 256, 257, 278; biological species, 44–7; Kantian, 256; phenetic, 44; phylogenetic-cladistic, 45; sortal, 12, 81, 254–7, 260, 262, 268, 269, 271 conceptual idealism, 232, 233, 243, 246 confusion hypothesis, 25 constructive empiricists, 200 constructivism, 48, 49, 54n., 200 contextualism, 123n. contraposition, 24, 27 corpuscular model of electrons, 222–3, 226–8 correspondence theories, 13, 274, 277 counterfactuals, 287–8 Cracraft, Joel, 59
306
Index
creeping minimalism, 8, 9, 123–80 creeps, 66–8, 81, 82 Crookes, William, 202, 221, 222 cross-categorial relations, 273 Curry’s paradox, 30, 33 cutoffs, 18, 25
D Davidson, Donald, 6, 89n., 90–8, 119, 120 deflationism, 127, 131, 132 De Moivre’s theorem, 116 denotation, 201, 205, 206–8, 211–18, 220, 223, 225, 228–31 Descartes, René, 49, 88, 89, 195 determinables, 88 Devitt, Michael, 4, 5, 100 dialectic space 36–8 disciplined syntacticism, 143 discourse, 4, 11, 75, 113, 133–4, 139, 142–4, 157, 159, 160, 198 Doppler effect, 222 Dreier, James, 123–7, 144, 159–61, 163–4, 177–80 D-thesis, 12, 254–67, 271 dualism, 85–8, 96
E ecological niches, 45 electrons, 198–231 electron theory, 198–231 eliminativism, 85, 86, 87 empiricism, 119, 200, 207n. entailment, 13, 273–7, 283 epistemic gaps, 19 epistemicism, 20 epistemology, 58, 99, 100, 102, 116, 118–20, 162, 170, 193, 213n., 239, 242 equivalence principle, 13, 274, 275, 282 equivalence relations, 80 equivalence schema, 8, 128, 133n., 148n., 149 equivocation, 107 Ereshefsky, Marc, 4, 44, 47, 51–7 error theory, 126, 157, 160, 161, 162, 175, 178, 179, 180 eternal sentences, 136n. ether theory, 206, 210, 213, 220–5 evidence, 4, 7, 49, 99–103, 105–7, 135, 151, 165–7, 170–1, 185, 193, 198, 223, 226, 238, 242, 243, 267
existence, 53; mind-independent, 4; of properties, 7; realism and, 4, 48; sortal-dependence of 258 existence dimension, 48; existential quantification, 152, 153, 212, 214, 286 explanations, 50, 55, 56, 59, 60, 62, 63, 65, 94, 125, 142, 156, 177–8, 221 explanatory realism, 4, 50, 55 explanatory significance, 4, 50, 53, 65 expressivism, 142, 143, 158, 159n., 161
F fallacies, of appeal to authority, 120; of equivocation, 107 falsity, 1, 17, 19, 26, 34, 36, 196 fictionalism, 119 fictionality, 7, 72–5 Field, Hartry, 81, 113, 115, 215 “fig-leaf” realism, 48, 100 Fodor, J. 85 Frege, Gottlob, 117, 127n., 169 functionalism, 87, 101, 102
G gaps, 2–3, 17, 18–25, 37–40, 49 Gem, the, 10–11, 232–53 genera, 44, 60 generalizations, 289–90 Goldstein, Eugen, 209, 210, 218, 220, 221n., 224, 229 Goodman, Nelson, 181 Grelling’s paradox, 31, 33 Griffiths, Paul, 45, 61, 62
H Hacking, I., 199, 200, 205 Hale, Bob, 117, 118 Hamiltonians, 230–1 Hardy, G. H., 112, 114, 115, 119–20 “heap” paradox, 29, 30 Hellman, Geoffrey, 182 Hempel, Carl G., 9 Hempel’s dilemma, 9, 181–97 heterogeneity argument, 52 Hilbert, David, 120 Hittorf, Johann, 209, 218, 229 Hobbes, Thomas, 195 Horwich, Paul, 8, 124, 126, 127–135, 137n., 141, 142, 178 Hull, David, 43, 46, 47, 52
Index Hume’s principle, 143n. “humility”, Kantian, 239–43; quaentities and, 243–6 Hyde, Dominic, 25 hyphenated entities, 11, 233–5, 237, 252 hypothetico-deductive method, 116
