JOURNAL OF SEMANTICS
AN INTERNATIONAL JouRNAL FOR THE INTERDISCIPLINARY Sruov OF SE MANTICS OF NATIJRAL
THE
LANGUAGE
MANAGING EDITOR. PI!TER BoscH (IBM Saennfic Ccnrre, Hetdelbcrg and Umvermy of Osrubriick) ASSOCIATE EDITORS MANFIIED K Rmu. (Umvermy ofTeX2� Ausnn) R o B VAN DER SANDT (Umvernry ofN':)mcgen) REVIEW EDITOR TtBOR Ktss (IBM Saennfic Centre, Hetdclbcrg) ASSISTANT EDITOR BIANL\ BUSCH BECK-W o tP (IBM Setennfic Ccnrre, Hctdclbcrg) EDITORIAL B OARD. N AsHER (Umvemry ofTexas, Ausnn) R. BARTSCH (Uruverstry ofAmsrcrdam) M. BtERWI!CH (MPG and Humboldt Umvemry Be rim) B BocURAEV (Apple Computer Inc) M. BoRJuo (Uruvemry ofToulouse) K BROWN (Umvemry of�x) G CHtERCHtA (Uruvemry ofMtlan) 0 DAHL (Uruvemry ofSrockholm) S C GARROD (Uruverstry of Glasgow) B. GEURTS (Umverstry of Osrubriick) M HERWEG (IBM Saenofic Ccnrre, Hetdelberg) L R. HORN (Yale Uruvernry) J JACOB! (Uruvemry ofWuppeml) P N JoHNSON-LAtao (Pnnceron Uruverstry) H KAJ..p (Umver51ry ofSrurrg:m) EDITORIAL ADDRESS
S LEVINSON (MPI Ntjmegen) S UIBNER (Uruvernry ofDusseldorf) StR JoHN LYONS (Umverstry ofCambndge) A. MANASTER-!V.MER (Wayne Scare Uruvemry) J McCAWLEY (Umvemry ofClucago) M MoENS (Umvemry ofEdmburgh) F J PELLtmER (Umvemry of Alberta) M PtNKAL (Umvemry ofSaarbriicken) T SANFORD (Uruvemry ofGlasgow) R. ScHA (Uruverstry of Arnsrerdam) A VON STECHOW (Umverstry ofTiibmgen) M STE EDMAN (Umvemry ofPenruylvama) W W AHLSTER (DFKJ, Saarbriicken) B WEBBER (Umverstry ofPennsylvama) H ZEBVAT (Uruvemry of Amncr
Jounul ofSemano� c/o Dr P Bosch, IBM Germany Saennfic Cenrrc, Vangerowsrr. 18, D-6911 5 Hetdelbcrg, Germany. Phone (-49-6.Z21-) 59--4251/«37 Telefax (-49-6121-) 59-3200 Emul:
[email protected]
0 Oxford Umvcrnry Press pan of thts pubhcanon may be reproduced. stored tn a remcval system, or rransnutted 1n any decrromc. mechamcal, photocopymg, recordmg, or otherwue wnhour e1ther the pnor wnnen pe iTlillSlon of the Pubhshers, or a hcence perrmmng resmcred copytng 1ssued 1n rhe UK by the Copynght Ltcensmg Agency Ltd. 90 Torrenham Coun Road, London W1P 9HE. or m the: USA by the Copynght Clearance Center, lll Rosewood Onve, Danvers, Massachusem 019lJ, USA. All ngbn reserved, no
form or by any means,
journal
of &m..ntra (ISSN
0167 s
1 33) u pubh.sbed quanerly m February, May, August and November by Word 130 per year jountal ofSnMntra u dumbuted by MAIL
Umverstry Preu, Oxford. UK Annual subscnpnon u USS
Amenca, lJ2J Randolph Avenue, Avenel, New Jersey 07001, USA. Second class postage p:ud a1 Avenel, New Jersey, USA and a1
US
addinmul entry poma..
POSTMASTER:
send address correcnons 10 jauntal of SnMntJCJ. c/o MAIL Amenca, lJlJ Randolph Avenur.
Avenel. New jersey o7001, USA.
JOURNAL OF SEMANTICS Volume 13 Number
2
CONTENTS ARNIM VON STECHOW
The Different Readings of Wieder 'Again': A Structural Account
87
FRANCIS RENAUD
The Defirute Article: Code and Context
139
UNGUISTICS
an interdisciplinruy journal of the language sciences
Editor-in-Chief:
Wolfgang Klein, Max-Planck-Institut, Nijmegen,
The Netherlands
LINGUISTICS was founded in 1963 as an international review and reconstituted under a new Editorial Board in 1979. The journal proVIdes an international and interdisciplinary forum for research in the language sciences. LINGUISTICS publishes articles and short notices in the traditional disciplines of linguistics: pragmatics, semantics, syntax, morphology, phonology. In addi tion, it will accept articles and notices from neighboring disciplines that are engaged in the study of natural language: experimental phonetics, psycho linguistics, and neurollnguistics, and first and second language acquisition. LINGUISTICS also publishes occasional special issues and welcomes pro posals for these. LINGUISTICS accepts work from other disciplines, inBofar as it is deemed to be of
interest to lmguists and other students of natural language; for example, cer tain studies in logic, artlflclallntelligence, social interaction, physiology, and neurology would meet this requirement
LINGUISTICS is published in one volume of six issues per year (approx. 1200 pages). As of 1979 the numbering was changed; Volume 34 corresponds to Issues 341-346. Subscription rates for
34 (1996): OM 628.00 /tiS 4699.00 I sFr. 594.00 {postage extra) OM 315.00 (includes postage) OM 111.00/tiS 666.00 /sFr. 106.00
Volume
Institutions I libraries Individuals* Single issue
Prices in US dollars for subscriptions in North America only: $ 430.00 {postage extra) Institutions/libraries $ 207.90 (includes postage) Individuals* $ 72.00 Single issue • Sublcrtptioru at the lndlvlduolntte ore for personal use only ond mull be prepaid and ordered directly from the publWlar Prepayment may be made by check or by credit card MollerCard (Acceu]. EuroCard, VIM ond American Expreu Orden� placed for lnstltutl01u will bo Invoiced ot the lrutitutionol ntle The lndlvlduol ntle I• not ovollable In the FRG, Swl118rland or Auttria
Institutional subscriptions and single or back issues can be ordered from your local bookseller or subscription agent, or directly from MOUTON DE GRUYTER (a division of Walter de Gruyter) at the following addresses:
For North America:
Walter de Gruyter, Inc. 200 Saw Mill River Road Hawthorne, NY 10532 USA
For all other countries: Walter de Gruyter 8: Co. Postfach 30 34 21 D-10726 Berlin Federal Republic of Germany
Scope of this Journal The JOURNAL OF SEMANTICS pubhshes uncles, notes, diSCUSSIOns, and book reVIews m the arc;� of natural language �manncs It
IS
exphacly mterdiSClplmary, m dut It auns at an integranon of
philosophical, psychol<:>g�cal, and lmguiSnc semanncs
as
well
as
semannc work done m arnfiaal
intelligence md anthropology Conmbuaons must be of good quahry (to be JUdged by at least two referees) and should relate to quesnons of comprehensiOn and mterpreution of sentences or te:xts m natural language. The editors welcome not only papers that cross mdmonal disoplme boundanes, bur also more speaahzed conmbuaons, proVIded they are accessible to and mreresang for a WJder readerslup Empmcal relevance and formal correctness are paramount among the cntena of accepunce for pubhaaon.
Information for Authon: Papers for pubharion should be submmed m 3
cop1es
to
the
mamgmg editor. They should be typed on A4 (or similar format� one-s1ded, double-spaced, and
WJth a WJde margm and must be accompamed by m approx. 200 word summary Not� md bJbho
graplucal references must appear ;t the �nd of the rypescnpt. All bibhograplual references m the text by author's sunume and year of pubhanon.
Dugr.lms
must be submmed camera-ready. All
papers submmed are subJect to anonymous refereemg and are considered for pubhanon only on
the assumpnon that the paper has neither as a whole or m part already been publiShed elsewhere nor
has elsewhere been submmed or accepted for pubhcanon Unless speaal arrangements have been
made, copynght rests WJth Oxford Uruvemry Press. Authors receive 20 offpnnts of their publiShed
uncles and 10 offpnnts of the1r pubhshed reVJews, free of charge. Larger numbers an be supphed at cost pnce by advance arrangement
Subscriptions: The Journal ofSemmacs IS pubhshed qwrrerly.
Institutional: UK and Europe £72 oo; USA and Rest of World US$130.00 (Smgle Issues. UK and
Europe £2 1 .00, USA and Rest of World US$38.00) Pmonal:• UK and Europe £15
oo,
USA and Rest of World US$66 .00. (Smgle ISSue: UK and Europe
£ 10.00; USA and Rest of World US$19 00) • Prnonal raus apply only whrn coptes art srnt to a pnvau addrro and payrnrnt ts made by pasonal chequdcredtt card Pnces mclude posuge by surface ma1l or, for subscnbers m the USA and Canada by Airfreight or m
Japan, Ausrraha, New wland and Ind1a by Au Speeded Post. Airmail rates are available on request.
Back fuues The current plus two back volumes are avatlable &om the Oxford Umvemry Press,
Walton Street, Oxford OX2 60P PreVIous volumes can be obumed from Dawsons Back Issues,
Cannon House, Park Farm Road, Folkestone, Kent CTI9 sEE. tel +44 (o)IJOJ 8SOIOI, fax +44
(o)1303 850440. Volumes 1-6 are ava1bble from Swets and Zeahngcr, PO Box 830,2 16o SZ LISSC, The
Netherlands
Payrnmt 1s reqmred WJth all orders and subscnpnons are accepted and entered by the volume Pay ment may
be made by cheque or Eurocheque (made payable to Oxford Umvemry Press), Nanonal
G1robank (account soo 1056), Credit cards (Access, Vw, Amencan Express, Dmers Club), or
UNESCO coupons.
Please send orders and requests for sample cop1es to the Journals Subscnpnons Department, Oxford Umvemry Press, Walton Srreet, Oxford OX2 60P, UK. Telex 837330 OXPRES, tel +44 (o)1865 267907. fax +44 (o)1865 267485
Copyright: It IS a condmon of pubhanon m theJournal that authors as.s1gn copynght to Oxford Umvemry Press. Tlus ensures that requests from third pames to reproduce amcles are handled
efficiently and consiStently and WJII also allow the amcle to be :IS WJdely diS�mmated as poss1ble In
:lSSigntng copynghr, authors may u� their own matenal m other pubhcanons proVIded that the Journal IS acknowledged :IS the ong�nal pbce of pubhanon, and Oxford Umvemry Press IS nonfied m wrinng and m advance.
Advertising: AdvertiSements are welcome and rates WJII be quoted on request. Enqumes should be addressed to Jane Parker, Oxford Journals Advertising,
19 Whitehouse Road, Oxford OX 1 4PA,
UK. Tel!fax: +44 (o)1865 794882. Emarl·
[email protected].
C Oxford Umvrrmy Pro:ss 111')6
The Different Readings of Account
Wieder
'Again': A Structural
ARNIM VON STECHOW
Uniwrrity ofTubinJ.ffl
Abstract
I In Stechow
(I995)
THE S COPE OF THI S STUDY1 I defended an analysis of the different readings of the
German adverb wieder 'again' in terms of scope. The account was supposed to compete with proposals by Dowty (I979) and Fabricius-Hausen (I983), who account for the different readmgs by meaning postulates, though in different ways. Dowty assumes a lexical ambiguity for the adverb, whereas Fabricius Hausen has a posrulate that yields different readings depending on the kind of entity modified by the adverb. I will not repeat my earlier criticism. In Appendix I, I report some further crincism of Dowty's (I979) meaning postulate, which is due to Ede ZimmermaJUL Furthermore, I will comment on
a recent proposal by Fabricius-Hausen, which is different from her earlier proposals but similar in spirit. In this srudy I will defend an analysis in terms of scope by applying it to a wider range of data. In addition, I try to combine this approach with Kratzer's (I994) syntax and semantics of the voice phrase. The starting-point of the analysis given in Stechow
( I-I )
(1995) is the following
conrrast in interpretation found in German: a.
Ali Baba Sesam wieder offnete (restirutive/repetitive) Subj Obj agam opened
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
I wtll defend a purely structural accounr of the dtfferenr readmgs ansmg from the Genn:m adverb wittier 'agam'. We will be concerned With the so-called repennve/resritunve ambtgmry. The claun is that the ambigmry can be resolved ennrely m terms of syntacnc scope The theory assumes a rather abstract syntax.. In parocular, abundant use is made of Kratzer's (1994) votce phrase, which plays a central role for the denvanon of repennve readmgs. One of the leadmg tdeas of the analysis IS that the structural accusanve p<)S!non has Wide scope With respect to the agent relanon expressed by the head of the votce phrase. If Wittler precedes an accusanve object, a repentive readmg 15 obhgatory. If Wittler follows the accusanve object, two readmgs are avail able due to two posstble p
88 The Different Read1ngs of Wieda 'Agam' b. Ah Baba wieder Sesam offnete(only repetitive) Subj
again
Obj
opened
The two German sentences exhibit the word order of dependent clauses, which is commonly regarded as the underlying word order in German, at least as far as the position of the finite verb is concerned: the finite verb occupies a final position (Bierwisch 1¢3). In this article most of the examples will be verb-final sentences. The semantic contrast between the two examples is now this:(1-1a) has both a restitutive ('inner') and a
repetitive ('outer') reading, whereas(1-1b) has only a
repetitive reading.2 The restitutive reading presupposes that Sesame had been it before; it is not even necessary that anyone had opened 1t before. On the other hand, the repetitive reading presupposes that Ali Baba himself had opened the mountain before. The crucial fact in these examples is that the posmon of
wieder relative to the
object may disambiguate the readings: if wieder appears in front of the direct object, only the repetitive reading is available. Stechow (1995) argues that this can be explained if we decompose the verb djfnen 'to open' into 'CAUSE + open'. This explanation can be traced back to Generative Semantics.3 The novelty of the proposal lies in the claim that the data are explamed by a syntactic theory which assumes movement of arguments to Case positions(see, e.g. Chomsky 1989). The classical treatment, which comes from Dowry (1979), assumes the same kind of decomposition, but the decomposition is not located in the syntax but rather in a conceptual semantic language only. In Dowry's approach, the adverb
wieder 'again' would occupy the same position in the syntactic structure, but the adverb is lexically ambiguous, and a meaning postulate has the task of creating the restitutive reading. The ambiguity of (r-ra) is explained by assuming the D-structure (1-2a), whereas the non-ambiguity of(1-1b) follows from the D-structure (1-2b): (1-2) a. SpecAgrO again [VP Ali Baba [scagain[sc Sesame open)) CAUSE) CASE repennon restitution b. again [AgrO-P SpecAJrO [yp [Ali Baba [sc Sesame open) CAUSE]) CASE repetition To facilitate the reading, I have translated the German words or morphemes into their English counterparts. I will adopt this method without further comments in the following text as well. As I said before, the structures assume that German is head finaL Hence, abstract verbs like CAUSE appear at the right side of the recursion. This feature of the analysis is not essential, however. To be sure,(1-2a) is short for two D-structures, each containing only one occurrence
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
open some nme before, but it is not required that Ali Baba himself had opened
Anum von Srechow 89
(1-3)
a.
die Tiir wieder aufging (restitutive/repetitive) the door again open-went 'the door opened again' a'. SpecAgrS again [VP[sc again [sc the door open]] BECOME] CASE repennon restitunon b. wieder die Tur aufging (repennve) again the door open-went b'.again SpecAgrS [VP[sc the door open] BECOME] repetition CASE
A problem for this analysis is that it seems to presuppose that the subject moves to the nominanve posinon [Spec, AgrS] for Case reasons. Syntacncians of German, however, generally assume that the norrunative can be assigned VP internally (den Besten 1 98 5). This assumption is in conflict with Stechow's (1995) explanation, because a VP-internal case assignment, say, via agreement with an invisible expletive in the subject position, is compatible with a restitutive position of wieder (c£ Sternefeld 1 98 5): (1-4)
a.
Expl1 wieder die Tiir' aufging b. SpecAgrS again [VP[sc again [sc the door open]] BECOME] repetition restitution Expl
It follows that this cannot be the correct structure, or Stechow's ( 1 995) explanation is not on the right track.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
of again. If again is generated in the higher position, it has wide scope with respect to CAUSE and we have the repetitive reading. In the lower position, again has narrow scope with respect to CAUSE and we have the restitutive reading. I have marked the two adverbial positions with the labels 'repention' and 'restitution'. At the surface, the subject Ali Baba moves to [Spec, AgrS]. The subject of the small clause, namely Sesame, moves to [Spec, AgrO] in order to get the accusative. Obviously, the two movements are compatible with two different base positions of again. This is the explanation of the ambiguity in the example (a). In the second example, the surface position of again makes it clear that it must be attached to a node higher than VP. Thus only the repetitive reading is possible. The explanation relies crucially on certain syntactic assumpnons: we have to localize the abstract morpheme CAUSE at an appropnate place in the syntax, and we have to assume an Accusative position 'above' the CAUSE-VP. Furthermore, we have to make analogous assumptions with regards to an abstract BECOME morpheme for intransitive verbs of transition, because they show the same pattern in interpretation:
90
The Different Readmgs of W1tda 'Agam'
(r-s) Als Anna wieder das Haus verlieB, war es dunkel (repetitive/restitutive) When Anna agam the house left, was it dark
The restitutive readmg 1s perhaps more naturally expressed by a sentence in which wieder follows the direct object, but (r-s) certamly can have that reading. Examples like this one were not considered in Stechow (1995), and we will have to talk about such cases. To anticipate the solution, we will argue that verbs of this class have a more complicated lexical decomposinon which licenses two accusative positions, namely a repetitive and a restitutive one. Another objection to a scope solution comes from Fabricus-Hansen, who pomts out on several occasions4 that such an enterprise has to face empirical difficulties having to do with verb pairs like tofall/to rise, which may be called counterdirectional antonyms.5 Consider the following pair of sentences:
(r-6) a. das Barometer WIEDER fiel (repetition)
the barometer again fell b. das Barometer wieder FIEL (restitution?)
Clearly, (r -6a) has a repetitive sense: the barometer fell after having fallen before. But what about the restitutive reading? Such a thing does not seem to exist. On the other hand, we have a very clear intuition about what (r-6b) means: the barometer fell after a rising. This reading is called counterdirectional by Fabricius-Hausen. The terminology is well motivated because the sentence does not seem to express the restitution of a state but rather continuation of motion or traflSitlon in 'reversed direction'. In this study I will try to subsume these examples under a 'repetitive/ restitutive' meaning of wieder. Roughly speaking, the account will be that a
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The examples show that Case theory 1s crucial for the explanation. We will therefore have to speak about the precise locations of the CASE positions in German. Another problem lS the status of abstract morphemes hke CAUSE and BECOME. If they play a crucral part in the explanation of the facts, they must be located somewhere in the syntax. The result is some rather abstract syntactic representatlon, regardless of whether we localize the distinctions directly in the syntax as Stechow (1995) did or recur to a logical language between the syntax proper and the interpretation as Dowty (1979) does. A genuine altematlve is provided by Kratzer's Davidsonian approach, which we will discuss in Section 3· There we can express the different readings without CAUSE. It wtll not be possible, however, to get rid of the BECOME-functor, as we will see. The theory sketched here predicts that we will obtain the repetinve reading if wieder 'again' precedes the object. Yet there are certain result verbs which do not fall under this generalization. For example, consider the following sentence:
Arnim von Stechow 91 'restitutive' falling is the reaching of a degree of height which the barometer had had before. Let me say at the end of this introduction that the aim of this article is not
purely descriptive; if the approach is along the right lines, it tells us something
about the German verb phrase and about the morphology/syntax/semantics interface.
2
THE MORPHOLOGY/SYNT AX I NTERF A CE
The general idea for organizing the syntax/morphology interface is due to into D-srructure, or they may be formed in the syntax by head movement or incorporation . Baker (I98 5 b/ I98 8) speaks of modular morphology .6 The idea is that morphology is a component or 'module' of the grammar which checks the
morphological well-formedness conditions, m particular morphological
selection restrictions.
Consider the sentence
Charlotte lachelte (' Carlota smiled'). A D-srructure in
the style ofStowell (I98I) is someclung like the following tree:
(2-1)
The past morpheme
- te
!(NFL) selects a stem of category V.
is a bound morpheme which selects a verbal stem.
Generally, an inflection of category
For the following, let us distinguish between stl«tion) and phrasal stlection (max-selection ).7
morphological selection (min
In order to check phrasal selection of YP by X, we need the configuration
(2-2)
[x· YP X] or [x· XY].
and in order to check morphologtcal selection, we need the configuration
(2-3)
[x YX] or [x XY].
We may call structures of the latter kind words if we like. This terminology does
in toto from the lexicon or that they morphology which is independent of the
not, however, imply that words are taken are the output of a component called
syntax proper. Some words are formed in the syntax, others are not. As far as they are morphological structures, they invariably satisfy the morphological
well-forrnedness conditions, especially morphological selection.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Baker ( 1 988). Words may be built in the morphology and subsequently inserted
92 The Dtfferent Readmgs of Wrtdn- 'Agam'
We may assume that phrasal selection is satisfied at every level of grammatical representatioiL This is a version of Chomsky's (1 98 1 ) projection principle. Baker ( 1 988: 1 40) assumes that morphologt.cal selection 1s satisfied at S-structure, a pnnciple called the Stray Ajfu: Filter, which goes back to Lasnik ( 1 98 I) . Suppose now that the finite morphology -te both min- and max-selects V; i.e. it selects 0> ( - V) and Jill-(- VP). It follows that V must be incorporated at S-structure mto !(NFL). An appropnate theory of head movement will ensure that the S-structure of Charlotte Iachelte must be something like the following tree: (2-4)
Tlus structure sansfies morphological and syntactic selection. Note that morphological and phrasal selection are independent of each other, either may occur without the other. The following enmes illustrate the point: (2-5)
bar-level selected rmn max 3.pers.sing. accusative lachelCharlotte
·
+ +
category selected
+
v
+
N N none
The interesting entry is the INFL-morpheme 3.person singular, which has conflicting selecnon reqmrements: both a V and a VP are selected. The conflict is resolved by checking the phrasal selection (max-selection) at 0-structure and the morphological selection (min-selection) at S-structure. Recently, a lot of so-called functional nodes have made careers for themselves. For instance, the category !NFL has been split into the nodes T (tense) and Agr (agreement). Many linguists do not stop here: they split Agr into AgrS (ageement subject) and AgrO (agreement object). Furthermore, a Neg-node has been introduced (c£ Pollock 1 989 and Chomsky 1 98 9) . In what follows, I will argue that the 0-structure for German must be something like the tree in (2-6). The structure is more or less identical to that proposed in Chomsky ( 198 9) . The difference is the Voice projection, which 1s taken from Kratzer ( 1994) . Kratzer assumes that the Voice node is somehow tied to inflection, and that there might be active or passive voice. Together with Stechow & Sternefeld
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
-te
Amim von Scechow 93 (2-6)
nommative
Ag .
� r----___ �� � VOice
Agrt)
Subject
Object
v
(r988), I will assume that there is no morphological active/passive distinction m German. The voice head is e1ther an empty node, or 1t lS, as we will see in sec tion I4, realized as a 'passive' aUXIliary such as werden , kriegen!bekommen , or sein + zu. Semantically, the morpheme may be filled by certain thematic relations hke agent (of the acnon under consideranon) or holder (of the state under col1Sideration), or it may be empty. Following Kratzer (I 994), I will assume that the subject is generated in (Spec, Voice). In other words, subjects are not gener ated in VP. This point will become clearer as soon as the theory 1s applied to examples. It is important to distinguish the abstract voice morpheme from its semantic interpretation, which is not unique but is determined by the content of the verb. For il1Stance, an 'unaccusative' verb like kommt 'comes' has an ACTIVE voice morpheme, but there is no thematic relation located m the Voice node. (Spec, AgrO) is the strucrural accusanve position. Kratzer (r994) assumes that the accusative position is the highest (Spec, V) under VoiceP. This might be a viable alternative as well, but I was not able to check it in all details. The Case Theory presented in Section 4 will closely follow Stemefeld (r99s).