I idealism, 11, 48, 49, 232, 236; conceptual 232, 233, 243, 246 identity, 12, 94, 110–11, 206, 236n., 239, 254, 255, 274, 284 idiomatic equivalence, 68 impetus, 9, 101, 187, 188–92 impredicativity, 31 inclosure schema, 32–8 independence dimension, 48 individuation, 12, 254–71 internal quantifiers, 154 irrealism, 142, 158, 178
J judgments, biological, 4–5; theoryladenness, 4, 57, 58, 65
K “Kantian humility”, 239–43 Kant, Immanuel, 48, 232, 236, 239–53, 256 Kim, Jaegwon, 6, 90, 92, 97 Kitcher, Philip, 44, 47, 50, 52, 53–8, 63, 65 knowledge, 2, 19, 20, 58, 72, 75, 77, 80, 94, 101, 103–5, 108, 109, 116–19, 123, 138, 140, 141, 143, 145, 152, 159, 161, 162, 164, 166–7, 169, 194–5, 198, 200, 210, 213, 216, 218, 229, 233, 236–43, 245–6, 250, 252, 258, 264, 266, 270 Kreisel, G., 120 Kroon, Fred, 10, 11 Kuhn, Thomas, 10, 48, 243–46, 253
L Langton, Rae, 239, 240, 241, 242 language, 6, 8, 9, 10, 12, 66–84, 97, 152, 170, 171, 178–9, 181, 197, 198, 278, 280; games, 75, 78, 83; minimal entities in, 126, 170; philosophy of, 99; stars in, 22, 24; transparency, 19;
307
unsettledness of, 17–28; see also metalanguage language dependence, 6, 8, 72, 76, 97, 168–70, 177, 178 laws of nature, 288–9 Leibniz, Gottfried, 100, 101, 107, 108, 110, 111 Lewis, C. I., 240n. Lewis, David, 117, 242 liar paradoxes, 3, 29, 30, 31–40, 279 linguisticized metaphysics, 5, 85–7 linguistic physicalism, 9 linguistic turn, 9, 85, 181, 183, 196–7 logicism, 32, 113 Lotke-Volterra equations, 62n.
M Maddy, Penelope, 116 Mallett, James, 60–1 materialism, 94, 109 mathematical entities 7; see also numbers Maxwell, Grover, 215–17 Mayr, Ernst, 44 meaning, 13 mechanisms 32, 35–6, 55 Meinongian semantics, 119 Mellor, Hugh, 12 metaethics, 124, 166 metalanguage, 13, 279, 280 metaphysics, 10, 12, 99; linguisticized, 5, 85–7 Mill, J. S, 240 mind-body problem, 88, 91 mind-brain identity theory, 101, 111 mind-dependence/independence, 4, 160, 162, 164, 165, 167, 170, 174, 179, 198 minimal entities, 8, 9, 75, 123–80 minimalism, 7–9, 11, 66, 123–80 Minimalist theory of truth, 127, 128, 129, 130 Mishler, Brent, 62n., 64 Modern Synthesis, 60 modes, 88 molecular predicates, 27 monism, 6, 43–5, 47, 57n., 85–98, 133, 155, 156 monophyletic groups, 61 moral realism, 166–7 Moretti, Luca, 7–8, 11 M-thesis, 267, 268 Musgrave, Alan, 5, 11, 232–243, 246, 252, 253, 271, 274
308 Index N naturalism, 50, 117, 118, 119 necessary existence, 174 negation, 17, 19, 21–8, 34, 143, 279, 282–3, 285, 286 Neurath, Otto, 181, 182 Newcomb problem, 37 Newtonian theory, 10, 11, 234, 243, 244, 250, 251 Nola, Robert, 10 nominalism, 7, 49, 84, 97, 98, 119, 120 nominalization, 147, 148, 151, 175 nomological character of causation, 90 noncognitivism, 125, 158, 159, 160, 161 nonepistemic approach, 17 nonidentity, 101 nonobjectual quantifiers, 154 nonphysical qualia, 117 nonrealism, 8, 9, 124, 125, 157–8, 168–80 nonreductive physicalism, 90–1 N-rays, 208 numbers, 1, 7, 80, 82, 112, 118–20, 159, 162, 179–80, 227, 236, 284, 288; see also mathematical entities
O Okasha, Samir, 44, 45 ontological commitment, 275–7, 286 ontological economy, 282 ontological maximalism, 66 ontological minimalism, 66, 69, 76 ontology, 13, 70, 86, 97, 120, 140, 156, 207, 210; pleonastic, 146–57 O-terms (observation terms), 211, 212, 214 O-theories, 200n.