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Ag
accusative
94 The Dtfferent Readings of Wied" 'Again'
3 T OW ARDS SEMA N T ICS A ND LF In this section we will develop our theory in detail. Recall that we want to explain the following contrast in readings: (3 -r) a. Ali Baba Sesam wieder offflete (restitutive/repetitive) Subj Obj again opened b. Ali Baba wieder Sesam offflete (only repetitive) Subj again Obj opened
a. SpecAgrO again [ypAli Baba again[sc Sesame open] CAUSE] b. again [AgO-P SpecAgrO [VP Ali Baba [sc Sesame open] CAUSE]] Remember that (3-2a) is short for two D-structures, each containing only one occurrence of again. If again is generated in the higher posinon, it has wide scope with respect to CAUSE and we have the repetitive reading. In the lower position, again has narrow scope with respect to CAUSE and we have the restitutive reading. At the surface, the subject Ali Baba moves to [Spec, AgrS]. The subject of the small clause, i.e. Sesame, moves to [Spec, AgrO] in order to get the accusative. Obviously, the two movements are compatible with two different base positions of again. This lS the explanation of the ambiguity m the example (a). In the second example, the surface posinon of again makes it clear that it must be attached to a node higher than VP. Thus only the repetitive reading is possible. What IS still missing is the semantics for the morphemes involved. That will be provided shonly. The explananon assumes a small clause in order to account for the ambiguity. This is morphologically plausible for causative verbs derived from an adjectival base like Ojfnen 'to open', because the adjectival base could be considered as the head of the small clause, whereas the verb is formed by incorporation in the syntax. The same kind of ambiguity, however, also arises with verbs where we have no morphological indication of decomposition:
(3-2)
(3-3) a. Randi den Bockhirsch wieder fing (restitunve/repetitive) Randi Bockhirsch again caught b. Otto wieder den Bockhirsch fing (only repetitive) Otto again Bockhirsch caught
The first sentence can mean two things: r. Bockhirsch (-our cat) had been a prisoner before, and Randi caused him to be a prisoner again. This is the
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
As mentioned in Section 1, the ambiguity of (3-ra) can be explained by assuming the D-structure (3-2a), whereas the non-ambiguity of (3-1 b) follows from the D-structure (3 -2b):
Arnirn von Stechow
9S
restrictive sense. 2. Randi had caught Bockhirsch before and she caught him again. This is the repentive reading. The second sentence has only the repetinve mearung. These examples require decomposition of the verb fongen 'to catch'. The D-srructure accounting for the ambiguity of {3-3a) is something like: (3-4) SpecAgrO [again [VP Randi [again [sc Bockhirsch PRISONER)] BECOME] CAUSE]
(3-s) prisoner(Bockhirsch(s) This means 'Bockhirsch is in the state of being a prisoner'. In other words, this 1S a DaVIdsonian approach which assumes 'eventualities', i.e. actions, events, and states. Following Kratzer {I994), I reserve the variable s for states proper. The variable e ranges over eventualines of any kind, i.e. states, activities, achieve ments, accomplishments, and perhaps others. The variables may have indices, of course. A note on the semantic language is in order. I will not elaborate on every formal detail. The language will be an extensional typed language in the style of Gallin (I 97s). Since no intensional constructions are considered, I don't need the type of worlds. Rather, the type s refers to eventualities (following Kratzer I994). e is the type of individuals, and t 1S the type of truth-values commonly assumed m the literature. Thus the symbol prisoner is of type (e, (s, t)). The meaning of 'Bockhirsch is again m the state s of being a prisoner' 1s represented as: {3-6) again(A.s(prisoner(BockhirschXs)])(s) - again{prisoner(Bockhirsch)){s) As the formulae show, again is of type ((s, t), (s, t)). Its meaning is given in (3-7), a definition inspired by Egg {1993):
(3-7) Let P be a property of eventualities and let e be an eventuality. ' �againJI(P)(e) is defined only if3 e [I!MAxll{P)(e')- I & e' < e] . Where defined, llagainii{P)(e)- I iffP(e)- I.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
This is a highly abstract analysis of the verb in the style of Generative Semantics and Dowty {I979). One would certainly like to have a more concrete treatment requiring less decomposition. After becoming acquainted with Kratzer's proposals, I thought that I could adopt them in order to obtam a syntactically simpler treatment To anticipate the result, I was able to replace the CAUSE functor with the thematic role agent located in the VoiceP, but I could not get rid of the BECOME-functor. Let us start with the semantic analysis of the examples in (3-3). Consider the following formula:
<)6 The Dlfferent Readmgs of W1rdn- 'Again'
The definition presupposes that P is a property of eventualities. < is the relation of temporal precedence. It is true of any two eventualities if the first is temporally located entirely before the second. MAX(P)(e') means that e' is a maximal P-event. More formally, (3-8) MAX is a symbol of type ((s, t), (s, t)). �MAXJKP)(e)- I ifiP(e) and there is no e' such that e is a proper part of e' and P(e')- 1. Thus, (3-6) can be read as 's is a state such that Bockhirsch is a prisoner at s and there has been a maximal state of the same kind before'. Next, consider the restitunve reading of the sentence (3-9):
Using an 'event version' of Dowty's BECOME-functor and Ignoring tense, we can formahze this as: (3-10) BECOME(again(prisoner(Bockhirsch))Xe) BECOME is a symbol of type ((s, t), (s, t)) and has the folloWing meaning: (3-II) (BECO�I(PXe)- I iff e IS the smallest event such that P is not true of the pre-state of e but P is true of the target state of e. The pre-state of e is the state that holds immediately before the event e occurs. The target state is the state reached at the end of the event. Both notions are primitives of the theory. Similar notions are used by many authors working with events. For instance, the notion pre-state is used by Fabricus-Hansen (I994a, b). The term target state is used by Klein (1994) and Kratzer (1994). A more careful statement of the meaning rule would confine the BECOME predicate to properties of events which have target states. In an event approach. we do not need the predicate CAUSE any more.8 Its role can be taken over partially by the thematic relanon agent(xXe), wruch means that x is the agent of action e. This predicate IS more or less Dowty's (1979) predicate DO. Ignoring tense, the two readings of example (3-3a) can now be expressed by the following two formulae:
(3-I r )
a.
3e[again(Ae[agent(Randi)(e) & BECOME(prisoner(Bockhirsch)Xe)])(PAST(e))] b. 3e(agent(Randi)(e) & BECOME[again(prisoner(Bockhinch))](PAST(e))]
This semantics assumes a defirute intepretation for the past tense in the spirit of Partee (1973):
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(3-9) Bockhirsch wieder ein Gefangener wird Bockhirsch agam a prisoner becomes
Arnim von Stechow
97
(3-12) PRES and PAST are symbols of type (s, s). a. �RES!J(e) is defined only if e temporally overlaps with the speech time te Where defined, �RESI!(e)- e. b. IJPAST1(e) is defined only if e is temporally located before the speech time tc)· Where defined, IJPASTJI(e)- e.
(3-13)
BECOME
den Bockhlnch PRlSONER Bockhlrscb prison..-
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Thus tenses impose restrictions for the temporal locanon of the event argument of the verb. Gtven these meaning rules, we can calculate the followmg truth condtnons for (3-IIa) and (3-ub), respectively. 0C3-IIa)iK is defined only if g(e) is temporally located before tc, where g is any variable assignment. If this is fulfilled, then g(e) is an event with Randi as agent, and Bocldursch is not a pris oner at the beginning of e, but he is a prisoner at the result state of e, and there has been such an event before. This is the repetitive meaning. The restitutive meaning (3-u) can be restated in English as: �3-rrb)IK is defined only if g(e) is temporally located before tc, where g is any variable assignment. If this is fulfilled, then g(e) ts an event with Randi as agent, and Bocklmsch lS not a prisoner at the beginning of e but he ts a prisoner at the result state of e, and Bockhirsch has been in the state of being a prisoner before. If we ask ourselves where the sernannc information which builds up the two
98 The Different Read1ngs of Witdrr 'Agam'
(3-14) a. Rand1 BockhirSch[t�;uBJ[wieder[[toBJPRISONERJBEC]JAGENTJ resrirunve SpecAgrS SpecAgrO agam b. Rand1 BockhirSch [wieder[[tsUBJ[toBlRISONER)BEC]AGENTJl repenrive
agam
If, on the other hand, wieder appears in front of the direct object at the surface, it must be generated outsi4e VoiceP, for instance, as an adjunct of AgrO-P. Thus the 0-structure of the unambiguous sentence (3-3b) IS: {3- IS) [AgrO-P wieder [AgrO-P AgrO [volceP Randi[yp Bockhirsch fang] ACTIVE]]] again
catch
This structure can have only the repetitive reading, as an Inspection of the tree (3-13) immediately shows: the information which again has to modify is (3-16a), and the result of the modification is (3-16b), which is the subformula of (3-1 I a) wruch determines repetitivity: {3-16) a. agent(RandiXe) & BECOME (prisoner(Bockhirsch))(e) b. again[l-e(agent(Randi)(e) & BECOME(prisoner(Bockhirsch))(e))](e)
If this explanation is on the right track, we have evidence that the direct object must move to a structural accusative position like AgrO; in other words, assign ment of structural accusative is not possible in situ . Furthermore, in German this movement must have occurred at the surface and not at LF.9 Suppose per impossibik that accusative marking of the object's 0-position. i.e. (Spec, X],
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
meanings is located in the LF, the answer is pretty obvious for the most part. The thematic role agent is located in the head of the VmceP. The repetitive positions of the adverb again must be higher than VmceP, whereas the restrictive position must be below Vmce P. The abstract predicates BECOME and prisoner belong to the VP, and BECOME must have wide scope with respect to prisoner. Thus a transparent LF for the VoiceP might have the following shape depicted in 3-13. A remark on the notation is in order. At the nodes of the tree, I have indicated which semantic operations are needed to combine the meanings. I think the notation is self-explanatory. The cruaal feature of the structure is that there are two modifier positions, indicated by the labels REPETITIVE and RESTITUTIVE MODIFIER. which trigger two dif ferent interpretations. If the subject and the object move to their Case positions, the D-position of wieder 'again' is no longer uniquely Identifiable from the surface, because the Case positions are outside of the VoiceP; hence the ambiguity. The following list summarizes the two structures, omitting tense:
Arrum von Stechow
99
were possible. Then sentence (3-3b) could have the following S-structure, where head movement (and movement of the subject) is diSregarded: (3-17) R andi [y01ceP tSUBJ [yp[xp wieder [xp Bockhlrsch prisoner]] BEC] Acn NOM
restitutive again
ACC
1.
.....
_______
catch
(3-19) Randi caught Bockhirsch. Then he escaped. Tristan caught Bockhirsch agam.
The last sentence of this short story reports the repetition of the acnon of catching Bockhirsch: the action had occurred before, though with a different subject. If this is one particular reading of the sentence, then we can represent it by giving again intermediate scope. But we need not analyse the sentence that way. Taking the sentence in its restitutive sense is compatible with the scenario as well. So it is hard to tell whether the intermediate scope reading is real. Thus the sernannc motivation for this additional scope possibility is weak, and morphologically it would certainly be more attractive to have only one verbal stem. After becoming acquainted with Kratzer's (1994) proposal, I thought I could attain a more concrete syntax by defining result verbs in her style. She would analyse the verb catch as a relation which holds of an individual x and an event e if e has a target state that has the property 'x is a prisoner'. This is a meaning rule which does not use the BECOME-functor. I did not adopt this analysis, because I haven't been able to express the restitutive reading with it. I will comment on the difficulties in Appendix 1. I should add some remarks on the location of the accusative positioiL In a number of publicatioru, Diesing has argued that (referential) definitive terrru are scrambled out of the VP in GerrnaiL10 To be sure, Diesing has in mind the
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
This configuration has the restitutive reading. So it must be excluded as a possible surface structure for (3-3b). This follows from the assumption that the direct object has to move to (Spec, Agr-0] for the purpose of Case checking. A certain drawback of the analysis might be its abstractness. We have decomposed the finite verb form fing "caught' into Voice + BECOME + PRISONER The voice head is well motivated for the reasoru gtven in Kratzer (1994). But there iS no morphological evidence for a decomposition of the stem fang- 'catch' into BECOME + PRISONER The monvation is purely semantical-u is a relict of Dowty's (1979) analysis and iS JUStified because it gives the correct readings for the examples discusssed. It might, however, yield too many possible readmgs because it permits again to occur in a position between agent and BECOME, a possibility that doesn't seem to be realiZed. Therefore, let us see whether we can find examples which require such a position for the adverb. A possible scenario could be this:
100 The Different Readings of Wrdtr 'Agam'
classical VP, which we identify with Kratzer's voice phrase. If Diesing's claim about German were true, the examples(ra) and(rb) would be insufficient for determining the accusative position. Diesing is certainly right that referential defirute terms (without contrastive stress) are scrambled out of a negated VP, but I am not convinced that this holds for unnegated VPs as well. Be that as it may, it is safer to consider examples with indefinite terms or quantifiers as direct objects. (3-20)
a.
According to Diesing, existentially interpreted indefinites are in the VP. The data behave as expected. (3-21)
a.
Frank jedes Fahrrad wieder reparierte(repetitive/restitutive) b. Frank wieder jedes Fahrrad repanerte(repetitive/restitutive?)
In (3-21 b) the repetitive reading is certainly the prevailing one, but I am not sure that the restitutive reading is totally impossible. If this is the case, then I have no explanation for thts fact It is interesting to notice that the position of wieder supports Thesing's claim that pronouns are obligatorily scrambled out of the VP, because wieder in front of an accusative pronoun is quite unacceptable.: (3-22) a. ??Chrigel wieder ihn gerade gebogen hat Chrigel again hm straight bent has b. Chrigel ihn wieder gerade gebogen hat Chrigel him again straight bent has 'Chngel straightened It(- the power earner) again' We can derive this contrast if we assume that wieder occupies the accusative position in example(3-22a). For some reason, ihn has to be scrambled out of the 'classtcal' VP.11 Therefore the sentence is ungrammatical It IS not entirely clear what the 'classical' VP which Diesing has in mind is exactly in our approach. Assume that it is the VoiceP. In (3-22b) the pronoun has been scrambled out of the VoiceP. Therefore it is no longer visible whether wieder attaches to VoiceP or to VP, and the sentence should be ambiguous between a repennve and a restitutive readtng. And so it is. I will finish this section with a remark on the localization of tense at LF. The relevant projection of the tree has the following form:
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Frank ein Fahrrad wieder reparierte(repetitive/restitutive) Frank a btcycle again required b. Frank wieder ein Fahrrad repanerte(repetitive) Frank again a bicycle repaired
Arrum von Scechow ror
�
(3-23)
PA PAST(eJ
Cn VoiceP
T p
ast
(3-24) Ae(again[Ae(agent(Randi)(e) & -
BECOME(prisoner(Bockhirsch))(e))](e))(PAST(eJ) again[A.e(agent(RandiXe) & BECOME(prisoner(Bockhirsch))(e))](PAST(eJ)
If we existentially close the formula by some default procedure, we obtain the repetitive interpretation represented by the formula (3-1 I a).
4 VOICE P HRA SE S AND CASE THEORY In tlus secnon I will first present Sternefeld's (1995) theory of votee and case. Then I accommodate it into the present framework. The reason for not adopting Sternfeld's theory as it stands 1s that he generates the subject VP internally, where VP is understood in the classical sense. Only at LF is the sub ject raised to [Spec, Voice]. This is not compatible with Kratzer's approach, and I will not follow Sternefeld in this respect. There is another important difference beween Sternefeld's and Kratzer's theories of voice phrases. For Kratzer, the voice node is always a functional category belonging to the inflection of a verb. For Sternefeld, an auxiliary may be the head of a voice phrase; this enables him to give a principled account of the different 'case absorption' propernes of the different passive constructions m German. Consider the following three verbal passive constructions: (4-1) Accusative passive (Sternefeld 1995: (7<)a.lb)) a. Ich..om schenke dem Fritzdat einen Cognac I give ARTc�arFritz ART.cc Cognac b. Ein Cognac wurde proacc dem Fritzdat (von mir) geschenkt ARTnom Cognac was ARTdat Fritz (by me) given
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Like Stowell (1993) I distinguish between morphological and semantic tense. The former is located in the head of TP and has no meaning. The place of the latter is the [Spec, TP]. We can represent it as PAST1 (or P RESJ at LF, and we translate it into the expressiOn PAST(ei) (or (PRES(eJ), which is of type i. To this we then apply the event-abstract formed of the VoiceP by A.-abstraction over the event variable. Assume, e.g., that the node T' of the tree is interpreted by the formula (3-16b). The value of the entire ttee would then be:
102 The Dtfferent Readmgs of Wied" 'Again' The idea of the analysis 1S that the passive auxiliary werden licenses an accusanve
pro in [Spec, Vmce), which is the imphcit subject of the collStruction. Thus there is no 'accusative absorption' in the GB style. The accusative is rather
asstgned to the non-overt subject.
(4-2)
Dative passive {Sternefeld 1995: {79c)) Der Fntznom bekommt prodat den Cognacacc (von nnr) geschenkt
ARTnom Fritz gets prodat ART.cc The auxthary
bekommen
Cognac
(by me)
licenses a dative pro in [Spec, Vmce).
given
As
Sternefeld
correctly points out, we would need three different lexical entries for the
geschenkt
,
.
(4-3) Zu-passtve a. Der Fritznom
hat den Motorace zu repaneren ARTnom Fntz has ARTace motor to repair b. Der Motornom ist pro.cc {von Fritz) zu reparieren {by Fntz) to repair ARTnom motor is
According to Sternefeld, the voice auxiliary sein licenses an accusative pro as Its specifier. Again, 1f the GB account of the passive were correct, we would need two entnes for infinitives, one that doesn't absorb accusanve and another that does. Before I go into the details of Sternefeld's theory, let me indicate the syntax he assumes. Thts is the voice phrase for the dative passive example {4-4-) [volceP e[VP
(4-2):
prodat den Cognac.cc der Frittaom geschenkt) bekommt) subject object ind. object vmce
Stemefeld assumes that the mdirect object
is
the most deeply embedded
argument of the verb projection. Furthermore, structural Case (nominative, accusative, dative) can be assigned in an arbmary way to the grammatical functiollS, but in the order mennoned and only once: {4-S) Case 'a.
Assignment {Sternefeld 1995: {75))
Within the projecnon of a verb, nominatives can be assigned by default, accusanves can be assigned if nominative has been assigned,
and dative can be assigned if accusative has. b. Assignment of structural Case is possible only once; i.e. if two Cases in the domain of a verb are the same, one must be either a lexical Case or an agreement Case.'
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
if a passive theory in the GB style were correct an active participle, a passive participle, which absorbs the accusative case and a participle which absorbs the dative case In Sternefeld's theory, there is only one participle but there are different votce heads. The same holds for the so-called zu-passive in German: participle
Arnim von Stechow Now, pro and
Case
103
have to be licensed. Let us consider the licensing of
(referential) pro first. According to Sternefeld, the only licensers in German are the passive auxiliaries, �Wtden , bekommtn I kriegen , and
sein . As said
before, the
licens1ng position is their specifier position, i.e. [Spec, Voice).
(4-6)
Pro Licensing (Sternefeld
1995:(74))
'The head of a passive voice phrase must license (via spec-head agreement) a pro that bears a subject theta role.' 'Theta role' is not used in the Davidsonian sense by Sternefeld. The term simply means that the phrase in question is a semantic argument of a verb. This principle entails that pro moves to [Spec, Vmce) at the level where it is licensed, for
Case
licensing, the nominative is hcensed by a tensed INFL, as
commonly assumed in the literature.
(4-7) Nominative Licensing (Sternefeld 1995:(76))
'Tensed INFL can license nominative Case m 1ts specifier position.'
is linked to the direct and the indirect object GF.
The exact narore of the licensing of the accusative and the dative is left open by Sternefeld, but the licensing
Putting all this together, this leads to the following further principles:
(4-8) Licensing of accusative and dative Case (Sternefeld 1995:(77)) 'a. (i) Accusative Case is licensed on a direct object theta role;
(ii)
Dative Case is licensed on the indirect object theta role;
where the exact nature of the licensing head is left open. Thus, licensing can be executed either by default or by V within the projection of a main verb, or by functional projections like AgrObj or AgrlndObj. b. wadtnp and seinz hcense accusative Case marking on the subject theta c.
role.
kritgenp and bekommenp license dative Case marking on the subject theta role.'
The subscnpt P means that the auxiliary selects a participle. The subscript
Z
means that the auxiliary selects an infinitive prefixed by zu . We have been
is our AgrS, of course. It follows that the S-strucrore of
assuming throughout that accusative Sternefeld's tensed INFL
(4-4) is(4-9): (4-9)
case
is licensed in [Spec, AgrO], and
[AgrS-P der Fritznom . . . [AgrO-P den Cognacacc [vokeP produ [VP tdat t.cc �m geschenkt] bekommt]] . . .]
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
say S-structure.
. As
104 The Different R.e2dmgs of Wid" 'Agam' Note that any other Case markmg leads to ungrammaticaliry: the subject has to
be pro,ut' because
bekommen
licenses only a danve subject Furthermore, it
cannot be an open subject in VIew of pro Licensing which requires a pro as the subject of passive auxiliaries. Assigning accusative to the indirect object and dative to the direct object would violate the licensing principles for accusative and dative Case licensing. Thus the Case disrnbutton is described exactly correctly and in a very elegant and transparent way. In particular, no recourse to Case absorption is necessary, a significant progress, it seems to me. The
(c£ German man) if the subject is not introduced by a von
semantics is reasonably clear as well. The pro subject of passive constructions is a generic pronoun
phrase ('by' phrase). In the latter case, pro 1s bound by the object of von . There are paper.
As it stands,
the analysts isn't complete yet for it doesn't allow us to derive
impersonal passives and mfinitivals with PRO as the subject Concermng impersonal passives, Sternefeld ( 1 995) argues convincingly that they are logically independent from the pnnciples governing 'case absorption'. He
assumes that nominative marking IS optional and that we might have a pro without case in German. This pro IS licensed by
werdenp as well. Thus the voice
phrase for sentence (4-1 oa) is (4- 1 0b), and the interpretation of the whole is
(4- 1 0c): (4- 1 0) a. Hier wird nicht gelacht!
[volceP pro [VP gelacht] wird]
Here is not laughed b. Hier
c. -.3e[here(e)
& agent(proXe) & laughing(e)]
As for PRO, it plays the same role as the norninanve as far as Case assignment is concerned. In other words, we can assume PRO for any GF we like and then proceed as if nominative had been assigned. But the licensing conditions are different, of course. PRO is licensed as the subject of an infinitival and perhaps of other non-verbal projections.'2 Let us accommodate Sternefeld's theory to our assumptions now. The first
modification is that passive auxiliaries express the agent relation. Furthermore,
the pro subject is base generated in (Spec, Voice]. It follows that the domam for structural Case assignment can no longer be the narrow VP; the voice projec tion has to be included (together with the different projections due to decom position). Furthermore, we assume that AgrO selects Voice. I think the revisions are clear enough. I nstead of wnttng them down formally, I illustrate them with the analysis of the dative passive example. The voice phrase IS this:
(4- 1 1)
[volceP
prodat bekommr]
[VP [VP den Cognacacc der Fri tznom geschenkt] BECOME]
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
several posstble ways to develop this further, but that is not my concern in dus
Arnim von Stechow
105
Recall that the voice head bekommt translates mto agent. Making all the parameters explicit, this structure therefore receives the following inter pretation: (4-12) agent(proXe) & BECOME(As[given(Fritz)(the cognac)(s)] )(e) 1S a state in which the cognac 1S owned by Fritz. As before, Fritz is the lowest argument of the verbal decom position. It is the indirect object of the voice projecnon. Note that the licensing principles do not forbid that an indirect object has a nominative, because the nominative is not linked to any speaal grammatical function. Fritz could have the danve Case as well, but then no dative would be left for the pro subject. It doesn't come as a surprise that the interplay of wieder with the direct object is exactly as before:
given(Fritz)(the cognac)(s) means that s
To illustrate the two senses, Fritz had a Cognac which he gave to Mary. Mary forgot who the donator was and gave it to Fritz at another occasion. Thus Fritz got the Cognac back. This is the restitutive sense. The Cognac was his bmhday present for Mary. At Chrutrnas, Mary gave Fntz the same Cognac as a present. This is the repetitive meaning. The explanation of the ambigmty is as before. The direct object moves to the accusative position m order to Lcense the accusative Case. If wieder precedes the direct object, the adverb must be higher than VoiceP, and we obtain a repetition. s
WIEDER A N D THE U N E R G A T I V E / U N A C C U S A T I V E D I ST I N CT I O N
Following Perlmutter ( 1 978), 1t is widely assumed m the literature that the subjects of unaccusative verbs are D-obects.u A well-known property of these verbs is that they don't passivize. In the present context, it is interesting to observe that unaccusatives exhibit the repetitive/resultative ambiguity with respect to wieder, whereas this ambiguity is not observed with the so-called unergative verbs. In other words, we obtain two readings when wieder surfaces after the subject of an unaccusative verb, but only when it occurs in front of the subject of an unaccusative verb. Let us see how our theory explains these facts. My first remark concerns the inability of unaccusatives to undergo passiviza tion. Under the present account this follows from the decomposition of these
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(4- 1 3) a. der Fritz den Cognac wieder geschenkt knegt (resntutive/ repetitive) the Frirz.,om a Cognac.cc again giVen gets b. der Fritz wieder den Cognac geschenkt kriegt (repetitive)
I o6 The Different Readings of Wi�n- 'Again'
verbs; they are all verbs without an agentive component, and passives, at least German passives, have an agent-component (or at least a holder-component) located in Voice. If we did passiVIZe unaccusative verbs, they would change their meaning drastically, but passiVIZation should not be a meaning changing operation. To be concrete, consider an example. (s- I )
a.
Der Brieftrager kommt the postman comes b. 3e[BECOME(here(the postman))(PRES(e))]
(5-2)
[AgrS-P the postmaDj PRES [VP [XP y here] BECOME]]
Suppose now we had an impersonal passive: (5-3) •(weil) gekommen wird because come IS According to our prevtous assumptions, the voice phrase and its Interpretation are as follows: (5-4)
a. (volceP pro (VP [XP e here] BECOME] werd-] b. agent(pro)(e) & BECOME(here(?))(e)
It is not at all clear what the empty subject e of HERE (not an event variable!) could be or how it is mterpreted. The trouble is indicated by a question mark in the formula. Note that, contra Sternefeld (1995), no raising is possible in our analysis because the pro subject is a semantic argument of agent. Ratsing would violate the Theta Cnterion. The formalization assumes the following mearung rule for here: (s-s) IJhereJ (x)(s) - I iff s is the state of x's being at the place of the utterance.14 The main role of the state parameter see ms to be temporally to determine the predicate of bemg here. We wtll now try to explain the second observation, which was that many (all ?) unaccusatives exhibit the repetitive/restitutive ambiguity while unerga tives do not. Consider an unaccusative verb first. (s-6) a. der Brieftrager wieder kam (repetitive/restitutive) the postman again came b. wieder der Brieftrager kam (only repetitive) again the postman carne
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(b) is my logical analysis of (a). There is no agent in clus representation. There fore we need either no voice morpheme in the lexical representation or a semantically empty one. If we choose the former option, it follows that the $-structure of (s-1a) is (5-2):
Arrum von Stechow 107
The posonan had come before, and he came again. This 1s the repetitive reading. The posrman had been here before, then he went away, and finally he came back. That's the restitutive reading. The first sentence has both meanings, the second only the first. Next, we have a look at an unergative verb: (s-7) a. Irene wieder geigt (repetitive) Irene agam played the violin b. wider Irene geigt (repetitive) The lack of the restitutive sense is due to the fact that the decomposition of
violin-playing? In (s-7a) the subject has moved to the (Spec, AgrS] for Case reasons, and thus there are two 0-sttuctures with two Interpretations: (s-8 ) a.
[voiceP Irene [VP wieder geig-] agent] [voiceP Irene [VP geig-] agent]
agent(Irene)(e) & again(violin-playing)(e)
again (Ae.agen t(Irene) (e) & violin-playing( e))(e)
b. wieder
(s-Sa) expresses mere repetition of an acnon. In secnon 3 I left it open whether this read1ng exists. Here are some more examples of unaccusative verbs, all of which exhibit the repennve/restitutive ambiguity 1f wieder follows the subject (s-9) a. das Flugzeug wieder !andere (repetitive/restitutive)15 the aucraft again landed b. wieder das Flugzeug !andere (only repentive) again the aircraft landed The repetitive mearung is that the aircraft repeated a landmg. The restitutive meaning is that 1t went down to earth again, after having been in the air (trus rmght be its first land1ng). Sentence (s-9a) has both meanings, sentence (s-9b) has only the repetitive one.
(s- 1 o)
das Barometer wieder fiel (repetitive/resntunve) the barometer again fell b. wieder das Barometer fiel (only repetitive) again the barometer fell a.
The barometer had been down, went up, and then fell agaiiL It is not required that it had been falling before, but 1t must have risen in order to have become high. This is the restitituve reading. The barometer had fallen before and went
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
geigen 'to play the violin' doesn't contam a BECOME-fUnctor whose scope interacts with that of wieder. But what about the decomposinon into agent +
108 The Different Readmgs of W1edn- 'Ag.un'
down further. Thts 15 the repetitive reading. For sentence (s- wa) both meanings are avatlable. Sentence (s-wb) has only the second one. The analysis is straightforward if we decompose Ianden 'to land' as BECOME + on tkground andfallen 'to fall' as BECOME + down . The repetitive meaning of (s-Ioa) is then represented by the formula (s- 1 1 a), whereas (s-1 1 b) represents the resntutive readtng of (s-9a):
(S-I 1 )
a. 3e(again(BECOME(down(the barometer)))(e)] b. 3e(BECOME (again(on-the-ground(the aircraft)))(e)]
I
(5-1 2) a. Unaccusative verbs do not project Voice (they are voiceless). b. Unaccusative verbs have the semantic structure BECOME + stative. It could be that the second hypothests is too strong; there might be purely stative unaccusatives. The important generalization is the first one, for it entails that unaccusatives never 'govern' the accusative Case, since the latter is asstgned m the voice phrase. One might think that thts result is empirically empty because intransitive verbs don't have an object, and therefore there is no GF that could bear the accusanve Case. This, however, is not so. The following contrast, which goes back to an observation made in Carrier & Randall ( 1992), shows that accusative marking has nothing to do with transitivity: (s-I 3)
a.
Olga betete ihren Sohn gesund Olga prayed her son healthy b. *Die Preise fielen den Aktionar arm The prices fell the stockholder poor
Note first that there 1s no semantic reason for the deviance of (s-1 3b), for while the first sentence may be paraphrased as 'Olga's praying had the result that her son became healthy', the meaning of the second sentence can be rendered as 'The falling of the prices had the result that the stockholder became poo r'. In our theory, the contrast is an outcome of Case theory. Betm 'to pray' is an agentive verb and projects a VoiceP. We have to make sure that the subject of the embedded small clause counts as a direct object of the VoiceP; then it can bear accusative Case, which is checked in [Spec, AgrO]. Fallen 'to fall' is an
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
think these representations are almost satisfactory, but the analysis is too simple after all. The meaning of 'to fall' should bring out the idea of continuous change of position and should not be analysed in absolute terms: at each moment of a falling, the falling subject is lower that it was at the preceding moment I will take up this aspect of the verb's meaning in Section 9· To resume the discuss1on of this section, the behaviour of wieder wtth respect to the subject of unaccusatives can be explained if they have the following properties:
Amim
von Stechow
109
unaccusative and doesn't project VoiceP. Hence we cannot have an accusative object, and ( s - 1 3 b) is ruled out. In a way, our theory encodes 'Burzio's generalization', which says that a verb governs the accusative if it has an external theta role (Burzio 1 99 1 ). AgroO selects Voice. Unergative verbs have voice, but unaccusative verbs have none. Therefore unerganve verbs can 'govern' the accusative, whereas unaccusative verbs can't. The reader may check for himself that the subject of a result small clause behaves exactly hke other dtrect objects wtth respect to
wiedrr.
This is as tt
should be because our decomposi tion of result verbs and unaccusanve verbs is modelled after small clauses-they are visible decompositions.
So
TWO C L A S S E S O F RE S U L T VERBS
far, our theory predicts that we will obtain the repetinve reading i f wieder
'agam' precedes the object. In this section I will mvesngate a class "o f result verbs which do not fall under this generalization. For example, consider the following sentence:
(6- I )
Als Anna wieder das Haus verlieB, war es dunkel (repetitive/restitunve) When Anna again the house left, it was dark
wieder follows the dtrect object, but(6- I ) certainly can have that reading.
The restitunve reading is perhaps more naturally expressed by a sentence m which
Wtllh Geuder (personal communication) observes that verbs which behave in this way (or at least some of them) have the semantic property that the qualification of the target state speaks about the object
and
the subject of the
verb. On the other hand, the relic verbs discussed m the preceding sections all had the property that only the object was a 'constituent' of the target state. I will show that we can explam the new behaviour of
wieder
by mirroring the
difference between the two classes of verbs in the semantic decomposition. The solution is largely based on suggestions by Willi Geuder.
To see what is happening, let us explicitly compare the relic verbs
open' and
(6-2)
Ojfnen 'to
verlassen 'to leave'.
a. Er offnete dte Tiir (RESULT: Object is open)
Anna verlieB das Haus (RESULT: Subject is outside Object) 'He did an acnon whose result was that the door was open'
b.
'Anna did an action whose result that she was out of the house'
obJ�t result verbs, and verbs of the second class holdrr + object result verbs. Other verbs of the latter kind which allow a restitu tive reading with wieder in pre-object position are found in the following list
Let us call verbs of the first class
(the examples come from Willi Geuder):
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
6
1 10
The
(6-3)
Different Readmgs of Wttdtr 'Again'
weil er wieder seine Sachen zunickbekam {RESULT: Subject has Object) because he again his belongings back got 'There was an event whose result was that he had his belongings back again' b. daB er nach dem Unfall wieder eine posinve Emstellung zum Leben bekommt (RESULT: Subject has Object) that he after the acadent again a positive attitude towards life gets c. daB sie nach dem Schlaganfall wieder das Gehen lernt (RESULT: Subject masters Object) that she after the stroke again the going learns d. deB er wieder den Hut aufsetzte (RESULT: Subject has Object on his head) that he again the hat on put
a.
gave me the following interesting minimal pair:
(6-4) a. weil er wieder seine Sachen zunickgab (repetitive/restitutive) because he again his belongings gave back b. weil er wieder seine Sachen zuriicksendete (only repetitive) because he again his belongings backsent The referee comments on these data; 'Notice that zuruckgeben implies a change of possession (subjeCt and object are involved), whereas zuriicksenden arguably only Implies a change oflocation of the object.' It is interesting to observe that the contrast disappears If we make the addressee of the giving!sending exphcit: (6-s) a. weil er dem Fitz wieder seine Sachen zunickgab (repetitive/ restitutive) because he ARTdat Fntz again his belongings back gave b. weu er dem Au tor wieder sein Manuskript zuriicksendete (repetitive/restitutive) because he ARTdat again his manuscript back sent
The generalization that emerges is that when wieder precedes the object we can have the restitutive reading only in those cases m wruch a 'subject' enters the description of the target state. This subject may be identical with the agent (Geuder's examples), or it may be different (the referee's examples). The subject of a state is always a holder. This is why I speak of holder + object result verbs. Verg� 'to forget' and verlin-en 'to lose' should behave in the same way, but for some reason they do not fir into the picture-again an observation by Geuder. The anonymous referee observed that the situation is slightly more complicated: there 1s at least one use of verlieren where we find a restitutive reading with wieder preceding the direct object. This is his example:
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
An anonymous referee
Arnim von Stechow I I I
(6-6)
Zwar
habe ich im
Indeed have I habe schon
ElsaB funf Kilogramm zugenommen, aber ich
in-the ElsaB five kilogramms gained,
wieder zwei Ktlo
have already again
but
I
veloren.
two kilos lost
In order to explain why these verbs allow a resntutive reading even if
wieder
precedes the object, let us have a closer look at their meamng. Consider the example with
(6-7)
verla.ssen
'to leave' again.
a. Anna verheB wieder das Haus Anna left b.
again
the house
Recall that our assumptions on Case marking state that accusanve Case is
GF of VoiceP. This suggests that the accusative of the object das Haus IS not assigned in the VoiceP headed by agent but rather in a subordinate assigned to a
VoiceP headed by holder. In order to make our Case theory work, we have to complicate the syntactic decomposition of the new class of result verbs. The relevant part of the structure could be as
in (6-8)
the object moves to the
BECOME
_7---v-obolder(ce �
das Haus the bouse
out
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
3e[agent(Anna)(PAST(e)) & BECOME(,J..s [again(,l..s[out(the house)(s) & holder(Anna)(s)])(e)])]
I I 2 The Different Readmgs of Wrtd" 'Ag2m'
Accusative posinon where its Case is checked. Wieder is attached to AgrO-P. Since AgrO is semantically empty, this structural position is equivalent to an adjunction to the lower VoiceP and we obtain the restitutive reading as desired. The structure presupposes that a voice phrase is not necessarily dominated by AgrO-P. To be sure, AgrO m-selects Voice, but nothing forces us to have that node in the syntax. It is inserted optionally. The analysis of unergative verbs, whiCh have an accusative object only if there is a small clause, motivates the same assumptimL Furthermore, this optionality is needed to derive the repetitive reading of sentence (6--9a), which is (6--9b):
BECOME(As[out(the house)(s) & holder(Anna)(s)])(e)])]
An mspection of the formula shows that wieder must be base generated higher than the agent projection, which is the highest VoiceP in the tree (6-8). Since the direct object precedes wieder, ir either has been moved to the (Spec, AgrO] pos1tion determined by the highest V01ceP, in which case the lower AgrO-P is nussing, or it has been scrambled from the lower (Spec, AgrO] to an adjunction position and the higher AgrO-P is missing. Finally, let me quote a crincal observanon of an anonymous referee: I find clut quannfied NPs resiSt the posmon in quesnon. A sentence hke wd u witd"fastj(des Buclr zuriick�kam seems to have the repennve readmg only, m conrrast to werl ufastjdt> Buclr wrtd" zuriick�kam. If nght, thiS means that the Accusanve postnon m (6-8) IS not a full DP postnon.
(6-1 0) Ich habe inzw�schen wieder jedes Buch/die meisten Bucher zuriick I am not sure about this. It seems to me that the sentences
bekommen I have meanwhile again every book/most books backgot
can have the restitutive reading.t
7
WIEDER I N P R E D I C A T I V E A N D A T T R I B U T I V E P A R T I C I P I AL P H R A S E S
The point I want to make in this section IS that the readings observed with wieder in participial phrases suggest that attributive partiap1al phrases of unergative verbs may have pass1ve voice, whereas there IS no passive in predicative participial phrases. The voice projection is responsible for the repetitive/restitutive ambiguity.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(6-9) a. Anna das Haus WIEDER verlieB b. 3e(again(PAST(e))A.e[agent(Anna)(e) &
Anum von Srechow I I 3
(7- I ) a. Der Raupenschlepper ist wieder repariert (restitutive/repetitive?) the power carrier is again repaired b. ?Der Raupenschlepper ist wiederum/erneut/noch einmal! the power carrier is again..e/again..e!once more/ ein weiteres Mal repariert (only repetitive) one more nme repaired c. Der Raupenschlepper ist wiederum/erneut/noch einrnaJJ the power carrier is again,.e/again..e!one more/ ein weiteres Mal repanert worden (only repennve) one more time repaired been The data to be analysed are:
(7-2)
a.
Der Raupenschlepper ist wieder heil the power carrier is again intact b. ??Der Raupenschlepper ist wiederum/erneut/noch einmaV The power carrier is again..e!again..e!once more ein weiteres Mal heil time intact one more
Perhaps we can explain this distribution if we require that the repetitive adverbs mentioned require the presence of a voice phrase}6 Let us therefore tentatively risk the following generalization: (7-3) wiederum, erneut, noch einmal, ein weiteres Mal, . . . arrach to VoiceP and possibly to higher projections). The examples suggest, then, that 'statal passives' do not contain a voice phrase and are therefore not passives, a claim defended for independent reasons in Kratzer {1 994). The examples in (7-1 b) are not totally unacceptable, but there is a clear conrrast between them and the (7- 1c)-examples, which are instances of the verbal passive ('Vorgangspassiv'). My guess is that we reinterpret the (7- 1 b) sentence as verbal passives if we accept them. The situation is different with attributive participial phrases; here the repetitive adverbs can be inserted without any problem: (7-4)
a.
der wieder reparierte Raupenschlepper (repetitive/restitutive) the again repaired power carrier
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The predicate in these sentences 1s what German grammanans call a Zustands passiv 'statal passive'. A term more familiar in Generative Grammar 1S adjectival passive. Here wieder seems to have only the restitutive reading. An anonymous referee has pointed out to me that certain German adverbs have only the repetitive reading: erneut, noch einmal, and ein weiteres Mal, among others. These adverbs combine poorly Wlth the statal passive. The contrast 1s observed even more sharply 1f we mod1fy an underived adjective:
1 14
The Different Readtngs of Witdn- 'Agam'
b. der emeut/ein weiteres Mal!noch einmal!dreim.al repariene Rupenschlepper I tentanvely conclude that attributive participial phrases may have a voice projection. This conclusion is not in agreement with Kratzer ( 1 994), who claims that attributive pamcipial phrases (her verbal participial phrases) are purely verbal. I say something about her motivation at the end of this section. Let me present the analysis first. Concerrung predicative pamcipial phrases, we follow Kratzer ( I 99+ 39) with slight changes which are due to our decompositional approach. The LF of the predicative partictptal phrase in (7- r a) is this:
In other words, the head of the phrase is the adjective morpheme PERF, the perfecnvuer, which min- and max-selects V and VP, respectively. Kratzer's semantics for PERF is this:17 (7-6) PERF is a symbol of type ((s, t), (s, t)). fERI1J(P)(s) - I iff3e [P(e) - 1 & s - ftarget(e)] , where P ts any property of events with a target state and s lS a state. PR01 is interpreted as �; hence, this symbol has to be brought to an operator position at LF.18 �� is a function which assigns relic events their target states. The notion of target state is moovated in Kratzer ( 1 994: 32 f£) and in Klein ( I 994). Given these assumpnons, the LF translates mto the following formula: (7-7) � [PERF(BECOME(again(intact(xj)))(s)] Given an appropriate semantics for the 'control copula', which lS the identical mapping,19 we can predicate this of the subject and obtain the wanted restitu tive reading, namely that the power carrier is in the target state of a repatring after having been broken. What happens ifwe attach wieder to VP or to A'? We would obtain formulae expressing 'intermediate readings' (c£ Section 3� i.e. the power carrier has been repaired before but not necessarily by the same people.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(7-5)
Arru m von Stechow
I Is
(7-8) a. �[PERF{again(BECOME{intact(xJXs)) b. �[again{PERF{BECOME{intact(xJXs))
(7-9)
e
V ceP
-------
again
A.e:VolceP
p�Wce
�Voice -�
n l.j
Intact
BECOME
agenl(x)(e)
And here is the semantics of STATE
(7-1 0) STATE is a symbol of type ((s, t), (s, t)). f:STATEIJ{P) - [PERF1(P), ifp is a property of relic events. 1STATEIJ{P) - P otherwise.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Predicate (a) applies to the power carrier if it is in target state of a repairing and if it had been repaired before; (b) applies to the power carrier if it IS in the target state of a repairing and if it had been in the target state of repairing before. Both predicates are equivalent and express repetitions. I find these readings hard to get, but see no obvious way to block them in my account. Let me turn to attributive participle phrases next. I speculate that they may contain a voice phrase. Furthermore, these participles have adjectival inflection. This suggests that the sequence of morphemes IS Verb + Voice + A + INFL. Thus a possible LF for the participal phrase in (7-�) could be (7-9). As we said before, [Spec, A) licenses PRO. We could leave it there, but given that we have a [Spec, I) position, we use the latter as the operator posmon. Similarly, we have to say that the morpheme STATE-the semanncs will be given in a moment selects Voice. And Voice has to license proacc. See also Rapp (1995), where a rather similar analysis is proposed.
1 1 6 The
Different Readings of Wmla 'Ag:un'
This somewhat ugly disjunctive definition takes account of the fact that �ERF1] is not defined for properties of non-relic events. We will cite an example of Kratzer's in a moment. First, let us consider the translation of the LF mto a formula. After some !-conversions we obtain: (7- I I) �3s(STATE(again(t..e [BECOME(intact(xJ)(e) & age nt(proXe)] ))(s))
(7- 1 2) a. E in von zwei Rappen gezogen-er Wagen (Kratzer I 994: (6)) a by two black horses drawn-agr cart b. *Der Wagen ist von zwe1 Rappen gezogen (Kratzer I 994: (ro)) the cart is by two black horses drawn (7- I 3) a. Die heute morgen gemalt-en Stilleben (Kratzer (1994: (7)) the today morning painted-agr still lives b. *Das Stilleben war heute morgen gemalt (Kratzer 1 994: ( 1 1 )) the still life was today morning painted She says (Kratzer ( 1 994: 49) that the ungrammaticality of the (b)-examples show that reduced relatives, i.e. attributive parocipial phrases, have verbal and not adjectival participles. She also mentions an observation by W1lmanns and Paul 'that the aspect of reduced relatives formed from verbal particrples matches the Aktionsart of the verb from which the participle is formed. She concludes: 'This shows that the verbal parnciple affix is not an aspectual operator (or it would have to be a triVial one). In particular, it is not a perfectivizer.' I was not able to appreciate these remarks fully because the manuscript I have doesn't contain a detailed analysis of reduced relatives. I have no problems with process verbs. Kratzer's analysis of the verb ziehen 'to pull' would be (c£ Kratzer (1994: 39): (7-1 4) zieh-*
hle(pull(x)(e)) 'e is a pulling of x'
(The asterisk indicates the translation into the logical language.) Neglecting the adverbial von zwei Rappen , the parncipial phrase in (7- I 2a) is perhaps translated as something like: (7. 1 3) We[pull(xXe)) We can intersect this with cart and obtain (7-I 4) We[pull(x)(e) & cart(x)]
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The formula expresses the property true of an individual a iff there is a state which is the target state of someone's (- pro's) action of repairing a, where a has performed such an action before. Clearly, this is the intended result. Kratzer ( I 994) argues against an analysis which regards attributive participial phrases as passivized. Here are some relevant examples (Kratzer I 994= 48):
Armm von
Stechow
I I7
This is the property rrue of individuals which are carts and are in the process of being pulled. Let us look at the result verbs next. I don't want to consider
Kratzer's example {7-1 3 a) because verbs of creation like
malen
'to paint' raise
further complications. Let us consider a simplified version of our original example instead: (7-I s ) der reparierte Raupenschlepper the repaired power carrier Kratzer would analyse ( 7- 1 6) repaner-•
reparieren 'to repair' as something like this:
Applying the same method as before, we could obtain something like the following translanon for the modified noun: (7- 1 7) We [repaired(x)(�t(e))
& power carrier(x)]
This is an attractive analysis. The reason for not taking 1t over 1s that I believe that Kratzer's decomposition of telic verbs is not general enough. See appendix 2. My analysis is more complicated. Stated in Kratzer's terms, the complex noun is translated as: (7- 1 8) J..x3s3 e [repaired(x)(s)
carrier(x)J
& s - �t(e) & agent(pro)(e) & power
This result is achieved by the perfectivizingjob of the STATE-operator, which licenses a 'passive' vo1ce head. A remark on the ungrammaticality of the examples (7- 1 2b) and (7- 1 3 b) is in order. Presumably, the von-phrase in (7- 1 2b) has to bind a pro at LF, although the details have to be worked out. In any case, the presence of such adverbials is a classical diagnostic for passivization, and predicanve participles are not passivized, although I am not sure how reliable the diagnostic is. As for (7-1 3 b),
this example contains the frame setter arguably has a position outside the VP.
heute morgtn
'this morning', which
Finally, let me stress that I do not claim that attribunve participial phrases are obligatorily passivized. Unaccusative verbs cannot be embedded under Voice, as we know. Take the following example: (7- 1 9) der wieder abgestiirzte Raupenschlepper the again crashed power carrier
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
A.xA.e[repaired(xX�t(e))J 'e is an event at whose end x is repaued'
1 1 8 The
Different Readings of Wrrda 'Agam'
Everything is as before, except that the voice phrase is missing; i.e. the repetitive reading of the participial phrase has roughly the following analysis: (7-20) [a.PR01(M, tj' (VP again (VP(XP tj down) BECOME]) STATE) INFL) If we have two possibilities of analysing attributive participle phrases, which one should we choose? Kratzer (199+ 4) points out the criterion of possible coreference between (implicit) subject and object, which might help: (7-2 1 )
Das Kind war gelcimmt (Kratzer 199+ (7a)) The child was combed (stative) b. Das Kind wurde gekammt (Kratzer 1 994= (7-b)) the child got combed (eventive) a.
(7-22) Das sorgfaltig gelcimmte Kind the carefully combed child If the pro-subject of a passive construction were coreferennal With the PRO moved to an operator position, a violation of principle C would arise. This would predict that the comber couldn't be the duld, contrary to the facts. On the other hand, the Case theory outlined in section 4 entails that we should have a voice projection if we have a structural dative in the participial phrase.20 (7-23)
Die ihrn gegebene Macht war gewaltig the lumdat given power was enormous b. Er trocknete die ihm gesalbten Fiiik he dried the hi�t rubbed with ointment feet a.