P paradoxes, 3, 29–40, 279 particle models, 222 Pasadena paradox, 37 Peano’s axioms, 118, 119 perception, 198, 264, 281 phenetic concepts, 44 phenomenal experimentation, 217–19 phenomenal properties, 219 phenomena-noumena distinction, 239, 241 phylogenetic-cladistic concepts, 45 physicalism, 9, 90–1, 95–8, 181–97, 288
physical theory, 9, 182, 183–4, 186, 189–92, 195–7 planetary model of spin, 219n. platitudes, 135–8, 141, 153, 154, 155, 157, 218 Platonism, mathematical, 112–20, 151n., 179; pleonastic, 5, 66–84; Quinean, 7, 116 pleonastic entities, 5, 8, 70–2, 75–8, 147, 151, 153, 168, 172 pleonastic ontology, 146–57 pleonastic Platonism, 5, 66–84 pleonastic transformations, 5, 68–74, 79, 171 Plücker, Julius, 202–9, 217, 219, 230 pluralism, 43–5, 47, 53, 54, 55, 57n., 58, 133, 134, 135, 142n., 156n., 159n., 177 pluralistic realism, 54, 55 plural quantification, 117 Poland, Jeffrey, 182, 185n. positive predicates, 21, 27 potential evidence transcendence, 165 predicates, 2, 5–7, 27, 53n., 85–97, 128, 143, 151, 153, 172–3, 183– 8, 191, 256, 278–80, 285–7; atomic, 21, 23, 28 Priest, Graham, 32, 36, 37, 40n. properties, 6, 7, 85–96, 104–10, 132, 146–150, 152–3, 155–6, 159, 162, 166–8, 173, 174, 177, 179, 183–92, 196, 198, 205,206, 209, 210, 215, 216, 219, 225– 31, 235, 236, 238–44, 248–51, 253, 278, 285–90 property dualism, 87 propositions, 2, 5, 13, 70, 72, 76, 83, 105, 106, 116, 128–32, 136, 138, 146–52, 155, 174, 177–8, 217, 237, 273–90, atomic, 285–7; identity, 255, 258; molecular, 287–8; primary, 273, 275, 281–5, 287–90 prosententialists, 127, 128 psychological thesis, 12 Putnam, Hilary, 85, 87, 135, 155n., 278
Q qua-entities, 233, 236, 243, 246, 249, 253 qua-names, 233, 236, 241, 245–50, 252–3 quantification, 147, 148 quantum theory, 86n.
Index quasi-realism, 142, 143, 158, 159, 160 “Quine’s test”, 286 Quine, Willard van Orman, 9, 48, 79, 80, 85, 86, 97, 104, 106, 107, 112–14, 169, 196, 197, 286
R Ramsey, F. P., 114–16, 127n. Ramsey–Lewis theory, 201, 211–17, 218, 220, 225, 228, 289 “Ramsey’s test”, 286, 288 Ramsification, 211–13, 217, 222–3 realism, 43–54, 83, 86, 93–8, 157–80, 198, 207, 277–8; about entities/ theories, 199; and antirealism, 3, 43, 47, 126, 135, 160, 162–8, 178, 180; and bivalence, 3; and existence, 4, 48; and knowledge, 236–9; and minimalism, 8; and nonrealism, 8, 9, 125, 157–8, 168–80; and pluralism, 43, 47, 54, 55; and truth 1, 7–14, 158; as semantic doctrine, 4; biological 43–65; commonsense, 165; explanatory, 4, 50, 55; “fig-leaf”, 48, 100; kinds of, 10, 199; metaphysical, 237; moral, 166–7; scientific, 10, 78, 83, 84, 198; selective/unselective, 53 reality, 1, 3–7, 170 realization, 213–17; imperfect, 229–31 reductionism, 85, 86 reference, 8, 130, 131, 154, 170, 171, 172, 178 reference fixing, 207 reflectance, 100, 101, 104, 110 relativism, 11 representational fallacy, 5 revenge problems, 38, 39 Russell, Bertrand, 30–4, 113, 204, 212 Russell’s paradox, 30–4 Russell’s theory of descriptions, 204, 212 Ryle, Gilbert, 85