It seems to me that the dative pronoun cannot be coreferential with the implicit subject of the participial phrase. Thus, clus seems to be the correct prediction. If it were true that the repetitive adverbs mentioned did require a voice projection, one would expect passivization in the attributive participial phrase: (7-24) Das noch einmal sorgfaltig gekammte Kind wurde sehr bewundert the agaiDnp carefully combed child was very admired The presence of a voice projection would predict that the child could not have combed hersel£ This is the prevailing reading, but I am not sure that self combing is totally excluded. All this is very subtle, and more is to be said about it. I will leave the matter at this stage. It seems to me that nothing in this section contradicts the general thesis of this paper that the different readings encountered with wi�der should be analysed in terms of syntactic scope. The
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
She comments: '7(a) is compatible with the child having combed herself, 7(b) is not.' The non-coreference in (b) follows from the Binding Theory for the analysis of the passive she assumes. It follows that the attributive participial phrase in (7-22) should not be analyzed as passivized:
Arrum von Stechow 1 1 9 problem of this section is not that we do not have enough structural distinc
tions. There might be too many.
8
WIEDER
A N D S L O P P I NE S S
(
Terms become sloppy i fone o f the parameters which determine their reference
is bound by an operator. In this section we will derive Fabricius-Hansen's I 98 3) observation that sentence (8- I a) exhibits a sloppy reading with respect to the
term den Namen ihres Mannes 'the name of her husband', which is not present in
sentence (8- I b):
sloppy)
(because) Anna again the name of her husband takes
b. Anna den Namen ihres Mannes wieder annahm (only strict)
Anna the name of her husband again takes
It is plausible that the availability of the sloppiness has to do with whether 'the
name of her husband' is in the scope of wieder.
Wieder mvolves two different
events, and the husband might have changed in between. The explanation of the data will be that if the term 'Anna's husband' is in the scope of wieder, it will
have sloppy interpretation if its event variable 1s bound by
wieder. If the event
vanable remains free, we will have a strict interpretation. If'Anna's husband' is
not in the scope of wieder, the event variable Will not be bound by the term will have a strict interpretation with respect to
wieder, and
wieder.
Let us comment on the different readings of the sentence first. Sentence
(8- 1 a) may describe a situation in which Anna was first married to Otto and had
taken his family name. Then they got a divorce anq Anna top� her maiden
name again. Mter a while, she again took Otto'� family name, ei�er by
marrying him again or by some other legal act. This is the strict reading for 'the name of her husband'.11 The sloppy reading suggests two marriages to two
separate men, with a divorce in between, and m each case Anna took the family name of her husband.
The explanation of the data is that the term 'the name of her husband' has an
event (or temporal) parameter which can be bound by the A.-operator associated with
again. When this is the case, we have a sloppy reading. Binding requires
that the binder c-command the bound material. This is the case in example
(8- � a) but not in {8- 1 b). Note that
annehmen
'to take on' is a
holder + object result verb and
(8- ra) can therefore have a restitutive reading although position.
that
wieder is in a pre-object
Let us derive the sloppy/strict ambiguity of sentence (8- I a). In order to
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(8- 1 ) a. (weil) Anna wieder den Namen ihres Mannes annahm (strict!
1 20
The
Dlfferent Readmgs of W1�a 'Agam'
simphfy the formulae, we represent the term 'the name of x's husband' by the symbol t:(x), whose meaning is defined as follows: (8-2) £:(x): - the name of the person who happens to be x's husband at the time of e. Even if x is kept constant, the value of fe(x) may be different for different e's. Suppose first that the object is in the accusative position embedded under hol der. The formula representing the sloppy reading of clns configuration is the following (after several /,.-conversions):
holder(Anna)(s)])(e))){PAST(e))]
The formula expresses a repetition: Anna married agam and took the name of the new husband. Suppose next that the direct object is in the accusative position above VoiceP. The sloppy reading created by this representation is given by the following formula: (8-4) 3e [again(Ae[l..x [ag.(Anna)(e ) & BEC.(As [poss.(xXs) & hold .(Anna)(s))Xe)](£:(Anna))])(PAST(e ))] - 3e [again(l..e[ag.(Anna)(e) & BEC.(As [poss.(£:(Anna)Xs) & hold. (Anna )(s)])(e)])(PAST(e))]
The difference between the two readings is very subtle. In (8- 3 ) the name of Anna's husband is determined only after the completion of the action, whereas in (8-4) the name of the husband is not affected by the action. The first scenario is very unplausible. Thus, for pragmatic reasons, both readings amount to the same 'sloppy' repetition. Let us turn to the restitutive sloppy readings of(8-Ia) next. In principle, there could be three of them, because again could mod1fy the possess phrase, the possess + holder phrase, or the BECOME phrase. In fact, the poss1bilines are reduced to two because the direct object is in the accusative position, which embeds the possess phrase. After performing several /,.-conversions, we are left with the following two formulae: (8-s)
3e(ag .(Anna)(e) & BEC.[aga in (A.s[poss.(£:(AnnaXs)) & hold.(AnnaXs)])](PAST(e ))) b. 3 e(ag.(Anna )(e) & aga in(BEC. A.s [poss.(£:(Anna))(s) & hold.(Anna)(s)])](PAST(e )))
a.
In (8-sa) Anna did some clung which caused her to have the name of her current husband, and she had the name of a husband of hers before, either the same one or another. If he is the same person, the interpretation coincides wuh the
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(8- 3 ) 3 e[ag a in(Ae[agen t (AnnaXe ) & BECOME{M[possess (£:(Anna)Xs) &
Armm
von Stechow 1 2. 1
restitutive strict reading, which we will discuss in a moment If it is another husb;nd, then the reading coincides with the repetitive sloppy reading for pragmatic reasons. {8-sb) is the 'intermediate' reading. I won't comment on it, since it presupposes a rather peculiar scenario. Strict readings are obtained by leaving the event variable of £:(x) free in the scope of again. We can also achieve this effecet by QR-ing the direct· object at LF. Let us choose the former option; i.e. we can leave the event variable simply free in the scope of again. The reader may check for himself that we are left with three strict interpretations for (8-Ia) : {8-6)
3 e [again{Ae ' [ag.(Anna)(e ') & BEC.{As [poss.(({Anna))(s) & hold.(Anna)(s)])(e ')])(PAST(e))] b. 3e(ag.(Anna)(e) & BEC.[again {As [poss.(((Anna))(s) & hold.(AnnaXs)])] (PAST(e))) c. 3e(ag.(Anna)(e) & again [BEC. (As[poss.(((Anna))(s) & hold.(Anna)(s)])](PAST(e)))
a.
(8-7) a. The sloppy configuranon off.,(x) with respect to again: e is bound in the scope of again : again(A.e[. . . £:(x) . . .])(e) b. The strict configuration of £:(x) with respect to again: e is free free in the scope of again: again{A.e' [. . . f.,(x) . . .] )(e' ) Strict readings require the event variable of£:(x) to be free in the scope of again . A typical case is an LF where the term is moved to a Case position lugher than again: (8-8)
A typical strict reading configuration occurring with again fe( x)(l..y [ agai n(A.e [ ... y ... ] )(e)])
i
.j,
Movement to [Spec, AgrO]
If movement to [Spec, AgrO] did always go in tandem with a strict intepreta tion, this would mean that we could not reconstruct this type of movement at LE The matter deserves more scrutiny, however. It is not clear to me that the following example excludes a sloppy interpretation:
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The most natural strict reading is represented by {8-6a) . This is the repetition in which Anna twice took the name of the same man. This is the restitutive reading for the same husband. (8-6c) is the 'intermediate' reading. Let us resume the sloppy/strict configurations. For a term to be sloppy, the event variable must be bound by the A.-operator following again:
1 22
The Different Readmgs of Witdtr 'Agam'
(8-9)
Wladimir hat seinen Computer
wieder rumiert
Wladimir has his computer again rumed If a sloppy reading is possible here, we have to reconstruct the direct object in a position at LF where we can bind the event variable of the description.
9
C O U N TE R D I RE C T I O N A L
WIEDER ?
One of the central semantic ideas u nderlying the explanations given in this article 1s to reduce the apparent ambiguity of 'again' to one basic reading,
repttition)
as such
is
repetition .
restitution , and the ( restitution/
The terminology
rrrelevant. What really matters i s whether 'again' has
narrow or wide scope with respect to the relevant 'aspectual' operators, espe
(
cially to BECOME. I take 1t that the spirit behind the methodology of this
e nterprise follows Dowty I 979)
in the essential respects. In this section I want
to defend this 'classical' approach against possible objections arising from a recent proposal by Fabricius-Hausen. Fabricius-Hansen22 points out on several occas10ns that a scope solution for
the ambiguity has to face empirical difficulties to do with verb pairs like to
rise,
which may be called
counterdirectional
tofall!
antonyms.23 Consider the
following pair of sentences: (9- 1) a. das Barometer WIEDER fiel (repetition) the barometer again fell b. das Barometer wieder FIEL (restitution?) Clearly, (9- r a) has a repetitive sense: the barometer fell after having fallen before. But what about the restitutive reading? Such a thing does not seem to exist. On the other hand, we have a very clear intuition about what (9- 1 b) means: the barometer fell after a rising. This reading is called
by Fabricius-Hausen. The terminology
counterdirectional is well motivated because the sentence
does not seem to express the restitution of a state but rather the continuation of motion or transition in 'reversed direction'. In recent
talks,u
Fabricius-Hansen generalizes the concept of counter
directionality by defining a relation CONTRA between properties. Her idea is
that the old notion of restitution is a special case of the new, more general concept. Fabricius-Hausen assumes rwo operators,
RES and PRE, which assign
the target state property and the pre-state property to a temporal property. For lnStance, (9-2) a. PRE(fall(x)) b. RES(fall(x))
- higher(x) - lower(x)
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
namely the repetinve one: the repetition of a state is called a repetition of a becoming is called a
Arnim von Stechow
123
(Cf Fabricius-Hansen 1994b: Fig. 1a.) The relation CONTRA holds of two properties iff they have reversed pre-state and target state properties: {9-3) CONTRA(T, T') ifiRES(T) - PRE{Tj & RES{T*) - PRE{T) (Fabricius-Hansen 1994b: de£ 12a)
T and T' are properties expressing transitions. It follows that 'fall' and 'rise' stand in the relation CONTRA. The final concepts we need in order to understand Fabricius-Hansen's new theory are the descriptions Prer(e) and R�r{ e), which denote the pre-state and the result state of the transition e, where e has property T:
J,.. assigns to any event 1ts pre-state. !target is Kratzer's target state function. Fabricius-Hansen's (1 994b: de£ I JC) meaning rule for 'counterd1rectional resntunon' 1s the following:
{9- 5) againc{T)(e) 1ff T(e) & 3e* IT*(CONTRA(T, T') & T*(e*) & TIME(ej < TIME(e) & ResMAX(T")(e*) - PreMAX{T)(e)] , where T and T' are properties expressing transitions.
(I am using the mdex 'c' for counterdirectional again in order to avoid confusion with my own defirution.) TIME is a function that assigns every event its running time. (It seems to me that the formulation of this meaning rule is unnecessarily strong. As far as I can see, the last conjunct could be replaced by the simpler condition �Xej, and thus the descriptions in (9-4) would be dispensable for the theory.)23 The 'counterdirectional-restitutive' reading, which is expressed by (9-rb), is now formalized as (9-6):
(9-6) 3e[againc(fall(the barometer)){PAST(e))]
The formula expresses correctly that the falling occurs (immediately) after a nsmg. At this point the question arises as to how the repetitive reading, i.e. sentence
(9-1a) , is represented. I am not quite sure what Fabricius-Hansen's position actually is. In her (199�) she simply assumes a lexical ambiguity. Let us denote this meaning by � · In other words, (9-1 a) is represented as (9-7) : p (9-7) 3e[�p(fall(the barometer)){PRES(e))] The meaning of � is the same as that described in rule (9-20) below. It means that the property in question is instantiated at the evaluation time and had been instantiated at an earlier time. The necessity of introducing a lexical ambiguity is a disadvantage of this proposal S techow (199 5) raises this criticism against Dowty's (1979) attempt to assume a lexical ambiguity for again.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
{9-4) a. Prer(e) :- LS. �(e) - s & T(e). b. ResT(e) :- LS. Lr t(e) - s & T(e).
1 2-4
The Dlfferent Readmgs of Wiedrr 'Again'
Another point, which is more troubling, is that there ts no way to explain the
loss of the resntutive-counterdirectional reading if wieder precedes the subject at surface structure. One of the crucial
aims
of the theory defended in this
article is precisely that of offering a structural explanation in terms of scope for this fact.
An attractive feature of Fabricius-Hansen's proposal is that 1t applies to 'relic transitions' as well as, for instance, to the verb pair einschlafen! aufwachen 'to fall asleep/to wake up'. The properties expressed by these verbs stand in the CONTRA-relation. Hence, the meaning rule (9-5 ) applies to them.26 For instance, we can formalize the meaning of (9-8 a) as (9-8b):
because Otto again up woke
b.
3e (againc(wake-up(Otto)){PAST(e)))
The formula correctly expresses that Otto first fell asleep, then Otto slept for a whtle, and finally woke up. A sentence with repetitive
�P' of course. Again. we
wieder is formalized with
would like to have an explanation for the fact
that we have only the repetinve reading if wieder precedes the subject Next, consider a pair of transitive CONTRA-verbs, i.e.
Ojfnen/schliefien
open/to close'. In Fabricius-Hausen ( 1 994b) we find the entry x y
'to
ofiD.en 'x
open y' for 'to open'. Carried over into a Davidsonian approach, tlus means that the verb has to be formalized as
open(y)(x)(e).
Therefore the restitutive
counterdtrectional reading is represented as (9-9b): (9-9) a. Ali Baba Sesam wieder offnete
c
Ali Baba Sesame again opened
b. 3(again (open(SesameXAli)(PAST(e ))))
Let us assume that
PRE(openv(SesameXx)) - closedA(Sesame) - RES(closev(Sesame)(y)) and
PRE(closev(SesameXy)) - openA(Sesame) - RES(openv(SesameXx)) where
x and y are arbitrarily chosen subjects. It follows that the transitional open(SesameXx) and close(Sesame)(y) stand in the CONTRA
properties
relation. Therefore formula (9-9b) says that Ali opened Sesame after Sesame
had been closed by someone else or had closed by itsel£ These truth conditions conform to our intuitions.
Again. it would be nice to be able to explain why the sentence
(9-1 0) weil Ali wieder
Sesam offnete
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(9-8) a. weil Otto wieder AUFwachte
Armm von Stechow 1 2 5
has only a repetitive reading. Why does it have exactly the same logical form (9-9b), with the only difference being that we have to choose agaiDnp Instead of againc? I have no idea how Fabricius-Hansen's new theory would answer this question. Let us take up the question of how 'counterdirectional' readings are treated in our approach. Ignoring tense, the main idea for an analysis of (9- I b) is resumed in the following paraphrase:
(9- I I ) e 1s an event at whose end the barometer is again in a lower state.
This is the restitution of a state of the needle. Ignoring tense, the formahzanon is this: Here beg(e) is the beginning of the event e, i.e. its left boundary. In this formula, MORE 1s the (abstract) comparative morpheme. For my purposes, the comparative semantics of Seuren ( I 97 3) will do: (9- I 3) IMOREII(P, Q) iff3d [P(d) & -.Q(d)) , where P and Q are properties of degrees.
Together with Seuren, I assume that properties of degrees are downward entailing in the sense that if a degree has some property, then any lower degree has that property as welL d.:.low.(x) means that the 'lowness' degree of x at state s is d. Increasing lowness means decreasing height, of course. For a construction of degrees of lowness, smallness and so on, see Stechow ( I 984b).27 Thus the formula (9-12) means that at the end of e, the barometer is down to a level which is lower than its level at the beginning of e, and the presupposition is that at some earlier time it has been down to a level lower than its level at the beginning of e at some earlier time. At this point, it becomes obvious that the terminology resitutitive versus rtpetitive is not always appropriate. (9- 1 2) does not express the restirution of a particular state of the barometer because the fall of the barometer might be longer than its previous rise; 1.e. the final state of the falling might be lower than the state at the beginning of the rising. So we might speak of the resnrution of a lower degree: the comparative MORE contains an existential quantifier and this one is in the scope of again. I would not call such repetition a restirution. The terminology is simply out of place. An anonymous referee comments on this analysis as follows: There 15 one situation in which this aru.lysu seems implausible, and this is the one actually mentioned . . . Assume times to. t1, t2 � where to is the first time under consideration. Assume the following barometer values: • •
to:
Soo, t1: Sso, t2: 9()0, t3: 850, t4: Soo, �: 750, t,s: 700, r8: 650
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(9- 12) the barometer{Ax.BECOME [again(As [MORE [Ad.d-low.(x), Ad.d-lowbeg(e)(x)]]))(e))
1 26
The Different Readings of W1td!r 'Agam'
I thmk we on say that for all points between t1 and rs. da5 Baromtttrfollt WJtti!T. But nonce that for the mterval, say, (�. �]. there IS no barometer value m the past that was lower than the barometer value at t5, 750. The present analysiS would say that only the mtervals (t1, t3], (t3, r.J make the sentence true.
(9-14) the barometer(Ax.again [BECOME(N;(MORE(Ad.d-low.(x), Ad.d-lowbeg{e)(x)]])](e)) In other words everything is as before. What about counterd1recnonality? In the classical approach. this term has no theoretical significance. Regardless of whether we have a repetitive or a 'counterdirectional' reading of our sentence, there must be a last preceding rising. The repentive reading entails that there was a previous falling as well. The 'counterdirectional' sense does not require this, but does not exclude 1t either. It could be the case that we have in mind the last ns1ng when we express the 'counterdirectional' meaning. If there were a prevwus falling, we should have uttered a sentence expressing the repentive reading according to the Gricean maxim of informativeness. It is therefore plausible that we have in mind a previous falling when we utter a sentence expressing the repetitive sense. All this is pragmancs and psychology, not semantics, one might argue. Returning to our analysis of'to fall', I have to add that the decomposition is the same which Dowty (1 979: 88 ff.) seems to have in mind for what he calls 'degree-achievements'. Typical examples are these: (9- 1 5)
a. The soup cooled for ten mmutes b. The soup became cooler for ten minutes c. *The soup became cool for ten minutes
If 'the soup cooled' is analysed as 'the soup became cooler', the expression denotes a process and can be modified by the durative adverbial 'for ten minutes' (perhaps combined with the invisible quantification 'for each relevant subinterval of', as Dowty (1979) assumes for that adverbial). (9-1 5c), on the other hand, expresses a genuine achievement and cannot be combined with that adverbial (There is a reading in which the sentence is acceptable, namely
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Yes and no. If we understand das Barometerfollt in the perfective aspect, i.e. the event time coincides exactly with the reference time, then I would agree. But in German we have no grammaticalized aspects. If we evaluate the sentence with respect to the interval (t5, �]. we silently switch to the imperfective aspect. That is, we consider the interval as part of a larger interval, say, (t2, �]. which can make the sentence true. In other words, we evaluate the 'progressive' meaning 'the barometer is falling again' with respect to the interval (t5, �]. For convenience, we will keep Fabricius-Hansen's repetinve/restitutive distinction. The meaning of(9-1 a), a repentive read1ng, is then represented by
Armm von Stechow 1 27
when the adverbial qualifies the result state.) It 1s the inherent comparativeness which makes 'degree-achievements' processes. Telic trarlSltions are achieve ments or accomplishments and therefore cannot be combined with adverbials of the said kind. The decompositional approach gives a uniform explanation of these apparently unrelated facts. In order to describe the restitution of stative properties, Fabncius-Hansen (1 994b) introduces a further meaning rule (her rule 1 4):
(9- 1 6) again_te(ZXs) iff Z(s) & 3e3T(Z - RES(T) & TIME(e) I< TIME(e) & againc(e)(T)] , where Z is a stative and T is a transition.
(9- 1 7) Anna wieder gesund ist Anna again healthy is would be formalized as: (9-1 8 ) again_te(healthy(AnnaXs) Some comments are in order. Fabricius-Hansen's semantics for this case is more complicated than that for transitions, because rule (9- 1 6) uses the rule for transitions in the definiens. In fact, the formula says that Anna's being healthy 1s the result of an immediately preceding becoming healthy for which there is a counterdirectional becoming sick. Since the latter requires Anna's being healthy to be a pre-state, Anrla must have been sick before. This analysis strikes me as somewhat tortuous. In any case, the treatment runs counter to the tradition. according to which the modification of statives is always the primitive case. In our more traditional approach, the formalization is exactly the same, but the semantics is very simple. The formula says that Anna's being healthy is later than a previous maximal state of her being healthy. Hence her health is restituted. Furthermore, it follows without stipulation that there is no 'restitutive/repetitive' ambiguity because the BECOME-functor is missing.28 The situation changes as soon as we have the auxiliary werden 'to become'.
Anrla wieder gesund wird (ambiguous) Anrla again healthy becomes b. wieder Anrla gesund wird (only repetitive)
(9- 1 9) a.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(The index 'state' is my own addition. I use it to keep this symbol apart from the other symbols for 'again'. !< means 'precedes immediately'.) It is not necessary to regard this rule as a further reflex of a lexical ambtguity of wieder 'again': the different meaning rules may rather be regarded as the qualification of different cases which the definition of the function expressed by the word has to distinguish. States are different entitles than transttions. Therefore, the function can assign them values m a dtfferent way. A sentence like
1 28 The Dtfferent Readmgs of Witdn- '�in'
The data are predicted by our theory. Fabricius-Hansen distinguishes a further class of properties, which she calls X These comprise achievements/ accomplishments which are not transitions of the kind mentioned. She illustrates this 'rest class' with examples such as the following ones: x auf lachen 'to burst out laughing', x winken 'to wave', x zwei Bier trinken 'to drink two beers', x zwei Kilo abnehmen 'to lose two kilos in weight', x 1 00 Meter fallen 'to fall r oo meters', etc. Whenever 'again' applies to such a prop erty, we have to apply the following rule (Fabricms-Hansen 1 994b: rule I 5c): (9-20) againe�.e(e)(X) 1 ff X(e) & 3e '[X(e ') & TIME(e ') < TIME(e)]
(9-21)
a.
Erst nahm Anna 5 Ktlo zu. Dann nahm s1e wieder zwe1 Kilo ab. First put Anna 5 kilos weight on. Then lost she again two kilos. b. Erst stieg der Ballon 200 Meter. Dahn fiel er wieder roo Meter. First rose the balloon 200 meters. Then fell it agam roo meters.
Clearly, the second sentences do not express the repetition oflosing a particular amount of weight or the repetition of falling a certain distance. Rather, they express the partial resntunon of an earher state: Anna regained the state in wluch she we1ghed two kilos less and the balloon regained a height it had earlier. FabriClus-Hansen can deal with these examples if she classifies the two properties as transitions. The counterdirectional property of losing two kilograms would then be the property of gaining two kilos. And the counter directional property of falling 100 meters is that of rising r oo meters. Thus the second sentence of the first diSCourse would mean that Anna lost two hlos and that this was after she had gained two kilos m werght. This would be true even if she had gained five kilos in weight. The repetitive sense is obtamed by the repetitive 'again'. The examples are interesting as such. So let us discuss the1r analysis in a decompositional approach.29 The restitutive sense of the second sentence in (9-2 1a) IS represented as (9-22): (9-22) BECOME(again(As(DIFF(2 kg}(MORE) (A<:l.d-LESS-WEIGH.(Anna), A.d.d-LESS WEIG�el(Anna)))])(e) DIFF(d) is the operator which accounts for differential comparanve readings. If MORE is of the determiner type, DIFF(d) is of the type which takes a determiner and yields a determiner as a value. It is defined only for MORE:
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(The last-resort subscript 'else' is my addition.) Fabricius-Hansen calls the readings described by this rule 'repetitive' ones. If I understand her correctly, she predicts that we find only a repentive readmg for each of the examples in the above list. Tlus, however, cannot always be correct, because somenmes we have ambiguities here, too.,
Armm von Scechow
(9-23) IDIFFft(d)(UM OR£ID(P. Q) - I iffiM O�(P, Q) - I & the max d*[P(d*) - I ] - the max d *[Q(d*) - I ]
1 29
- d.
(9-24) Erst ging Anna vowarts. Dann ging sie wieder riickwarts. First went Anna forward. Then went she again backward. This text strikes me as odd because the adverb T"Uckwtirts 'backwards' suggests that Anna did not tum around when she went backwards. Better continuations of the first sentence would he: (9-25) Erst ging Annas vorwarts. a. Dann kehrte sie wieder urn. Then turned she again around. b. Dann ging sie wieder zuriick. Then went she agam back.
It is a remarkable fact that these continuations pennit a decompositional analysis. Sentence (a) cannot be interpreted literally: Anna not only has to tum around, but also has to go back a bit 'to a point behind her back'. Clearly this is the resutution of a previous state. Or take the following sentence, a model case for counterdirecnonal 'again': (9-26) First, the tape moved to the right. Then, it moved to the left again. It allows a decompositional analysis as well; when the tape moved back to the left, it moved to a previous position. This is a restitution, with the proviso mentioned in the comment on the analysis (9-1 2). Finally, take an example pointed out to me by Willi Geuder: (9-27) Er strich den Satz wieder durch 'He crossed the sentence out again'
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The semantics for the relation LESS-WEIGH has to ensure that increasing LESS-WEIGH means the same as decreasing M ORE-WEIGH and vice versa. Thus formula (9-22) means that, at the end of e, Anna weighs two kilograms less and she had weighed that much before. This reading does not presuppose that Anna had lost weight before. The repetitive read1ng is expressed by giving again wide scope with respect to BECOME. Are there cases which are irred uably counterdirectionaL i.e. not analysable by means of an appropnate repetitition? I think such cases do not exist. If a 'counterdirectional' event is presupposed, then the ongoing event can always be described as the repetition or resutution of a state of that event. I have not proved that in general. I have rather convinced myself of the truth of this claim by gmng through the relevant cases. And here are some further ones. Fabricius Hausen ( I 994b: Fig. I b) gives the pair vorWrts gehen I riukwarts gehen 'to go forwards/to go backwards'. She does not discuss an example, so let us create one:
1 30
The Different Readmgs of Widn-
'Again'
A literal interpretation is not poSS1ble. The sentence means that he brought it about that the sentence was 'unactualized' again. Clearly, this ts the restitution of a previOus state of the sentence. My tentative conclusion is that there are no cases of irreduCible counter directionality. Thus our theory does not face emptrical difficulties. Further more, it can explain certain data which are still a problem for Fabricius-Hansen's new approach. A last point worthy of consideranon is that the decomposition method applies to other adverbs as well, e.g. to ambtguities arising with 'almost'. For a review of the relevant case, see Dowry (1979: Section 54). One would have to show how Fabricius-Hansen's theory applies to these data. C O N CL U S I O N
I think it ts fair to say that the present theory accounts rather well for a number of subtle facts. One thing is important for the explanation to work: movement to the structural Case position is visible in German, or at least movement to the Accusative position is. In the recent literature (e.g. Chomsky 1 992 or 1994), we find remarks that in English thts need not be so; i.e. Case is checked at LF. I have not sa1d anything about English Case. My explananon of the German data relies on the assumption that objects at least are at their structural Case positions at 'the surface' and adverbs do not move at LF. Under these assumptions, rather subtle facts of German can be explained amazingly well. I take it that this is a rather strong argument in favour of this proposal. ARNIM VON STECHOW
Received:
&minarfur SprachiVISSmschafi WilMlmstr 1 13 D-71074 Tubrngrn Gmnatty �-marl: armm.sw:how@um-tutbingtn dt
A P PE N D I X
r:
C O M ME NTS O N D OWTY's ( 1 9 7 9) ' a g a i n ' 2
The chssial examples whtch have monvated leXJal decomposition from the orne of
Generanve Semanncs on are ambigu ines ansmg from
agam
20.05 95
ReVISed vemon received: 01.02.9()
and
a/mast.
The most
complete survey and review of the relevant
luerature is found
m
Dowty (1 979� We have
not discussed almost, which mvolves quantifi
cation over worlds.JO Rather, we have con cenrrated on the Gerrnan version of nantely
agam,
Wltdn-. Th e classical analysis of agam
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
10
Arnim von Stechow 1 3 1
( 1) Otto makes the water run agatiL Suppose we represent that as (2), wtth the irrelevant dewls ommed:
(2) agamiThere IS an acnon x [Ono does x CAUSE water runs))
Dowty's problematic meamng postulate says that thiS statement means the same as (3) There IS an acnon x [Otto does x CAUSE again1 (water runs)] where again1 IS the repennve 'again', whose mearung we understand intumvely. Now Zimmermann assumes the followtng scen ano. Whenever Otto wants to make the warer run, he has to rum on the red up of the shower. It IS an essential property of the shower that if one operares the red up, the first flow of water is cold and the water heats up after a while. But there is a second, secret up known to the plumber only. When this one is rumed on, the shower delivers hot
water munediately. After the first insullanon of the shower, the plumber tested It by operating the secret up. Since the shower worked well, the first running of water was a rutm�ng of hot water. Let us assume that as a further essential property of the shower that the secret up can be used only once. after the first use It IS sealed for ever by some built-m mechaO!Sm. The second nme, the shower was used by Otto. He had to operate the red up. We can describe thts Situation by means of sentence (1� In VIrtUe of the parncular acnon done by Otto, Le. ruming on the red tap, (2) means the same as (4): {4) agamlthere IS an action x [Otto does x CAUSE cold water runs)) The mearung postulate says that synonymous wtth
thts
IS
(5) There is an acnon x [Otto does x CAUSE agam1 (cold water runs) ThiS statement entatls that the first runmng of water was a runrung of cold water, which was not so This shows that the mearung posrulate IS not tenable 1f we assume essennal propemes. At first sight, it IS a b1t puzzlmg that (2) enmls (4) 1he Situation IS similar for an entailment between Fritz IS human and Frrtz IS malt. If 1t IS essennal for Fritz to have the two propemes, the two statements are true in the same worlds. To state Zimmermann's argument some what d1fferendy: the mdiV!dual acnon e which satisfies (2) IS similar to Fritz-It causes the runrung of water, even of cold water Therefore, for dus pamcular e, (b) and (7) are synonymous: {6) Otto does e CAUSE water runs (7) Otto does e CAUSE cold water runs Dowty's postulate entails for this pamcular e that it IS the cause of the truth of the state ment. {8) agam1 cold water runs) Thus, Zimmermann's argument refutes Dowry's postulate mdeed. If we look care fully at the argument, we diSCover that the
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
IS also that of Dowry (1979� Dowty assumes decomposinon in the semannc language but not in the syntax. Scope amb1guines are accounted for by mearung postulates and by assuming two lexemes, again1 and agam2, where the first accounts for the 'outer' or 'repentive' readmg and the second for the 'inner' or 'resnrunve' readmg 3 1 Thts approach has been cnnazed m Stechow (1 995) for the followmg reasons Fmt, the method of assummg two different lexemes IS ad hoc. Second, there is no expbnanon of the fact that the surface pos1tion of agam resmcts the range of poss1ble readmgs. We have been concerned with such cases. Thtrd, Stechow (1 995) expresses doubts about the feas1b1hry of the mearung postulates mvolved. In hiS Anmmvorlesung {cf Zimmermann 1993) held m December 1 992 at the Uruver slry of Stuttgart, Ede Thomas Zunmermannn took up that cnncism and made It almost certam that mearung postulates of the sort adopted m Dowry (1 979) are not sound. To refute Dowry's postulate, Zimmermann cons1ders the restitunve read1ng of a sentence hke
1 32 The Different Readings of Wt�rr 'Agam' erroneous assumptton IS rhar we can discover those qual1ties of the result of an mdiVIdual event which are needed for descnbmg rhe 'sure resnrured' by invesnganng the event
A PPE N D I X
Irsel£ ThiS is nor possible. We have ro modify rhe 'urger sure mforrnanon' 1rself, i e� that property of srares which IS embedded under BECOME m our treatment.
T H E R E L A T I O N BETWEEN K R A T ZE R ' S fwgtt A N D D O W T Y ' S B E C O ME 2:
(I) catch :- A.xAe pruoner(xXf-r�e)} Let me comment first on Kratzer's functor J� It denotes a funcnon defined only for events or acnons which have a result. It assigns that result to the acnon. �,(e) IS what Klem (I994) calls the tawt sta� of an action. If P IS a predicate of srares, then P{f,..,..k)) always implies or rather pre supposes that P does not hold of an earlier stage of e. Suppose, we understand the funcnon JWZ<'. Then rule (I) correctly stares that the object of rhe catching IS a pnsoner only at the end of the catching. Recall that we analysed the verb 'to catch'
by a function of an event. Although derails are not ennrely clear to me, let us suppose that Kratzer's and our representation of the verb mean exactly the same. In other words, we assume that the followmg statement IS a necessary truth: (3) /...xA.e BECOME(e)(prisoner(x)) - /...xA.e prisoner(xXftars<,(e)) Recall, however, that the BECOME-functor has its own place m the synrax. Therefore It can have scope m the synrax. As far as I have understood Kratzer, no analogue synracnc decomposinon IS Intended for her verbs. The definiens m entry ( I ) may be regarded as a rwo-place funcnon, and that's It. It is no problem to formaliZe the repennve reading of 'Rand1 caught Bockhirsch agam' by means of Kratzer's verb meaning: (4) again[/...e(agent(R.andiXe) & catch(BockhirschXe)))(e) ThiS means that Randi earned our an acnon whose result was that Bockh1rsch was a prisoner, and there was such an acnon of Randi's before. A problem ames, however, when we want to express the resnrunve reading We can try to represent thiS by giving again(e) narrow scope With respect to agennvuy: (s) agent(RandiXe) & again[catch(Bockhirsch)](e)
(2) f..:J.e BECOME(e)(prisoner(x))
ThiS cannot be correct. If we evaluate (5) by means of the defirunons of catch and again, we obwn the following meaning:
In orher words, the result has to be described by means of a propositional functor, and not
(6) Comenr: agent(R.andiXe) & prisoner(BockhinchXfra(e))
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
In the arncle, I have been usmg Dowty's BECOME-functor. I have found lmle evidence for the syntactic representation of thiS morpheme. agam always seems to have narrow scope wnh respect to BECOME or scope over the VoiceP which conwns BECOME If we look at result verbs. In parncular, 1t was hard to argue for mter mediate readmgs, where again has scope berween agent and BECOME Thus, It would be ruce to get nd ofBECOME Kratzer (1994) has found a method of encodmg the result mformanon without the BECOME fu nctor. Here I would hke to explam the reasons why I didn't follow her m thiS respect. Kratzer (I 994) would presumably rep resent a result verb like 'to catch' as some thmg hke this:l2
Amim von Srechow 1 3 3 Presupposition:
3e•[MAX(prisoner(Bockhirsch))
(f...(e•)) & TIME(e•) < TIME(e)]))
ThiS means char Rand1 caught Bocklursch and char chere had �n a successful carchmg of rum before. We don't want rh1S reading. We rather wane che sraremenr: 'IUndi caught
Bockhirsch and h e had been a pnsoner before'. The conclusion I draw from chis reasorung IS that an analysiS of resulunve verbs m the sryle of (1) IS roo simple. We need something hke (2), which is essentially Dowry's (1 979) decomposition.
N OT E S
(I) (u)
Ali Baba Sesam Wieder OFFnete Ali Baba Sesam WIEder offnete
The firsr example exlubits verbal mess and strongly suggests a resnruove read ing. In the second example, we find rhe m:un srress on wi�tr and thmk of a repennve reading. The correlanon berween srress and mrerpretarion is nor wa.cerrighc, however. Consider rhe folloWing text
(m)
Jerzt FAHRT des Bus. Now moves che bus. b Jerzr bleibr er STEHEN. Qerzr bleibt er wieder STEHEN) Now Stands n snll. a,
Jerzt FAHRT er Wieder Oerzt fahrt er WIEDER) Now moves 1t again. d Jerzr ble!bt er wieder STEHEN. Qerzr bleibt er WIEDER srehen.) Now stands tr snll.
c.
In (a) che srress IS conrrasove, no doubt. The bus had been standing snll. The first vananr of (b) conrrasts che stopping With the dnVIng. The second vanant IS the resnrunve reading of wirdtr 'again', where the stopping IS presumably conrrasted With the dnVIng In (c), we can have con rrasnve srress. If so, we have the precedmg utterance/thought m mmd. If we lay the srress on WJ�tr, we have a repennve readtng and have the first utterance m mmd. Thus the conrrastive/resnrutive/ repennve d!Snncrion is not only a matter of the world. The interpreunon must be anaphonc m some sense. The present paper ignores the issue ennrely. Hope fully, no harm wtll anse from thiS lazmess. 3 See Dowry (1 979) and the references gtven chere. 4 Fabncius-Hansen (1 980) and { 1 983) 5 Fabricius-Hansen ( I 980) observes thar Enghsh agam , Danish rgm 'ag:un', Norwegian rgj(71 agam and German WJ�tr 'again' origtnally all meant 'agairut'. Smce the notion 'against' denotes a conrrast, it is plaustble that 'against' expressed a contrast in nme for counrerdtrecrional pairs hke 'to fall' and 'to rise'. Thus, 'again fall' could have '
',
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
1 Th e first verston of thiS amcle was wntten for the occaston of the Krerns conference on morphology Quly 1 992). I have changed some details, notably under the influence of Kratzer's (1 994) manu scnpt and a handout thereof from 1 993 The mam Ideas have remamed unaltered, however. I profited very much from diS cussing the matenal wnh Carhenne Fabricius-Hansen, S1gnd Beck, and Wilhelm Geuder The latter wrote through comments on the first draft, which com1derably Improved Its content. I also have to rhank rwo anonymous referees of rhe journal of &mantics for 1mpottant comments. 2 The rwo readmgs of { I - I a) may be diS ambiguated by meam of mtonanon. Constder the folloWing rwo our-of-the blue utterances:
1 34 The Different Readings of Witdn- 'Again'
6
7
9 10
II
12
1 3 The hiStory of the 'unaccusanve hypo thesis' IS rather complicated. See Pullum ( 1 99 1 ), Chapter 1 8. The first expliCit formul:mon of the hypothesiS seerriS to be contamed in a lerrer from Paul Postal to DaVJd Perlmurrer on 20 October 1 975. 1 4 A proper formulation of the meaning rule requues a theory of context dependency, of course. 1 5 I owe these examples to Manfred B1erwisch. 16 The facts are not as clear as one would like. The sa1d referee writes: 'Zummdest m�rut kann zweifelsfrei auch als Zustandsprad!ltat fungieren, vgl Das Fahrrad rst m��t kapu.tt.' Perhaps these adverbs are not purely repennve. There seems to be a scale from Wlrdn-, wh1ch has the broadest use, to adverbs which are very near to quannficanonal adverbs; noch einmal seerriS to belong to this other end of the scale Quantificanonal adverbs don't combme With stanves at all· (1) •Dte Tasse war noch eimnal! dreimal he1l the cup was once more/three nmes intact The mc)(Micanon of pred1canve par ttClpial phrases by adverbs of quantifica non IS not very good but certainly berter:
(n) die Tasse ISt dreimal geklebt the cup IS three rimes glued
ThiS
17 18
19 20
suggests that the mcxhficanon takes place under the PERF operator which rransforms the BECOME-verb mto a stanve. Kratzer ( 1 994: 39� Kratzer ( 1 99+ 70) interprets the mdex i of PR01 as A,, whereas PRO IS an NP wluch is invisible for the semantic componenL For the purposes of this paper, my slightly simpler account is suffiaenL C£ Kratzer (1 994: 69). Kratzer ( 1 994) tells a different story about the danve. For her, It is licensed by a VISible or mvisible preposmon. This preposition mcorporates into the verb,
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
8
meant somethmg hke 'agamst what we had before, we have a fallmg now'. As far as I can see, Borer's (1 988} para/It! morphology shares all essentul basic assumptions With Baker ( 1 985b� Ifl speak of a Bakerian approach, then that IS because I became acquainted With thiS way of theonzing through Baker's work. A careful hiStory of thought Will presum ably reveal many other sources, for msunce Generanve Semantics and early work in Generanve Grammar. In Chomsky ( 1 986} phrasal selecnon is called categorial selecnon (c-selecnon). I don't rake over thu tenrunology because I consider such a way of speaking IS miS leading-a category is selected both m the morphology and m the syntax. The relevant difference is the complex�ry of the 'bar-level'. This approach considers the mvnmal/m1rumal diStincnon to be pnminve. ThiS IS not compatible With Chomsky's ( 1 994) 'bare phrase structure', where rhe diStincnon IS relanonally reconstructed G1ven the simple extensional ontology used m this amcle, It IS not possible to spell out the mearung of rhe CAUSE predicate, for it IS a relanon berween sets of worlds. CAUSE(p, q) holds in world w Iff p and q are true m w and the counter factual 'if p were the case, then q would be the case' IS rrue m w. ThiS IS LeWis's (1 973) analysis, of course. In Chomsky's ( 1 992) terms, the move ment must be before spelling out. See Diesing ( 1 990'1). (1 990b), and (1992). 01es1ng (1995) suggesrs that rhe reason IS that pronouns are free vanables winch would be bound by her exiStenrul closure of the VP. Therefore they could no longer be used anaphoncally any more. If this were the reason, they should be bound by a genenc operator or an adverb of quanti fication in a scrambled posinon because scrambled mdefirute terms are always bound by such operators. Kratzer (1 994) argues that the subjects of APs may be PRO.
Amim von Stechow I 3 S but verbs wluch have mcor:ponted empty
26 The maximahty condition
IS
redundant
preposinons cannot incorponte further.
for tehc transinons because properties
ThiS expl:uns the contrast
expressmg them do not have the sub
[1) •0as Manusknpt ist ihr geschickt {Kntzer I 99+ so)
Interval propety. Hence the
m:oomality
condition is mvially fulfilled: 27 Stechow ( I 984b) recoTI5ttUcts 'positive'
the manuscnpt is herda, sent
degrees like tallness as mtervals [o, m],
(u) das Ihr gesch1ckte Manuskript
where o marks the begmrung of the
the herda, sent manuscnpt The mvJSible preposmon licensmg the
relevant scale and m is a number marlung
the distance from o With respect to some
dative is mcorpomed mto V. In (1), V has
appropnate measure. The degree d is
to mcorporate further mto
higher than d•,
IS
A Tlus
blocked. According to Kratzer, attnbunve no further mcorporanon is necessary m (1i). With open preposmons, the danve
IS
poss1ble m pred1cative parnCipial phrases
Without preposmon (examples (i) and
(u)
hcensed by [Spec, Agrlnd.O], wluch
requrres the presence of VoiceP.
This
explains the contrast (1Y(u). The dative m
(ill) is licensed exactly as Kratzer proposes.
;rhus there are two mecharusms for danve
co],
where
the
left open. [m, co]
IS
nght
h�gher
than [n, co] 1f m < n, because only then 15 IS
reconstrUcted as the diStance
diStance m feet from
2 I The analysiS requires that the descnpnon
'her husband' may refer to Otto, even if .presently not Anna's husband any
oo
to 6. We don't
know what tlus should be But '6 feet
smaller' makes sense because the dif ference between rwo degrees of smallness IS
defined. For [m, co] and [n, co] thl5 ·IS
1m - nJ; 1.e. the absolute value of the dif ference berween m and n.
28 The
hcensmg.
more
[m,
IS
sense hterally because It would be the
have to say that the srrucrural danve
IS
mtervals
boundary
from o to 6 m feet. '6 feet small' makes no
rhe ball was hi�, to-thrown
Otto
for [o, m] and [o, n] if
m > n. 'Negative' degrees are regarded as
feet tall'
(1ii) Der Ball war 1hm zugeworfen
IS
Iff d• IS a proper subset of
case
the second a proper subset of the first. '6
as well:
I
is the
same
holds
for
the
followmg
examples w1th stressed wit!tkr, which have
been diSCussed by FabnciUs-Haruen on
several occasions:
The only pngmanc constr:unt
seems to be that
Anna d1d not r�marry
{I)
a.
after the divorce. Thus 'her husband'
the captam ag:un sober·JS
means 'Anna:s husband at orne t', where t 15 contextually determmed. Such uses of terms are known under the
name
b. WIEDER 15t der Kapitan niich tem
amp/ra
tio in the tr:�dition.
der Kapitan WIEDER niichtem 1St
c.
Der Kapicin JSt Wieder NOchtem
22 Fabricius-Hausen (I 98o) and ( I 98 3� 23 See note S·
24 Tiiljmgen,July I 994 (- Fabricius-Hansen I 99�); Blwbeuren, October I 994 Fabricius-Hansen I 994b�
(
a sort of a JOke, which points
(b) or (c)
IS
out that
we
again have the remarkable
fact that the capt:un is sober. These rea.dmgs do nor presuppose that the
baro
captain was ever drunk, though rhey sug
'last' riSing before the falling and the
semantically berween - (a} and (b) on rhe
writing this into the meming rule.
think we interpret (b) and (c) silently
2 5 CoTI5ider sentence (9-u�
if the
meter fell after a ruing. then there was a barometer fell
after that. I see no pomt in
gest that. I
see no way to d15nnguJSh
one lund and (c) on the other lund. I so
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
participial phrases are verbal; therefore,
d. Tim
1 36 The Different Re4dings of Witdtr 'Again' dur they express genume 'repennons'
A
Appropnate paraphrases may be:
advocated here IS found m Abusch's
(2) Wieder sehen wtr den Kapitan niich rem
'Again we have a day when the
almost(p) means something like 'p IS rrue discussiOn. MPI I · 'Vx 'VP 'Vp againz( "CAUSE( "P(x), BECOME(p)) - CAUSE( " P(x),
As desired, these sentences do nor pre
BECOME( ·agam1(p)))
suppose that the capram had been drunk.
29 As far as I know, thiS has nor yet been ( I 979: 88) gtves some hmts for the rrear
3 2 C£ Kratzer ( I 99+ Chapter 2, p. 3 5)
t
Sun Peters asked me why the lower vo1ce phrase does nor undergo pass!VIta
'degree
non. The (sensmve) answer is that the
achievements', but he IS not sure whar the
passive auxi11anes express the relation
oor way of dealing wtth rhem would be.
agmt
RE FE RE N CE S Abusch, Dom (1 985), 'On verbs and nme', Ph D. diSserunon, Uruvemty of Massa chmens, Amherst.
Baker, Marc C. ( 1 98 5 a), 'The mmor pnnC!ple and morphosyntacnc explananon', Ling uJStJc lnquiry, 1 6, 3 7 1 -4 I 6.
Borer, Hagu (forthcom mg), 'The morphol ogy-syntax mterface. a study of auton omy', forthcoming
m
the Proceedings of
the Krems Conference on Morphology, July 1 992. Burz10, LmgJ (I 98 1 ), ' Intrarumve verbs and
Baker, Mark C. ( I 98 5b), 'Incorporanon: a theory of grammancal fu ncnon chang
Irahan
auXIhanes',
Cambndge,
Ph.D.
MA, MIT.
diSsertanon,
MA,
Carner, Jill & Randali, Janet H ( I 992), 'The
Baker, Mark C. ( 1 988), Incorporation. A Throry
of resultarives', Linguistic lnqutry, :tJ, I 73-
mg', Ph.D. d1sserranon. Cambndge, MIT
cf Grammatical Functton Changing, Umver
Slty
of Chicago
Press,
Chicago
and
London.
argument srrucrure and synucnc structure
2 3 5· Chomsky, Noam (I 98 I), LuturtS on
GoVffll
mmt and B1ndmg, Foris, Dordrecht.
Besten, Hans den ( 1 98 5), 'The erganve hypo thesiS and free word order m Dutch and
Chomsky, Noam ( I 986), Know/ttl� of Lan gua�, Praeger, New York.
German', m Jmdrich Thoman (ed.), StuditS
Chomsky, Noam ( I 989), 'Some notes on
tn Coman Grammar, Fons, Dordrechr 2 3-
economy of denvarion and represenu
64-
non',
BielWISCh, Manfred ( 1 96 3), 'Grammarik des deutschen
Verbs',
Studia
Grammatico ,
Akadem1everlag, Berlm between
compounds
and
consrrucr, m Gee rt BooiJ & Jaap van Marie
(eds),
Ytarbook of Morphology
Dordrecht.
Chomsky, program
Borer, Hagtt ( 1 988), 'On the morphol<>g�cal panllelism
MIT Workmg Papns m LmguJStiCs,
43-74-
1,
Foru,
Noam
( I 992),
for
hnguiSnc
'A
minimalist
theory',
Occasional Papns in LmguJStics bndge,
MA, MIT.