S Schiffer, Stephen, 8, 71, 77, 124, 126, 146–57, 178 scientific method, 117 scientific realism, 10, 78, 83, 84, 198 selective realism, 53 semantic ascent, 1, 9–10, 196, 197 semicausal accounts, 171, 172 sentences, 17
309
sets 117 set theory, 113, 118; paradox, 31, 32 sharp borders, 26–7 sharp cutoffs, 25 simplicity, 30 singling out, 262–6, 269, 270 skepticism, 49, 120, 237–9, 245, 254; “brain-in-a-vat”, 237n.–238n. “smidget” example, 20n., 23, 24 Snowdon, Paul, 12 something-from-nothing inferences, 147, 148, 156, 168, 170, 176, 180 something-from-nothing transformations, 146 sorites paradoxes, 3, 29, 30, 31, 37, 38–40 sortals, 12, 254–71 spatiotemporal boundaries, 262, 269 spatiotemporal coincidence, 261, 267 spatiotemporal/nonspatiotemporal entities, 17, 66, 77 speciation, 45 species, 4, 43–65; category problem, 45; pluralism, 43, 44, 45–7, 52, 54, 65; taxon problem, 45 spin, 199, 219, 229, 231 Spinoza, Baruch, 96 “spraying test”, 200, 205 Stanford, Kyle, 4, 44, 57–9, 65 Sterelny, Kim, 45, 61, 62 star exhaustion, 24 star gaps, 2–3, 25 star mates, see stars stars, 20–2, 24, 27, 28 Stoljar, Daniel, 9 Stoutland, Fred, 93, 97 Stove, David, 82, 232–42, 246, 252, 253 St Petersburg paradox, 37 Strawson, Galen, 117 Strawson, P., 127 substance dualism, 87 substitutional quantifiers 154 superassertibility, 137n. supervaluationism, 25–6 sweeping minimalism, 123–5, 134, 142, 145, 151n., 152, 154–9, 162–4, 167, 171, 173, 174, 177–80 syntax priority thesis, 150n.–151n. systematics, 45, 61, 64
T Tarski, Alfred, 120, 279 Tarskian semantics, 119
310 Index tautology, 174, 234 taxa, 4, 45, 51, 52, 53, 56, 58, 63, 65 theoretical models, 219 theory-ladenness, 4, 57–8 theory of fiction, 75 thick entities, 156 things-in-themselves, 241n., 243, 249, 250 Thomasson, Amie, 72, 75, 76, 170, 173 Thomson, J. J., 117, 202, 206, 214, 222–9 tolerance relation, 18 torrent of molecules model, 221, 222 transformations, 146 transparency, 26 truth, 1–3; 127–34, 272–90; and realism, 1, 7–14, 158; and reality, 1, 7–14; and vagueness, 29–40; as link between meaning and ontology, 13, 140; as property of propositions, 129; conditions, 278–80; conjunctive, 284; disjunctive, 283–4; folk/ordinary notions, 139, 145; Minimalist theory, 127–32; nature of, 1; necessary, 284; negative, 282–3, 290; primary/nonprimary, 281, 290; questions about, 1; transparency, 26; values, 1; Wright’s minimal theory of 134–46, 157, 174n. truth-aptitude, 142–4, 159 truthbearers, 12
truthmakers, 6, 12, 13, 67, 93, 97, 272–90 truthmaking 12–13, 272–7, 280–2 truth-platitudes, 136, 138, 142 T-terms, 211
U unary predicates, 17 uniform solution, 29–31, 35 unselective realism, 53 unsettledness, 17–28 unsettled predicates, 2 U-theories, 200n.
V vagueness, 1, 2, 3, 18, 21, 26–7, 29–40
W Walton, Kendall, 246, 247 wave models, 222 Wiggins, David, 12, 254–65, 267, 269, 270, 271; minimal metaphysical reading of, 257–8, 262 Wilson, R. A., 52n. Wittgenstein, Ludwig, 85, 87, 113, 114, 115, 117, 120 word magic, 5, 11, 69, 70, 73, 235 Wright, Crispin, 8, 80, 81, 117, 124, 126, 134–46, 157, 158, 174n., 178
Y Yablo’s paradox, 33–4, 36–7