10,
1
MIT
, Cam
Chomsky, Noam ( I 994), 'Bare phrase struc
ture', MIT OccaSJonal Papns in Linguistics 5 ,
Cambridge,
MA, MIT.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
done in detail tn the literature. Dowty calls
solunon
3 I Cf. Dowty (I 979· 267):
capram IS sober'
he
the
in the nearest world (bur not m the actual
der Kapuin niichrem ist.
what
to
one)'. See Egg ( I 993) for some recent
Wieder haben wtr emen Tag. an dem
of
pomnng
dlSSerunon. See Abusch ( 1 9 8 5 : 1 30). 30
'We see the capum sober agam'
menr
remark
Amim von Stechow I 37
Pluralttrmm und kptktklassm , Fmk. Munich. Lasnik, Howard (I 98 I ), 'Resrricring the theory of rraruformanons. a case study', m Norbert Hort!Stein & Dav1d L1ghtfoot (eds), £"planation in Lmguzstrcs Tlrt Log�eal Probltm ofLangua� AcquiSitron, Longman, LondoiL Lewis, DaVId (I 973), 'Causanon', journal of Phrlosophy, 70, S S -6 7 LeWIS, DaVId ( 1 979), 'Amtudes de d1cto and de se', Phrlosophrcal R=nv, 88, I 3-43. Partee, B. (1 973), 'Some srructural analog1t'S berween tenses and pronouns m EngliSh', journal ofPirilosoplry, 70, 6o 1 --9 Perlmurrer, Dav1d M (1 978), 'Impersonal pass1ves and the unaccusanve hypothesiS', Proutdings oftlrt Fourth Annual Mtttrng oftlrt &rkdty LinguiStiCS Socrtty, Berkeley, CA, I S7-89. Pollock, J. Y. (1 989), 'Verb movement, UG and the srructure of IP', LrnguiStiC Inqurry, zo, 365-424 Pullum, Geoffrey K. ( I 99 1 ), Tlrt Grrut Eskrmo Vocabulary Hoax and Otlrtr lrrtvtrtnt &soys on tlrt Study of Languagt, Umvemty of ChiCago Press, Ch1cago and London. Rapp, Irene ( 1 995), 'Parriz1p1en und seman nsche Srruktur: Zu pass1v1SChen Kon srruknonen mit dem 3 Status', dlSSCrtanon, U ruvemcit TiibmgeiL Seuren, P. A M ( I 973), 'The comparanve', m F. K1efer & N. Ruwet (eds), Gmtratrw Grammar in Euro�, Re1del, Dordrecht, 528-64Stechow, A von ( 1 984a), 'Companng seman nc theones of companson', journal of Stmantics, J, 1 -77. . Stechow, A von (I 984b), 'My reacnon to Cresswell's, Hellan's, Hoeksema's and Scuren's comments', journal of&mantics, J, I 8 3--99· Stechow, A von ( 1 995), 'Lex1cal decompos1non in synrax', m U. Egh. P. E Pause, Ch. Schwarze, A von Stechow & G. Wienold (eds), Tlrt Lo.1con rn tfrt Organization of Languagt, John Benjamtns Publishmg Company, Amsterdam and Ph1ladelphia, 8 1 - 1 1 8.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Dies1ng, Molly ( I 990a), The syntacnc roots of sem:�.nric parrinon', Ph.D. thesis, Amherst, Uruvemty of MassachusettS. Diesing, Molly ( I 990b), 'Verb movement and the subj ect posmon in Yiddish', Natural Langua� and Linguistic 11r�ry. 8, 4I-79· Dlesmg, Molly { I 992), 'Bare plural subjects and the denvarion of logtcal representa nons', LmguiStic Inqurry, l.J, 3 5 3-80. Dlestng, Molly ( I 99 S). 'Ytddtsh VP order and the typology of objeCt movement m GeiTIUruc', unpubhshed MS, forthcommg m Natural Langua� and LmguzstiC Tlr�ry Dowty, DaVId R. { I 979). Word Mtanrng and Montagut Grammar, Retdel, Dordrecht. Egg. �rkus { I 993). 'Akrionsart und Kom posmonahcit; Zur kompos1nonellen Ablemmg der Aknonsart komplexer Kategonen', d!Ssertanon, Umversltii.t Konstanz. Fabncms-Hansen, Cathnne (I 980), 'Lex1cal IISChe Dekomposmon, Bedeutungspostu late und witdu', m D. Kastovsky (ed.), Pm�ktr� dtr ltx�ka/JS(htn &mantrk, BouVIer, Bonn, 26-4 1 . Fabncms-Hansen, Cathnne (I983), 'Wteder em wrtdtr? Zur Semannk von witdtr', m R. Bauerle, C. Schwarze & A von Stcchow (eds), Mtanmg, u� and lnmprttation of Langua�, de Gruyter, Berlm and New York. ( I 994a), Fabncms-Hansen, Cathnne 'W1eder', handout, Umverstcit Tiibmgen, 6July I 994( I 994b), Fabricms-Hansen, Cathnne 'Wteder', handout, Blaubeuren, October I 994Galhn, D. (I 975), Intmsronal and Hrglrtr-Ordtr Modal Logic, North-Holland, Amsterdam. Jacobs, J. (I 98o), 'Loncal decompos1rion m Montague Grammar', Th�rttical LinguiS tics, 1, I 2 I -36. Klein, W. ( I 994), Timt rn Langua�, Roudedge, London and New York. Kratzer, Angelika ( I 994), 'The event argu ment and the semantics of voiCe', MS, Amherst, Univemty of Massachuserts. Krifh. M (I 989), Nommalrifmnz und ZLit konstitutton. Zur ZLmantik r,oon Massmttrmtn,
1 38
The
Stechow,
Different Readmgs of Wttda 'Again' A.
von and Stemefeld, W. ( 1 9l!l!), syntaktischrn Wi.ssnu , West deurscher Verlag, Opladen. Sternefeld, Wolfgang ( t 9l! s), 'On case and bmdmg theory', m Jindrich Thoman (ed.), Studttl tn Gmnan Grammar, Dordrechr, R.etdel, 2 3 t -8 5· Stemefeld, Wolfgang (t 99S). 'Votce phrases and rhetr specifiers', Semmar fUr Sprach WISsenschaft, S£5-Report-<>S-95. Untverst cir Tiibmgen. Baustrinr
Stowell, Ttmothy ( 1 98 1 ), 'Origms of phrase structure', Ph D. diSSertation, Cambridge, MA, MIT. Stowell, Timothy (1993), 'Synux of tense', MS, UCLA. Ztmmermann, Ede T. (1993), 'Zu Ristken und Nebenwrrkungen von Bedeurungs posrulaten', Ltnguistiscltt' Bmclzlt', 146, 26 3-
82
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Journol of&mantUJ
1 3: 1 39- 1 Ko
C Oxford Umvc.-rnty Press 11)<)6
The Definite Article: Code and Context FRANCIS R E NA U D CRLAO!EHESS, Paris
Abstract Numerous grammaocal Jtems can only be understood 1f the concexr JS taken mro account. I n
order t o proVIde a formal treatment of chiS problem, w e mttoduce the n onon o f know/rtf� h<m
defirure arncle /(. Tlus grammancal Jtem
has the parnculanry of not encodmg a fixed
mformanon stored m the lexicon, but of runrung a program of rnformatron proussr ng We show
char the referennal computanon can
be
made by a pre= algonchm workmg on a defirure
stratum of the reader's knowledge base. In
rlus framework, the uruqueness hypotheSIS of
Russell becomes a gmdelme to remeve memonzed mformatioiL The algonrhm accounts for de1ctic
as
well
as
anaphonc referenna! Jdennficanons. Our analysiS proposes a
unrtary account
for a large parr ofempmcal data and reconales the log:Jcal approach ofRussell-Moncague WJth the pragrnanc approach of Hawlans.
o I NTRODUCTION What IS the difference between grammancal items and other words of the language? Is the semantic value of the article the same nature as that of the noun
the m the noun phrase the table of
table?
To answer this question. we shall not take the point of view oflmgmsts who consider that the essential job of linguistics is only to build
logical
representations of sentences. This is because we believe that the meaning of an expression cannot only be in its logical representation. It seems more likely to be the product of a dynamicprocessing ofinformation. This is why we attribute two
()
complementary tasks to semantics: ( 1 ) building semantic representations of a logical nature; 2 handlmg these representations in specific cogninve contexts. In the recognition stage, the parser builds the semantic representations with the support ofthe syntactical structures. This relation between syntax and logic was first established on formal grounds by Montague, and remains an ideal for us to reach in semanncs, as the beaunful boo k ofKamp & Reyle ( 1 993) testifies.
But to narrow semantics to this construction stage has always seemed
insufficient to us. Why should we construct formulas if we then have no use for them? We would like to avoid producing formulas without any operational
force. To be effective, these representations have to be brought into play to
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
mto lmguJSncs. After haVIng worked our a hard core system of rules for analysmg numerals,
defirure arncles, and mdefinne articles, we g1ve a thorough srudy of the French smgular
1 40 The Definite Article: Code and Context account for inferences and dialogues. Moreover, they must be able to be subjected to empirical tests. All the empirical data we can gather on any linguistic problem (such as time, quantification, etc.) do not determine in a uruque fashion what the best representations are. Semantic theory is undetermined by observation. To overcome these Insufficiencies, we thought it necessary to borrow the notion of a knowledge base from artificial intelligence. It is a memory where not only the representations of the sentences analysed are recorded but also extralinguistic knowledge (whether accepted or tacitly possessed by the hearer at speech time). The model also contains a reasonable module which makes it possible to draw inferences and compute answers to quesnons by using the clearly between the semantic representation of an expression and its
meaning. We understand the mearung of an utterance as the whole set ofinferences one can obtain after entering its semantic representation into the knowledge base. Once an implementation of this model has been devised, it becomes possible to srmulate dialogues. We enter a small text into the computer and we ask questions: who does what? when? etc. Then, since to understand a text is to be able to answer questions on that text, it becomes possible to
hypotheses empirically:
test semantic
the best hypothesis is the one that gives the best
simulation of a dtalogue. We used this model to give linguistic analyses of spatia-temporal expressions in Chinese (Renaud
1 988 )
and in French (Renaud, forthcoming).
And when we began to study nominal determiners in French some years ago, we discovered that here, too, the notion of a knowledge base has a leading part to play in their semantic interpretation.
It is a truism m linguistics to consider that meaning depends on context.
There are countless examples of starred (inacceptable) sentences which, when
(a la
provided with an adequate context, become perfectly acceptable. This banal observation is wholly inexplicable in the traditional semantic frameworks
Montague or Kamp), since the computation of semannc representations consists in retrieving information stored m the lexicon and combining them
according to the syntactic structures.
A detailed study of the French definite article convincd us that they involve process contextual information. They trigger off a search in the
instrucnons to
knowledge base, the purpose of which is to identify a referent. If the procedure succeeds and returns a constant, definite articles acquire a logical representation m
the spirit of Russell. Thus we have the means of reconciling the logical
approach of Russell and the pragmatic approach of Hawkins(1977). This article' consists of three parts: first, we introduce our linguistic model; second, we give a sketch of a system of rules to parse some of the French determiners; third, we give a thorough study of the definite article
/e.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
techniques of resolution by unification. We thus have the means to distinguish
Francis Rcru�ud 1 4 1
1 FOR AN O P E RA T I O N A L S E M A N T I C S
auto. reasoning
Answer
When the computer/listener is gtven an assertive sentence, he parses it and bmlds a semannc representation wluch 1S stored m the knowledge base. Thus we can enter a small text and then ask questions on it. The representations of the quesnons are sent to the automatic reasoning module, which computes answers. 1.1
The grammatical module
This module is based on a linguistic theory that we can only sketch out here. The syntactic analysis is made by an unification grammar (Smolka 1992; Car penter 1 992). Within the limited needs of this article, we will often content ourselves with rewriting rules of context-free grammar. Each syntactic structure is associated with a rule of semantic construction. We have chosen to use a very powerful tool to build semantic representations: the lambda-calculus. It has the advantage over the DRS-construction rules . (Kamp & Reyle 1 993) of allowing an integrated analysis of syntax and
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
We are trying to model how speakers understand sentences-how they encode the information into semantic representations and how they process this information. We are only dealmg with recognition. For the sake of concreteness, we will present our model in the shape of irs computer implementation. It is composed of four modules: ( 1) a grammar module, consisting of rules for the construction of syntacnc and semannc representations; (2) a knowledge base, KB, where the semannc representations of the sentences are stored; (3) a parser, which employs the grammatical knowledge to build a semannc representation SR of the sentence; and (4) an automatic reasoning module, which computes the answers to questions.
142 The Definite Arocle: Code and Context
semantics. A simultaneous syntactic and semantic analysis is psychologically more plausible than a sequential one.2 Let us suppose we want to represent ( 1 ) Mary came by comt!(m), where m is a logical constant representing Mary. If we take A.P.P(m) (all the properties of m) as the SR(semantic representation) of the noun phrase N2, the computation will be made by
{
S - N2 V2 So :- S1.S2
So :-
Sr.S2 -(AP.P(m)).come
by which �reduction gtves come{m). Let us also take a rule of construction of N2 from a proper name NM
{ N2 - 1-P.P NM (S r ) So :-
where So and Sr are the semantic representations of N2 and NM. Wah these two rules, we can sketch a denvation of Mary came S,
N2, .t
I
come(m)
--- -...__
P.P(m)
V2,
come came
NM, m Mary 1 .2
The parser
For each sentence, the parser uses the knowledge of the grammatical module to build a formal representation. The grammatical module, as previously presented, could lead us to believe that it works in a compositional manner. But that is not entirely the case. The reason for this is that the management of spatia-temporal reference requires us to process information that has been previously recorded by memory. When we analyse the tenses of the French sentences: (2) Max enrra. II vit Anne. [Max came in. He saw Anne]
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
where So, S r , and S2 are the semantic representations of the sentence S, the noun phrase N2 and the verb phrase V2 respectively. The dot between S r and S2 represents the application according to lambda-calculus. The rule tells us that if the V2 came has the representation S2 - come and the N2 Mary has the representation Sr - A.P.P(m), then the sentence S will have the representation
Francis Renaud 1 43
I .3
The knowledge base
As soon as the sentences of a text have been analysed by the parser, their semantic representations are recorded 1n the knowledge base (KB). The knowledge base permanently contains the necessary information to give an 'intensional depth' to the predicates introduced from the lexicon (definition of properties of transitive relanons, links between process and state, etc.) and some general or episodic knowlege of the world.
We shall define the Knowledge Base KB as the set ofinformation (knowledge, beliefs, assumptions, t'tc.) accepted or tacitly possessed by the hearer at the speech time. As we have seen, the SR aSSOCiated with a sentence contains a lot of pragmatic information linked to the utterance situation: knowledge about the speaker and the addressee, of the speech rime, of the referent point, etc. Never theless, we do not consider that the whole meaning of the utterance is contained in its SR Numerous contextual effects can only be obtained by plugging this SR into the K.B. The predicates appearing in the SR obtain their meaning only if they are linked to other predicates of the language by a rich network of relanonships and if they are Instantiated in general knowledge schemas. It is more or less deep activation of the relations between the SR and the KB that defines the depth of the understanding. In a verbal interaction, not everything is said. To understand is to be able to restore all the missing information in order to build a coherent and complete representation of a 'world model' (Enjalbert & Victorri 1 994). TM mt"aning ofan uttaance is always diftned relative to a knowledge base. It is not frozen in a representation but is obtained by a dynamic processing of the whole information contained in the SR of the utterance and in relevant parts of the
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
we must be able to represent the succession of the events came in and saw Annt'. To do that, at the same time as we parse the event e1 vitAnne we must be able to search in the .memory for the previous event entrer(eo, max), indexed by eo, in order to note the succession end(eo) < beg(e1 ) (end of eo before beginning of e 1). In this same example, the analysis of the pronoun il (he) also supposes a processing of contextual information. In previous work (Renaud, forthcoming), we dealt with the management of the temporal reference point. Here we will consider the nominal reference of the French definite article le in its specific, countable uses. We will see that it cannot be represented by a predefined semantic representation rt>corded in the lexicon but by a program ofprocessing the contextual information . The running of this program will possibly return an identifying referent which will then be used to bmld the semantic representation SR of the noun phrase. Thus the SR contains some pragmatic mformation.
144 The Defirute Arncle: Code and Context
KB. This
dependency on the parts of the whole produces a holistic effect,
3
making semannc analysis particularly difficult.
1 .4
The reasoning module
This module is a mechanical theorem-proving system used to compute the answers to questions. Whenever the parser receives a sentence in the
interrogative form, it first computes its SR according to the rules of grammar and then dispatches the SR to the reasoning module. The answer is then computed by resolunon, a theorem-proving techruque using unification (as in Prolog). Finally, an answer generator translates the logical answer into a natural To make resolution proofs, it is necessary to ehminate existential quantifiers. Consider, for example, the followmg (3) Paul a epouse
une Chinoise. Elle est jeune et Johe.
[Paul mamed a Chinese woman. She 15 young and pretty]
If the indefirute article introduces an eXIstential quantifier, as is tradmonally assumed, the first sentence Will
be represented by:
(sq) 3e3x(epouser(e, paul, x) 1\ chinoise(x))
KB, each existential quantifier triggers offthe creation ofa
e IS an mdex which specifies the uniqueness character of the event At the nme of storing the SR into the
new constant'' (up nil
now not used). With e36, c37, these new skolem constants,
the clauses
{
epouser(e36• paul, C37) chi noise(c37)
will be added to the
KB. When the second sentence is parsed, the analysis of the
pronoun elle (she) will run a procedure of anaphor resolution. In the rule which introduces pronoun PRO
{
N2 - PRO
So
:-
A.P.P(srch_ant(U I))
the algorithm searching for antecedents
srch_a nt
uses the features of gender
and number (transmitted by the feature structure U 1 of PRO) to extract a possible class of candidates from a particular register refn--disc where the . constants and the heads of noun phrases are stored. In the present case, it will
KB, it will return the skolem constant
search for a singular feminine noun and will find a possible candidate in rifer-disc. Then by unification in the
c37. Thus the second sentence will be represented by
(we do not coruider the time).
Chinoise
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
language answer.
funos Renaud 1 4 S Without skolemization, we would have come up against the serious prob
lem of performing modifications on a previously built SR: if the representation sr3 had been put into KB, we would have to put the SR of the second sentence age(x, Jeu�) /1. aspect(x, joli) (with a free variable!) inside the scope of the
not have to modify structures built previously; we only have to add new information into the KB.
existential qrumnfier 3x of sr3. Thanks to skolemization, we do
Knowledge bases for non-lingmsnc knowledge and theorem proving are
still not regarded as centrally relevant to the study of natural language by many pure linguists. However, these traditional tools of artificial intelligence are essential if we want to srudy functional words such as the definite arncle, whose purpose is to process context.
representations. For us,
observed inferences.
SRs are not ends in themselves but means
to
account for
This is why this model has been designed to account for
dialogues. It gives us a precious means of testing semantic hypotheses: the best hypothesis is the one that simulates dtalogues better.
operational semantics has also been defended by many 1 980, for example) and AI computer closest to ours is that of Enplbert 1 989; Enjalbert &
This conception of an
psychologtsts Gohnson-Laird & Garnham sciennsts (the theory Victorri 1 994).
2 T H E H A R D C O RE O F N P R U L E S I t is qmte obvious that the introduction of formal representanons allows a full clatificanon of hypotheses and makes discussions easier. But, for us, the importance of formalization also hes in the fact that it makes it possible to build step by step a complex formal system integrating a large number of language properties.
Tlus IS the reason why, before launching into a thorough srudy of the
definite article, we begin by sketching a system of representations for most of the determiners. We plan to deVIse a hard core of rules which should not need reshuffling even when faced with the slightest difficulty. It would have been possible to expound the formal properties of the singular definite article with the help ofpredicates with
individual arguments. But such a
representational scheme could not be extended to account for the properties of
plural determiners. The system which we are about to present is backed up by a sufficient number of properties to be able to stand up to various extensions. It is
based on set type arguments. Although necessary, this typmg is not sufficient to
account for some properties. To prevent an oversimplified opposition of collective and distributive interpretations, we shall also introduce the notion of
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Contrary to HPSG (Pollard & Sag 1 994) and to most contemporary linguistic theories, we do not beheve that the ultimate goal of semantics is to bmld formal
1 46 The Definite Amcle: Code and Context cover. Unfortunately, it will not be possible within the scope of this article to give appropriate justifications for the notion in question. First, let us see why we need set type arguments. Our problem is with the representation of verbs with collective subject, such as se reunir (to meet):
(4) Tous les eleves se sont reunis. (All the students met] {s) *Max s'est reuni. (Max met) An SR of (4) with individual type can be
Vx(stud(x) :::> meet(x))
(6) (Taisez-vous), les eleves travaillent [(Be quiet,) the students are wodung] This sentence is true in each of the following situations: - etther each student works by himself, - or all the students work together, - or in any other intermediate sttuation, as when the students work m small groups. Rather than proposing three representations, we prefer to introduce the notion of cover that enables us to give only one representation, true in all possible simations. A cover of the set of students elm is a family of sets !ii - (C1, C2, Co) whose unron covers the whole set elm: •
•
•
eleve - c l u c2 u . . . u CD
Thus, sentence (6) will be represented by
3 .9f' cover( .st' , eleve) 1\ VX(X E .9f'
:::> travailler(X))
The distributive reading is obtained by taking a family of sets with one element whereas the collective reading is obtained by a family of one set of all students n-
1,
cl -
deve,
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
But from this formula and 'Max is a student' stud(m), we can deduce the formula meet(m) (which will be the SR of the agrammancal sentence ( s)). Consequently, we must introduce a set type argument meet(stud) which means that the meeting involves the whole set of students stud - J..x .stud(x) (restricted to a given group in situation). We choose set typing instead of the senulamce typing of Link (198 3). The latter sort of typmg seems to create complicanons without any explanatory advantages.5 Let us now consider the following example:
FranCIS Remud
1 47
This notation makes it possible to give only one representation for both the distributive and the collective readings as well as the intermediate readings, generally neglected by linguists. For a cover 5I - {C1, C2, C D) of the set C, we will note the union of its elements •
•
•
u. 5I - c l u Cz u . . . u CD - c
(7)
Les eleves ant deplace trois armoires. [The students moved three cupboards)
Tlus sentence can be true m numerous sttuations. But is 1t necessary to translate this multiplicity of values in a multiplicity of representations? Shall we, like Kempson & Cormak (I 98 1 ), give four different representations of Two examiners marked six scripts or, hke Davies ( I989), give no less than etght representations? Generally, disambiguating such sentences without context causes a lot of trouble for speakers. It seems psychologically implausible that they should first methodically bmld four or eight representations, then proceed to eliminate the representations which are incompatible with the context. This is why we choose here to take only one formula to cover the whole set of possihilities. To account for the scope problem in (7), instead of representing the plural the students by J..P.(P.student), the set of propemes of students, we prefer to take the set of properties of a cover of srudents: ).9' .351 (student - u.!ll 1\ 9'.51 ). For the singular relational noun phrase 'Ia fille de Sophie' {Sophie's daughter) we must take a singleton .91 - Q�Q.x) (for {{x))) as a cover A9'.31�fille.(Q.s).y) 1\ 3x{fille.(Q.s).x 1\ 9'.(Q .(Q.x)))
For more details, see the rules in the Appendtx. 2. 1
Type theo ry notation
Type theory is a generalization of predicate logic which makes it possible to introduce arguments of any type One thus obtains a language well suited to dealing with quantification and plurality in natural languages. We do not give here even a cursory introduction to type theory but refer the reader to the excellent boo k of Andrews (I 986). .
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
We take an individual type argument for a noun such as 'eleve' and set type arguments for a verbal predicate. A subset of 'eleve' will be noted such as: X s;;; eleve whose meaning is Vx(X (x) :::> eleve(x)). For a verbal predicate p and a set of individuals X, the formula p{X) Will mean that the whole set X takes part in the event p at a definite place and a definite nme. Now we add a few words on the problem of quannfication. Consider cases like the following:
148 The Defirute Arncle: Code and Context
Type theory is, as we have said, a generalization of predicate logic based on typed lambda-calculus. The two baste operations are application and abstraction. The application of a function for a predtcate p to an argument x Will be noted f.x and p.x. instead of the f\x) and p{x) of logic. If x is of type a (noted x:a), and f of a type a -+ � (noted fa -+ �). then fx is of type �- The abstraction is used to construct functions from terms. By the fl-reduction pnnciple, we shall consider that the set of girls is represented byfllle as well as by J..x(fille.x). If one takes ind to stand for the type of individuals and prop to stand for the type of (Boolean) propositions, then J..x(fille.x) is of type ind -+ prop.
Any term oftype a -+ prop will be interpreted as a set ofelements oftype a, and any of type a1 -+ (a2 -+ prop) will be interpreted as a binary relation over elements of types a1 and a2. term
a - x will be represented by Q.a.x and A.x(x - a) by A.x(Q.a.x), t.e. Q.a by fl-re duction.
Consequently (Q.a) corrtS]JOtUls to the singleton {a}.
If we now want a predicate with set type arguments, as in se-reunir.eleve, we shall have to take the following rypes: eleve:(md - prop)
se-reunir:((ind - prop) - prop).
and with a two-argument predtcate, we shall have aimer.X. Y where X and Y are of type ind - prop. We shall adopt che convention of association to the left of parentheses; for instance: atmerX.Y - ((aimerX).Y). Let us also see how to represent the mtersect:lon and the union of two sets. Let us constder the set of guls and the set of students. The1r intersection is composed by each x which is both girl and student J..x(girl.x 1\ student.x). Their union is composed by each x which is etcher a girl or a student J..x(girl.x V student.x). These terms Will be noted directly by girl n student and girlv student. So, we can adopt the following definitions: for X and Y of type a - prop, we define the new sets of the same type: X n Y -def /..x(X.x 1\ Y.x)
X V Y -def /..x(X.x V Y.x)
(with infixed operators as n and v, we shall omit the dot wruch notates application, so that X n Y must be understood as nX Y). One can also define the set inclusion by
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
To represent the singleton {a}, a set with only one element, one has to take the term A.x(x - a). With Andrews, we shall note equality With the prefixed symbol Q:
-der'v'x(X.x ::::> Y.x)
Francis Renaud
1 49
and the union of a collection !R of sets X�Y
-def Ax3X(!R X 1\ X.x) with the rype x:ind, X:ind .... prop, !R :(ind .... prop) .... prop. u./R
The set made up of two elements j and s is the union of two singletons U. s) UJ u {s), so, it Will be represented by ( Qj) u ( Q.s) written also as an abbreVIation such as Qj u Q.s . We can verify this; if we apply the definition (Qj u Q.s) - i..x(Qj.x V Q.s.x) - i..xG - x V s - x)
3/R (eleve - u./R 1\ VX(!R X ::::> travatllerX))
Let us also remember the meaning of the quantifier 31: 3,x(p.x) means: there is one and only one x such as p.x, i.e. 3x(p.x 1\ Vy(p.y ::::> y - x)). 3 T H E A R T I CLE L E A N D T H E U N I Q UE NE S S CONDITION I am going to study the specrfic uses of le N, with a singular count noun. I shall leave the plural rill later. In this classic field, which has already given nse to numerous studies, one can only be mdebted to all the authors who come before one-I'm thinking of all the people presented in the bibliography, especially Kleiber ( 1 98 1-1993 ) and Hawkins ( I 977). The survey of these studies has already been excellently carried out (Corblin 1 987; Heim 1 988; Galrniche 1 989; Kleiber 1 98 1 -1 983 ). Rather than adding one more survey, I prefer to rely on emptrical data and mathematical tools to develop the 'internal logic' of my analysiS. I shall attempt to present a comprehensive formal analysis with syntax and semantics explicitly given. Although the informal analyses of traditional linguists are very valuable, It is impossible to assess them theoretically without translating them into formal systems. This task has never been completed and I cannot do it for the sake of others. As for the work of Heim ( 1 988) (which we discovered after having
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
we find that x belongs to Qju Q.s if and only if x -j or x - s. Hence, jean and Sophie cohabitent Oohn and Sophie live together)' will be represented as: cohabiter. ( Qj u Q.s), but 'Les eleves ttavaillent (The students are working)' will be represented as
r 50
The Defirute Arncle: Code and Context
completed this study), we are pleased to note that her Novelty-Familiarity Principle has great similarities with our own rule of construal. When she says 'for every indefinite, start a new card; for every definite update a suitable old card', we would say 'for every indefinite, add a new skolem constant; for every definite retrieve old information'. J.I
First hypothesis
We shall first give a short contrastive analysis of le, un, and les. Let us begm with the following simple case:
Out of context, sentence (I) presupposes that Sophie has only one son and asserts that this son is ill. Generally, one considers that 'the king of France' presupposes that there is a king of France. As we shall see, one must narrow the range of this proposition to only one individual: 'France has one and only one king'. Moreover, we shall also see that withm a given context, this presupposi tion is valid if one restricts its application to the 'relevant' individuals (and we shall try to give explicit conditions of relevance). Still out of context, sentence (2), taken in the partitive reading, and sentence ( 3) both presuppose that Sophie has several sons. The first sentence asserts that one of them is ill and the second one asserts that all of them are ill. We call 'presuppositions' those propositions which are implied by (I), (2), and (3) and which remain valid even when these sentences are negated. For the time being, we use this notion of presupposition but later we shall see that it is unsuitable for the phenomena studied. To corroborate these very srmple observations, let us look at the examples below:
(4.1) Le pere de Sophie est malade. [Sophie's father is ill] (4.2) # Un pere de Sophie est malade. [#One of Sophie's fathers is ill] # Un pere de Sophie n'est pas malade. [#One of Sophie's fathers is not ill] (4.3) # Les peres de Sophie sont malades. [ #Soplue's fathers are ill] # Les peres de Sophie ne sont pas malades. [#Sophie's fathers are not ill]
{ {
The abnormality of the last four examples comes from the fact that the associated presupposition6 'Sophie has several fathers' lS in contradiction with the general knowledge of what fatherhood is. On the other hand, with
( 5.1 ) #La portiere de rna voiture ferme mal. [#My car door shuts badly]
(5-2) Une portiere de rna voiture ferme mal. [One of my car doors shuts badly] ( 5.3 ) Les portieres de rna voiture ferment mal. [My car doors shut badly]
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
( I ) Le fils de Sophie est malade. [Sophie's son IS ill] (2) Un (des) fils de Sophie est malade. [One of Sophie's sons IS ill] (3) Les fils de Sophie sont malades. [Sophie's sons are Ill]
Franru Renaud
I sI
we see that the abnormality of the first sentence comes from the presupposition that 'my car has only one door' is in contradiction with general knowledge about cars. To (s.I), my garage mechanic will retort 'Which one?' (in some contexts (s.I) can be correct, as we will see in secnon J.J.J.I). The phrase 'fils de Sophie' will be represented by ).xfils.(Q.s) .x, where s is the constant identifying Sophie. The uniqueness presupposition introduced by 'le' is thus represented by 31x fils.{Q.s).x
(Sophie has only one son)
On the other hand, the items 'un' and 'les' introduce a multiplicity presupposition: {Sophie has several sons)
where card is a cardinaltty operator, reckorung the number of elements7 of ).x fils. ( Q.s) .x. We can now generalize the observations obtained from the previous examples: with N I a noun phrase such as 'fils de Sophie', le N1 presupposes that the extension ofN I is I �N I � - I or in type theory notation 31x N.x (where N is the semantic representation of the phrase N I). But, un N1 and les N1 presuppose that the extension of N I is of cardinality superior to r: liN I � > I or in our notation (card.N) > I
Let us notice that the multiplicity condition card.N > I of un N1 applies only to its partitive reading un des N1 . After the presupposition, we shall have to look at the 'assert' part of the sentence ( r ) and (2): 'le/un ills de Sophie est malade (the/one son ofSophie is ill)': (sr2) 31x {ills.(Q.s).x 1\ 3E{etat.E.(Q.x).malade 1\ m E int.E))
where E is a variable used to locate the event with respect to the speech time m, and where int.E represents the validity interval of the state 'being ill '. In the case of sentence {I), since we presuppose that Sophie has only one son, we can content ourselves with stating that (sr1) 3x (fils.(Q.s).x 1\ 3E {etat.X.(Q.x).malade 1\ m E int.E))
We replace 31x by 3x, because of the following type theory theorem 31x N.x 1\ 3y{N.y 1\ P.y) I- 31y{N.y 1\ P.y) which correspnds to the obvious set-theortic formula: ifN - {a) and N n P f:. 4' then {a) n P - {a)
We might be surprised that in (sr1), which is the representation of (2) 'un (des) fils de Sophie est malade), we also need to resort to 31x to extract the only
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
card.(A:x fils.(Q.s).x) > I
1 52 The Defirute Arocle: Code and Context candidate for illness. But note that if Sophie has two sons who are both Ill, It would be infelicitous just to say that she has an ill son. If I declare to my � collector that )e possede un appartement (I own one flat)' when in act I own two, he could remind me precisely of a general property of numerals: numerals extract the maximum number of individuals satisfying the predicate. Finally, the representation of sentence
(3), with the definite plural, will be
I 1\ Vy (fils.(Q.s).y :::> 3E(etat.E.(Q.y).malade 1\ m E int.E))
We can sum up this first sketch of the properties of le, value un des) by the following lexical entries: card.(Ax fils.(Q.s).x) >
les, and
un (with the
first outline of the representation of le N is taken from Russell's for mula for definite descripnon. Let us remember that his formula, Within the
This
form given by Montague, Is: 3yVx((N(x)
a
(x - y))
1\ P(y))
Although it is different from our formula (LI ): 31xN(x) 1\ 3x(N(x)
1\ P(x)) (after haVIng taken the same typmg on mdlVlduals)
one can show that the two formulas are equivalent. The purpose of (L I) is to bring to the fore the distincnve feature of the uruqueness presupposition. Still, the parts asserted and presupposed of the representation, inherited from the leXIcon, are put on an identical footing. But we are going to have the uniqueness condition play an essential part in the dynamic construction of the SR on the basis of the context. So we will depart radically from Kamp & Reyle (I 993), who give the same static representation to the defimte and the mdefinite (c£ pp. and 254) i.e. /..N/.P . 3x(N(x)
I 22
1\ P(x))
by rendering their DRS m logic.8 We thus see that the difference between and un mainly originate in the uniqueness hypothesis 31x N.x.
le
What is more, for the time being, formula (L I) accounts for the fact that presupposition is preserved under negation, since the neganon Will be introduced through the variable 9', in the 'asserted' subformula, Without affecting the 'presupposed' subformula.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(LI ) k. Det, /..N/.. 9' (31x N.x 1\ 3x(N.x 1\ 9'.(Q .(Q.x)) les, Det, /..N/.. 9' (card.N > I 1\ 3.5\' (N - v. .s\' 1\ 9'.5R )) un (des), Det, /..N/.. 9' (card .N > 1\ 3 1x(N.x 1\ 9'.(Q.(Q.x))))
Frmcis Renaud 1 5 3
The role ofthe context
3 .2
Is all the information we draw just from examples ( I ) and (2) representanve of le and un ? A first survey of the difficulnes met by the
the properties of
representations of(L I ) wtll give us the opportunity to introduce the problem of the contexts ofinrerpretanon. In section 3·3· we shall try to present an inventory of these contexts. Let us take at random a concrete, nonrelaoonal, count noun without any determination, such as ite (island) in:
eruptions] with the first mention of an island, sentence (6) 1s infeliatous, unless one appeals
We want to ask first, 'what island are you talking about?' When we are faced to a srylisnc device commonly found in novels and m the press, by which a
badly identified object 1s introduced immediately by definite article and whose characterisncs are diSclosed later. We shall study this deVIce m section 3·3·3- I ,
but for the time being we shall look at the device used to introduce a new object
by means of an mdefinite article ('Le 27 aout I 88 3,
une ile a ete derruite par une I 88 3, an island was
sene d'eruptions volcaruques. [On the 27th of August
destroyed . . .']} or by means of a definite article, as in this text adapted from Wilson ( I 993: p. 27): (7) L'ile de Krakatau a ete detruite un lundi mann, le 27 aout I 88 3, par un serie de pmssantes eruptions volcaruques . . . Tandis que l'ile s'effondrait dans la chambre souterraine videe par !'eruption, la mer s'engouffrait dans la Caldeira ainsi creee. [On Monday morning of the 27th of August I 8 8 3 , the island of Krakatoa was destroyed by a series of volcanic eruptions . . . As the island collapsed in the underground space emptied out by the eruptioru, the sea rushed into the caldera thus created] If we apply our rules {L I ) to the first sentence, we obtain, with some simplifica tions:
1\ name.(Q.x).kra) 1\ 3x(ile.x 1\ name�Q.x).kra 3X3E(X � eruption 1\ detruire.E.X(Q.x) 1\ int.£ � [ I 88 } ,8,27))}
3 1x(ile.x
1\
(name.(Q.x).kra is a predicate which glVes the name 'kra' to the ennty x) which after skolernization gives ile.c10
1\
name�Q.c10).kra
1\
eruption
1\ ile.c 1 1 1\
name.(Q.c 1 1).kra
Vx((ile.x
1\
name�Q.x).kra) ::>
1\ detruire.e0.d12.(Q.c1 1) 1\
x
-
c10)
1\
d12 �
inte0 � [I 88 3,8,27)
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(6) L'ile a ete detruite le 27 aout I 88 3 par une sene d'eruptions volcaniques. [On August 27th I 8 8 3 , the island was destroyed by a senes of volcanic
1 54 The Defimte Arocle: Code and Context
Thus, thefirst mention of an island, even made with a definite article, introduces
a new constant c10 into the knowledge base, which will be assumed to be, by
default, different from all other constants (except for c1 1). Empirically, we
observe that the
second
mention of an island in the second sentence obviously
{Lr) of Ia whtch will produce a new constant, different from c10 (and c1 1). But refers to this same object. So, we cannot, in this last case, use the representation
we shall see that if we use the uniqueness constraint 31x ile.x to search m the knowledge base, we will be able to bring back the right constant c10. Then all
that remains is to instantiate the existential variable of the 'assert' part of le N1
to transmit the right information to the right place. We shall set out the exact
procedure to search for the antecedent later on, but we can for the moment say
le N I -
A.9' ((A.x(N.x 1\ 9' .(Q.(Q.x)))).Srch-info)
Example (7) provides us with four first mentions of de6mte noun phrases: 'le 27 aout I 88 J', 'la chambre souterraine', 'la mer' and 'Ia caldeira'. We can disregard the interpretation problem for nouns of dates and mass nouns, since our study only deals with count nouns. The two other nouns are cases of assoctattve anaphora: they belong to the special terminology and they should be understood as 'la chambre souterraine de volcan (the underground space of the volcano)' and 'la caldeira produite par le volcan (the caldera of the volcano)'. In his knowledge base, the reader can come across the extstence of a volcano through the mention of the volcanic eruption d12 introduced in the first sentence and through general knowledge about volcanoes: volcanoes usually develop in an underground space and their exploston may produce a caldera. On this basis, it becomes possible for an algorithm to prove the uniqueness of the space and of the caldera in the restrictive dtscourse universe. We shall later propose a rule for the treatment of this problem. We shall deal later with the case of nouns with a unique referrent ('mono referring' nouns, such as 'Ia volcanologie', vulcanology) and that of deictic uses (such as 'ferme Ia porte', shut the door) which can all appear in a first mention (c£ section J.J.J). The two kinds of anaphora illustrated in (7), the direct anaphora ('l'ile') and the associative anaphora ('Ia chambre souterraine'), can be taken into account if the article 'le' does not translate directly into a logical representation taken from the lexicon (as does rule L1) but into a representation built on the basis of contextual information supplied by the knowledge base. It has frequently been observed that the starred (unacceptable) context-free linguistic examples could become 'acceptable', when some appropriate context has been supplied. Thts very common phenomenon in natural language comes
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Srch info which will return a constant c, identifying the antecedent This will then be assigned to the variable x of 3x(N.x 1\ !7' .(Q.(Q.x)) as: that the 'presupposition' 31x(N.x) will be changed into a search algorithm
Franas
Renaud 1 5 5
from the fact that one part of the information is encoded and that the other has to be constructed from the context. The methods of taxonomic description, which have allowed linguists to make progress, seem utterly ineffectual in the face of the fantastic profusion of possible contexts. We assume that to account for the contexts we will not have to give a description of all the possible contexts, but only that of the algorithm ofcontextual information processing. 3·3
Contextual restrictions
We shall try to classify the specific uses of the definite article le (with count noun). We propose three kinds of use:
3·
either the lexicon guarantees the uniqueness (monoreferring nouns); or the information stored in the memory makes it possible to check the uniqueness; or all these search processes foil but the use of a noun complement or of a deverbal head is enough to guarantee that the uniqueness hypothesis will be compatible with the basis. We srudy these three processes of analysis in the following three paragraphs.
3 . 3 . 1 Encoded information is enough to guarantee uniqueness of
reference: monoreferring nouns
Monoreferring noun phrases are noun phrases wuh a definite amcle but with a rigid referent; such is, for instance, the case ofproper names. Thus, 'Ia Seine' and 'Paris' both only refer to one 1tem. We usually choose to represent them by their minimal forms: and
/.. .9'9'.(Q�Q.s))
f.. .9'.9'.(Q�Q.p))
where s and p are constants idenufytng the river Seine and Paris. Of course, to more precise we should accompany these lexical entries with additional information as 'a French river that flows through Paris . . .' or 'Paris, the capital of France . . .' be
( 8) f.. .9'(fleuve.s 1\ ds.(Q .s).Q.(fr)) 1\ name.(Q .s).'La Seine' 1\ . . . 1\ .9'.(Q�Q.s))) (9) /...9' (capitale�Q.fr).) 1\ name�Q.p).'Paris' 1\ .9' .(Q.(Q.s)))
ds is a predicate of localization such that ds.(Q .x)� l.y) requires that the indlVldual x is in the place of the mdiVidual y Q is a function referring to a place; one can also have functions corresponding to on, under, south, etc.). Into this category of monoreferrings, we put •
toponyms such Mecque, etc.'
as:
Ia (Loire/Seine, etc.), le Massif Central Ia Chine, Ia
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
1.
2.
'
1 56 The Defirute Arocle: Code and Context •
time nouns locating pericxls on the absolute time axis: le cretace, le Moyen Age, etc.;
•
well-known people or historical events: le Roi Soleil, la Revolution Fran�aise,
•
la Restauration, etc.; names of works Qe Ramayana, le Coran, la Joconde, etc.), names of news papers {Le Monde), trade names, company names, names of stars Qa lune, le soleil), etc. 'Le pape' [the pope) or 'le dalai' lama' are not included in this category since
they refer to different people at dtfferent times. They are count nouns dependtng on time. One can speak of 'deux papes' [two popes) or of 'l'histoire different value and 'l'histoire des Chmes' would be very cxld.
Also classified among the monorefernng nouns are a class of nouns (such as 'la campagne' [the countryside), 'la paleontologie, la realite virtuelle') wluch are neither strictly countable: trois (*campagnes/tables) [three (*countrysides/tables)) nor strictly mass nouns: (de la!un kilo de) (*campagne/confiture) [(some/one kilo oD(*counrryside/ j am)) Many of them are deverbal or deadjecnval nouns ('la construction, !'intel ligence') and will not be studted because of lack of space.
fu all these terms refer to unique entities, they should be charactenzed by unique identifying constants, possibly accompanied by some prototypical properties (a foreigner or a robot coming across the term 'la CGT m a text
would certainly be very disappointed if the leXIcon just told him it is the only
entity of its kind). Thus, if we look at the representation (8) of 'la Seine', we see that the uniquenss is mechanically imposed by the introduction of the constant · s and that the definite article plays no essential part. So syntactic rules j ust bnng up
the uniqueness already encoded in the lexicon . But, on syntactical grounds, it is
necessary to protect the autonomy of such nouns as 'Seine', 'France', or 'lune' because of such uses as 'une Seine endormie', 'en France', 'la pleine lune'. Later (at the end of Section J.p. I a), we shall give the rule of interpretation of the monorefernng nouns. But for the time being, let us emphasiZe the fact that the system of identification of world objects works correctly only if we subject the knowledge base to some constraints. It will be assumed that, by default, all the predefined constants are unique and distinctfrom one another. For most speakers, the lexicon will attribute the same identifying constant to 'la Corse' and 'l'ile de Beaute'. But we can also have speakers for whom these constants are different and who are subject to opacity effects in certain contexts.9
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
des papes' [the hiStory of the popes), but 'deux Chines' [two Chinas) will have a
fr=ru Renaud 1 S 7
It is only a posteriori and by default that this uniqueness of constant constramt is 1mposed on the knowledge base. If, for example, a text mentions a river for
the first time without naming it ('Nankm est situe sur les bards d'un grand
fleuve', Nankm is on the bank of a large river, etc.) and then subsequently gives
its name, 'Le Yangz1', we will have to identify the constant coming from the
angn):
Skolem constant generator (for ex. c 101 for a large river) with the constant idennfying the monorefemng phrase (for ex.g for the Y c101 - g. That is to say, as the analysis of the text proceeds, the urnqueness of constant
condinon can always be called into question. Thus 1t can only be dynamically
managed by the theorem prover which m real time and by default guarantees the urnqueness of each constant, unless the contrary is explicitly specified. for
i � j.
3 . 3 .2 There is not enough encoded information to guarantee
uniqueness of reference but the search for contextual speci fications succeeds We will consider successively the process of searching for informacion about
the speech situation. the previous context, and the general knowledge. We shall
call these restrictions respectively deictic, anaphoric, and cognitive (or general) restrictions.
3·3·2.1 Deictic restrictions
(a) Direct deictic identification
Consider the request
( w)
Pourrais-tu fermer la pone. [Could you shut the door)
Let us suppose that trus order 1S given to a robot equipped with the visual
captors and algorithms enabling it to recognize the objects in situatiotL Moreover, let us suppose that it stands
in a room with two doors: one open and
one closed: our robot will be able to identify and to represent in its memory one
open door c1, one closed door c2, one open window c3, one human interlocutor c4, etc. This information is
all stored in a particular stratum of its knowledge the situational (orpragmatic) stratum (coming from the sensory store in bemgs). Disregarding temporal phenomena, KBJit would contain.
base KBr�r
human
among other things:
pone.c1 A ouvert.(Q.c1)
pone.c2 A ferme.(Q.c2)
fenetre.c3 A ouvert.(Q.c3)
etc.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
c1 � cJ
1 58 The Definite Arncle: Code and Context
Each of these objects c1 is associated wtth a p1cture and 1S referred to a spatial coordinate system bound to the room Moreover, the base will have the general properties of the predicates used: thus we shall have 'si une porte est ouverte elle n'est pas fermee' (if a door is open it is not closed) Vx((porte.x 1\ ouverr.(Q.x)) ::::> -.ferme.(Q.x)) or in conj unctive normal form: -.porte.x V -.ouverr.(Q.x) V -. ferme.(Q.x)
(OF)
-.porre.x V X - c1 V X - c2
(CL)
To sum up, KBsit will contain the followtng information:
KB ..t
porte.c1 1\ ouvert.(Q.c1) porte.c2 /\ ferme.(Q.c� fenerre.c3 1\ ouverr.(Q.c3) -.porte.x 1\ -.ouverr.(Q.x) V -.ferme.(Q.x) -.porte.x V x - c 1 V x - c2 (CL) etc.
(OF)
If we now look at the order to shut the door ('ferme la porte'), the meaning of 'fermer' requires that before doing the action the door must be open ('ouvert') . In tlus case, 1t is empirically obvious that the uniqueness hypothesis linked to 'la porte' should not only be 31xporte.x but UNI - 31x (porte.x !\ ouvert. (Q.x)). Wuh the help of this restriction, it is now clear that we can prove
KBsit 1- 3 1x(porte.x 1\ ouvert.(Q.x))
The major point of interest in this demorutration sterriS from the fact that if we do it by refutation 10 we can get a substitution x ... c1 which will return the right identifying corutant of the object. After negation and skolemizanon of the goal 31x(porte.x !\ ouvert. (Q.x)), we get -.porre.x V -xmverr.(Q.x) V (porte.(f.x) 1\ ouvert.(Q.(f.x)) 1\ -.(f.x - x))
By resolunon of this goal on porte.c1 and then on ouvert. (Q.c1), we get porte.(fc 1) 1\ ouverr.(Q.(£c 1)) 1\ -.(£c1 - c 1)
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Finally, each stratum of the base is supposed to contain all the information on its reference domain. Thus the mention oftwo doors c1 and c2 does not amount to saying that the base knows the existence of two doors perfectly but is unaware of other possible doors. On the contrary, we wtil have to suppose that the information is complete, that is to say no other door is involved in the situation. This property of domain closure, studied by Reiter (1984) m relational data bases, will be formulated here as:
Francis Renaud 1 59
with the help of the substitution x - c1• The resolution of porte. (f.c1) on (CL) giVes £c1 - c1
V
£c1 - c2
which by resolution on the previous -. (f.c1 - c1) gives fc1 - c1 and by substitution of the last equation in ouvert. (Q. (f.c1)) we get ouvert. (Q.c1). Its resolution with the axiom (OF) gives -.porre.c2 V -.ferme�Q.cz)
1.. .9' (f..x(porte.x 1\ .9'.(Q.(Q.x)))).c1 -�1.. .9' (porte.c1 1\ .9'.(Q .(Q.c1))) Then the classical rules of sentence corutruction will gtve the following representation for the order (ro) ordre.(fermer.e.(Q.rob).(Q .c1) 1\ porte.c1) (the robot idennfying itself by the corutant rob and bemg able to localize c1• From this example, we will remember the following three points: - the uniqueness hypothesis does not only behave as a presupposition in the traditional meaning of the term; - the uniqueness hypothesis is not a logical representation trarurnitted from the speaker to the hearer. On the contrary, it is an instruction used by the hearer m his program of processing contextual information to identify a particular individual; - finally, the formulation of the uniqueness hypothesis 31(porte.x) is incomplete, at least in the illocutionary frame of the given example. To be effective, the additional predicate ouvert. (Q.x), corning from a presupposi tion linked to the verb, must be added. In the general case, we shall note this rutrictive pmlicau by R and thus the uniqueness hypothesis by 31x(N.x II
R.x).
We can verify that if the situation did not imply the presence of the two previous doors c1 (open) and c2 (closed) but also one third door c3 open, the proof KBdt
1-
3 1x(porte.x 1\ ouverr.(Q .x))
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
which to our delight can be resolved with porte.c1 and fermi. ( Q.c1) in KBdv and which in the end gives the desired contradiction (the empry clause) and the subsntuation x - c1 . Our first intuition, according to which the uniqueness hypothesis has to play a role m the idenn£cation of the referent, lS trarulated here by an effective computation procedure which returtlS the right identifying corutant c1 (and then a spatial localizanon). The last stage of the analysis of the phrase 'la porte' Will be to apply the substitution x - c1 to its representatiotL
I 6o
The Definite Article:
Code and Context
would have failed. In this case, one can agree that the hearer/computer will ask 'which door?' to launch a dialogue. We shall sum up the analysis of a phrase that in ( w)) as follows:
- First, a procedure
Srch-info. (N
n
R)
le N1
in an identifying context (as
which processes the siruanonal
information is run, with N being the representation of N I and R the
restrictive predicate. This procedure tries to prove by refutation
KBa�t 1- 31x(N.x 1\ R.x) Two possibilities arise:
x -
t, where t is a ground term (vanable-free) such that K.Boit 1- N.t
1\ R.t
(b) or the proof fails.
If we introduce the additional procedure
Test-Srch-info, one can then sum up
the alternative: (a) the proof succeeds and
Test-Srch-info. (N n R) returns 'true' Srch-info. (N n R) rerurns the term t
Test-Srch-info. (N n R) rerurns 'false'
(b) the proof fails and
- The final representation of le will be:
(L2)
le, Det, acc:(gr:masc 1\ nb:sg) ifTest-Srch-info.N
I..NA.9'
then {l..x(N .x 1\ 9' .(Q.(Q.x)))).(Srch-info.N) else answer-by-the-question 'quel N?'
(for the time bemg, we are not considering the restrictive predtcate R. It will be introduced later). The rule shows us that we can build general algorithmic
rules
applied to the
contexrual information. Thus, we see that semantics does not need to make a content analysts of the context. Nevertheless, we will show that if we want to
ad hoc contexrual information, comprised ofsmall units (which we call strarum).
avoid the appeal to
we
conceive of KB as being
(L I ) and {L2). Let us again take example ( I ) 'le fils de Sophie est rnalade (The son of Sophie is ill)':
Before srudying other contexts, let us see if it lS possible to give a synthesis of
- either no information on Sophie's children is available in the base:
KB 1-/- 31x fils�Q.s).x KB 1-/- card(fils.(Q.s)) �
I
1-/- means 'one cannot prove'
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(a) either the proof succeeds and then at the same nme rerurns a substirution
Franas Renaud
r6r
In this case , we have to assert the uniqueness and introduce an existential variable: f..gJ 31x fils�Q.s).x 1\ 3x(fils.(Q.s).x 1\ ,9J.(Q.(Q.x)))
- or we already know that Sophie has one and only one son, Paul: Then the proof of KB I-
31x fils.(Q.s).x
successfully halts and returns the identifymg constant p ofPaul. In this way 'le fils de Sophie' becomes f.. gJ (fils.(Q.s).p 1\ ,9J.(Q.(Q.p)))
KB 1--/- 31sx fils.(Q.s).x KB I- card(fils.(Q.s)) � r
in this case the hearer can show his disagreement With the presupposinon of the utterance ( r ) (or we are m the Situation depicted m section J .J.J.I ) Thus the synthesis of rules (Lr) and (L2) iS possible. With tlus aim m mind, let us introduce an addinonal procedure to test if KB I-
card.N � 1
can be proved. We will call it Test-Srch-info-NU. The old procedure which tests whether uniqueness KB I- 31xN.x
can be proved Will be called Test-Srch-info-UNI, and the procedure which brings back the substitution, Srch-info-UNI. With these notations, the rule appears as le, Det, acc:(gr.masc 1\ nb:sg) f.. N). gJ ifTest-Srch-info-UNI.N
then ((l..x.N.x 1\ ,9J.(Q.(Q.x))).(Srch-info-UNI.N)) else ifT est-Srch-info-NU.N then retort-by-the-quesnon 'quel N?' else 3x(N.x 1\ .9'.(Q.(Q.x)))
Let us label the first formula, when Srch-info-UNI returns an identifying constant c already known in the base /..m.9' (N.c 1\ .9'.(Q.(Q.c))) as the Ukntifying val.u
/..N/...9' 3x(N.x 1\ .9'�Q�Q.x))) as the existmtial val.u.
and the second formula
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
- or we know Sophie has several or no sons:
1 62 The Definite Article: Code and Context
"Vx31y pere.(Q.x).y c; Vx3y(pere.(Q.x).y 1\ Vz{pere.(Q.x).z � z - y)) which after skolemization by the function pr appears as: (PR) "Vx"Vy(pere.(Q.x).y 5 (y - (pr.x))) Parsing 'le pere de Sophie' will make the test Test-Srch-info-UNI. (pere.(Q.s)): (PR) . . . 1- 31x pere�Q.s).x This refutation succeeds and returns the substitution identifying value of the noun phrase will be: )..!FJ ((Ax pere.(Q.s).x
x
-
pr.s. Thus the
1\ 9J�Q.(Q.x))).(pr.s)) �f,.gJ (pere.(Q.s).(pr.s) 1\ fFJ�Q.(Q.(pr.s)))) ....
So rule (L3) brings this satisfactory representation into sentence(4-1 ): pere�Q.s)�r.s) 1\ etat.e.(Q�r.s)).malade 1\ m E inter.e
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
In this latter formula, we choose not to keep the assernon of the uniqueness hypothesis 31xN.x. although this choice is disputable, as we will see later. Since the Russell uniqueness presupposition associated with the existential value is used to direct the algorithm which 'pragmatically' searches for the referent, rule (L3) thus carries out the synthesis of the logical approach ofRussell Montague and thepragmatic approach ofHawkins ( 19 77, 1 99 1 ). We recall that when this search succeeds, the referent found is transmitted to the 'asserted' part of the Russell-Montague formula. Thus we have a precise and powerful tool which. contrary to the mtuitive notion shared set (or P-set) of Hawkins, can be empirically tested. What we have is not a simple list of world objects but a knowledge base, sufficiently rich in information to enable us to define an effective procedure for processing informatiotL Although Heim's (198 8) notion of 'file' is more precise than that of the shared set of Hawkins, we can apply the same kind of criticism to it. Unlike the 'files', the KB, a well-defined and well-studied notion of AI, plays an essential part in our theory of meaning, and its usefulness goes far beyond the problems involved in the interpretation of nominal determiners. Furthermore, we give an effective procedure for processing the information contained in the KB and we will see that it is possible to structure that information mto suburuts sufficiently well-defined to avoid the resorting to ad hoc information (see later the problem of 'accommodation'). This representation (L3) makes it possible to understand why(4.1) 'le pere de Sophie est malade (the father ofSophie is ill)' is correct, whereas (5.1) 'la portiere de ma voiture ferme mal (the door of my car shuts badly)' can pose problems. We just have to spur on some general knowledge of the base: 'tout humain n'a qu'un seul pere (every human only has one father)'
Francis Renaud 163
Thus with tht help ofgenaal knowledge about the world, the definite articlt can haw an idmtifying valut at itsfirst mention. No ad hoc information11 need be added to the
KB.
On the other hand, in example (s.I ) 'la portiere de rna voiture ferme mal', the test Test-Srch-info-Uni will fail if we know that a car has several doors. In this case, the rule (L3) correctly allows for the listener to ask additional information: 'quelle porte? (which door?)' (other possibilities will be considered later). Let us now test (L3) on the following example:
(I I )
le pull de Sophie a retreci au sechage. [Sophie's sweater has shrunk during the drying]
Seine, N, /...x(fleuve.s 1\ ds.(Q.s).Q.fr) 1\ name.(Q.s).'La Seine' 1\ x
- s)
we can verify that Srch-info-UNI succeeds and returns x +- s and thus gives J.. gJ ((/...x(fleuve.s 1\ ds.(Q.s).Q.fr) 1\ name.(Q.s).'La Seine' 1\ x - s 1\ gJ.(Q.(Q.x)))).s) -+ J,.g; (fleuve.s 1\ ds.(Q.s).Q.fr) 1\ name.(Q.s).'La Seine' 1\ gJ.(Q.(Q.s))). as the identifying value of'la Seine' (after having taken out the tautology s - s). So it is thanks to the uniqueness ofreferent encoded in tht lexicon that the definite article of a monoreferring noun obtains its identifying value. (b) Indirect deictic identification (hi) Use of spatia-temporal localization of the situation Let us consider this headline from Le Monde: ( 1 2) Le gouvernement accorde une aide de I , S milliard aux agriculteurs. [The government grants aid of 1 . 5 milliards to farmers] Anyone anywhere meeting this sentence in a French newspaper will understand 'le gouvernement' as 'le gouvernement fran�is'. The noun
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Let us suppose that tlus sentence has been spoken in front of a clothes-drier. To be feliatous, there must be only one sweater belonging to Sophie present in the speech situation. Of course, the dozen or so sweaters Sophie owns as well are not to be taken into account. Even if in front of the drier there are two sweaters belonging to Sophie, only one of which has shrunk. the sentence would be infelicitous-it would be more suitable to say 'un pull de Soplue a retreci12 (one of Sophie's sweaters has shrunk)'. To account for these propernes exactly, it is imperanve, as in the robot example, to take for KB,1t only the objects which are present in the speech situation and no others. Fmally, as announced in Section 3·3· 1 , we are going to ensure that rule (L3) makes it possible to interpret the monorefemng noun phrases correctly. If we take a singleton as the semantic representation of'Seine', as in the lexical entry:
1 64 The Defirute Article: Code and Context
'governement' is relational: it is the government of a given country at a given time. If we pay no heed to the temporal locahzation. the uniqueness hypothesis will be:
3x3y(gouvemement.(Q.y).x A pays.y A Vz(gouvernement.(Q.y).z � z - x))
Furthermore, the notion of 'gouvernement' (as with 'pere') is associated with a Junctional relation : 'chaque pays a un et un seul gouvernement (each country has only one government)':
(G) Vx(pays.x � 3y(gouvernement.(Q.x).y 1\ Vz(gouvernement.(Q.x).z � z - y)))
KBnt(&) I-
3x3y(gouvernement.(Q.y).x 1\ pays.y A Vz(gouvernement.(Q.y).z � z - x))
succeeds and returns x .... gv.fr (gv skolemization function) y fr. We just have to generalise the process by which we instantiate the variables of (L3) to get +--
>-. .9.gouvernement.(Q.fr).(gv.fr) 1\ .9'.(Q.(Q.(gv.fr))) Finally, we see that the representation of 'le gouvernement' is the same as that of 'le gouvernement fran�s'. Thus we are happy to see that the rule (L3), which has been constructed to account for other kinds of example, also works perfectly well with {I2). {b2) Use of knowledge speofically shared between speaker and listener The knowledge used in the previous examples was general knowledge shared between most of the native speakers. But it can also happen that some episodic knowledge shared only by the speaker and his addressee is activated. If l say to my wife:
( I 3)
Ce soir,je passerai a Ia boulangerie. [This evening, I'll go to the baker's shop]
only my wife will know which baker's shop I am speaking about. Although there are three baker's shops near our home, the term 'Ia boulangerie' will refer to the baker's shop where we usually buy our bread. One can understand that with the help of this restrictive predicate which will guarantee the success of the first test rule (L3) will return the correct referential identifying value. But it is interesting to note that this utterance (I 3) will also be comprehensible to a hearer who has no
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
To be able to make the same kind of analysis as with 'le pere de Sophie', we just have to select the right situational stratum. Knowing that the sentence ( 1 2) has been taken from a French newspaper, the reader, wherever he is, is going to activ ate his stratum of knowledge about France. Thus KB11t(&) will be composed of (among other things): paysfr (France is a country) and the general knowledge (G) on governments. One can see that the proof of
FranCIS
Renaud 165
information on the shops near our home. In this case , the two tests of(L 3) fail and the rule returns .\.9':«( boulangerie.x II 9'. (Q. (Q.x))) -the hearerjust has to extract any baker's shop. No shared knowledge is necessary to understand ( I 3) (contrary to Hawkins I 99I who appeals to 'some subset of entities {P}, in the discourse universe which is mutually manifest for speakers and hearer on-line'). We thus see that rule (L3) accounts for �llan's referential and attributivt readings (a very specific baker's or any baker's). For the addressee, the choice between one ofthese two readings depends on his knowledge ofthe world and not on what the speaker has in mind. On the other hand, when we arnve at a new place on holiday, instead of (I 3) I would rather say: { I 4)
Ce soir,je passera.I dans
une boulangerie. [This evening, I'll go to � baker's
smce the indefinite article does not call for a process ofsearch m the base. We thus find again the observations made by Damourette & P1chon {I9SO) or Christopherson {I939) who respectively uphold the thesis of the 'notonety' and 'familiarity' of the definite article (the definite article refers to the already known individuals). This famihanty effect simply comes from the referenoal 1dennfymg reading when the kowledge base is compatible. For rule (L3) to work well, it is of the greatest importance that Srch-info-UNI acnvate the appropriate stratum of the base: the previous examples all show that we have to take a stratum KBrit containing the information about the immediate speech situation With, if necessary, speaker-hearer shared informatiotL13 3·3·2.2 Anaphoric restriction: the retrieval of information &om the representation of the previous text The stratum of the base we are g01ng to study will be called the 'textual stratum' and Will be noted KBtxt
(a) Direct anaphora
{ I s) Vincent Van Gough s'installa a Aries en I 8 8 8 . . . 11 reproduisit les sites qui se trouvent aux environs de la ville [Vincent Van Gouch settled m Aries in I 8 8 8 . . . He painted the settings in the vicinity of the town]
Let us consider the example:
In such a simple text, it is clear that at the time when 'la ville' is parsed, the search throughout the semantic representation of the previous text will return the substitution x arl: +-
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
shop]
166
The Definite Article: Code and Context
if we represent 'Arles' by A.9'(ville.arl /\ name.(Q .arl).'Arles' 1\ 9'.(Q.(Q.arl)))
{I6) (Vincent sortit de la gare d'Arles) A sa gauche, il apercevait la grand-rue qui se dirigeait vers Marseille; devant lui se dressait la porte de la ville. [(Vincent went out of the station of Arles) On his left, he saw the main street leading to Marsetlles; before him, towered the gate of the town] For a human reader, it is quite obvious that 'la ville (the town)' refers to Arles, seeing that 'la porte de la ville (the gate ofthe town)' stood in front ofVmcent who had just arrived at Arles and that Marseilles is one hundred kilometres from there. Thus, through this chain of reasorung, we have a tail to discover a formal proof, making it possible to idennfy the antecedent correctly. However, this is on condition that we introduce a restrictive predicate in the uniqueness hypothesis: UNI
-
tl existe une ville x et une seule telle que Ia porte de x se dresse devant Vincent. (there is one and only one town x such that the gate ofx stands m front ofV.)
Thus we again find the problem of restrictive predicates already met in 'ferme la pone (shut the door)' (needing the predicate 'ouvert (open)') or 'le pull de Sophie (Sophie's sweater)' (needing the predicate 'present en situation (present in situation)'. In actual fact, the restriction is brought by the sentence in which the noun phrase appears. Sometimes, as in (r6), we j ust have to take the relative: 'la ville dont la porte se dresse devant lui (the town whose gate stands in front of him)', or in ( r 3 ) 'la boulangerie ou je passerai (the baker's shop where I'll go). To confirm this hypothesis, let us look at this new example:
(I 7) Pourrais-tu coller cette note sur la porte. (Could you stick this note on the door] In a situation where there are several doors, to make it possible for our dear robot to pick the right door, it would have to use the information given by the sentence
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
and if we appeal to Reiter's domain closure hypothesis. To our great surprise, the algorithm Srch-info-UNI seems able to resolve the direct anaphora! Unfortunately, such a simple computation is seldom seen. If we look at Chapter IX of Une Vie de Van Gogh (Sweetman I 990), which describes Vincent's time in Arles, in the first five pages we find, in all, four occurrences of 'Arles', two occurrences of 'Marseille' and nine occurrences of'la ville', all referring to Aries. The menoon of two towns in the context forbids the rough use of the uniqueness hypothesis in order to search through the textual stratum. Let us look at an example closely:
FranCIS Rem.ud
167
under parsing. For example, the robot would say to itself. ifhe wants to sticka note on a door, this IS to enable everyone to read it, therefore it's better to put It on the door of the corridor. Thus, in the same speech situation , the occurrences of'la porte' in 'ferme la porte' and 'colle cette note sur la porte' may have different referents! We
thus see that the information used to identify a referent IS given not only by the knowledge base but also by the sentence under analysis. We have here a beautiful
example of noncompositionality.14 We do not claim to put foward the solunon to the problem of anaphora resolution (of definite description). This is a very difficulc question which needs huge capacities ofsemantic information processing but which may probably also
use
further syntactic notions such as the DRS ofKamp or diSCourse focus.
To give a formal translation of our observations on the resmctive predicate,
predicate R lies in adding an abstraction AR, which will give the R to the search algorithm but will keep N intact on the 'assert' part of the representanon in order
to avoid getting: 'devant lui se dressai t la porte de la ville dont la porte se dressait
devantlui (before him, towered the gate of the town whose gate towered in front of him)' le, Det, acc:(gr:masc /1. nb:sg)
/..NARA 9' ifTest-Srch-info-UNI.(N n R)
(L4)
then ((Ax.N.x /1. 9'.(Q.(Q.x))).(Srch-mfo-UNI-{N n else ifTest-Srch-info-NU.(N n R)
R)))
then repliquer-par-la-question 'quel N?' else 3x(N.x /1. .9'.(Q�Q.x))) Of course, the difficulty is to introduce this restricnve predicate advisedly. In order to do it at the level of the rule N2 ..... Det N r , we must have SO : ((SI .S2 )Restr). But the predicate Restr, being dependent on the sentence under analysts, cannot be introduced in a compositional manner. We shall not stop at
this small technical problem while a lot ofmysteries still remain about the data. To sum up, we can say that the algonthm Srch-info-UNI searches through the knowledge base in relative darkness. The only trail followed is that given by the uniqueness hypothesis with its possible restrictive predicate and by limiting the search process to the restricted frame of discourse universe defined by the situational or textual stratum (KBat or KB""J. Let us remember that in ( r 2) the interpretation of'le gouvernement' clearly needs
{b)
Associative anaphora
to 'let in the clutch' on the speech situation. On the other hand, we will see that in a narrative, in which by definition the chain of events is not linked to the speech
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
we first have to modify the rule (LJ). A simple means of introducing a restrictive
r68 The Definite Article: Code and Context time, the mterpretation of le N will call cotext).
be a matter for the previous text (wluch we
Let us see an example with nonrelational nouns:
( 1 8)
Vmcent alia a Anvers. Il vit la cathedrale, l'Hotel de Ville, les rues tortueuses de la vieille ville mais c'est le port qui le frappa le plus. [Vincent went to Antwerp. He saw the cathedral, the town hall, the winding
streets
of the old town bur it was the harbour that struck him most]
To be able to locate the cathedral (and all other underlined nouns), the reader's KB must contain the following informanon:
'Chaque grand ville ocadentale a une cathedrale (each large Western town has
'La Belgtque est un pays d'occident (Belgium is a Western countty) ' Moreover, the lexical entty of'Anvers' must contain the information that 'Anvers est une grande ville de Belgique (Antwerp is a large town m Belgium)'. Let us sum
up this information after removing the role of the West Vx(ville.x => 3y(cathe.y 1\ ds.(Q.y).O.x) 1\ Vz(cathe.z ville.av (av idennfying constant of Anvers)
1\ ds.(Q.z).O.x) => z - y)))
Mter having skolerruzed with the function c and having applied one ponms, we wtll have in the base
(KB
)
at
I
modus
:���{
c�v)
ds.(Q.(c�v)).O�v) -.cathe.z V --.ds.(Q.z) .O�v) V z - (c.av)
When we were working with a
relational noun, i t was the presence of the first
argument of the noun (usually bound to a noun complement) that enabled us to retrieve the specifications from the base. Here, with a
unary noun
(or sortal
noun, i.e. with only one argument), we must mttoduce a localization predicate
that will be used to anchor the cathedral in a place. The umqueness hypothesis will be 'il existe une seule cathedrale se ttouvant dans un lieu (there is one single cathedral m a place)'
UNJ - 3x(cathe.x 1\ 3y(ds.(Q.x).O.y)
1\ Vz(cathe.z 1\ ds.(Q.z).O.y) => z - x)))
Now all that remains for us to do is to make the proof by refutation ofKBtxt 1-
UNJ to abram the substitution x - �v. y - av. Rule (L4) then gives us the speci ficanons we looked for:
/.. � cathe.(c�v) 1\ ds.(Q.(c�v)).O�v) 1\
.9�Q �Q.(c�v)))
Consequently, 'la cathedrale' is 'la cathedrale d'Anvers'.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
a cathedral)'
Fr:mCIS Renaud 169
{ I 9) J'ai emmene rna voiture au garage. Le coffre ferme mal. [I took my car to the garage. The boot shuts badly) Ifwe have available in memory the fact that 'cars have only one boot', the search algonthm will have no trouble in returning the right spectfication. Let us say in passmg that since we are searching through logtcal representations we could also succeed with an indefinite antecedent (2o) J'ai emmene une voiture au garage. Le coffre ferme mal. [I took a car to the garage. The boot shuts badly) whereas a surface analysts that tried to restore the noun complement would give 'le coffre d'une votture (the boot of � car)' and thus would mtroduce another car into the universe. Fradin (I 984) was one of the first to observe insightfully that the knowledge used in this kind of process was stereo typic knowledge. Kleiber (I 99 3 b) found a lot of arguments to support this view. So the textual stratum is composed of the representation of the previous text and of the associated stereotypic information. This information isgeneral knowledge, shared by most ofthe speakers, and which gives the necessa ry, prototypical or scenic (coming from a scenario) properties of objects. {As a rule, we will represent it by the Retter's default rules, Sombe 1 989)). LOhner (198 5) laid great emphasis upon the role of functionality: 'the crucial condition is that the head noun in these uses provides a general two-place functional concept'. At first sight thts seems to be right, since the algorithm Test-Srch-info-UM will succeed m its quest for uniqueness if it finds a functional axiom in the KB . But Kleiber {I 993b) clearly shows that this condition on the noun is not sufficient in itself It must necessarily be of stereot ypic nature. We can rapidly illustrate this particular status of knowledge by this famous example adapted from Kleiber {I993b):
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
If the previous text now mentions the town of Brussels, we again come up agatnst a problem such as that of choosing between Aries and Marseilles which we met in (I 6). In that case, we have to analyse the sequence of actions: 'A go to X', 'A see Y' with the help of some general principles: 'If an mdividual A has gone to the place X, afterwards he is in X', 'if an individual A located in X sees Y, then Y is in X'. We must also take into account the sentence under analysts 'Vincent saw a cathedral' 3y cathe.y /1. voirE.(Q.v).(Q.y). With the help of these items of information. which are all highly likely in the situation under consideration. It becomes possible to make the refutanon which will giVe us Antwerp rather than Brussels. Let us also point out the central part played by the part/whole relanon in the associative anaphora.
1 70
The Definite Article: Code
and Context
(z r ) Nous entrames dans un petit village. L'eghse dominatt les matsons. [We came into a small village. The church towered above the houses.] (22) Nous entrames dans un petit village. ??La super�tte etait fermee. [We came into a small village. ??The superette was closed]
J.J.2.J Cognitive restrictions Until now we have coruidered the process of retrieving information stored in more or less long-term memory. The preceding analysis of deictic and anaphoric restrictions made use of information originating from the sensory store (real-time perceptions of the objects in the speech situation) or textual shon-term information (immediate anterior text). On the contrary, the
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The prototypical relation between 'small village' and 'church' is enough to interpret 'l'eglise' as 'l'eglise du village'. But substituting 'l'eglise romane (the romanesque church)' for 'l'eglise (the church)' lS enough to make the sentence look very strange. As for sentence (22), it seems very odd in French. although everyone knows a lot of small villages with a superette. The functionality of the head noun, which of course guarantees the uniqueness condition in rule (L4), seems to be vahd for the associative anaphor only if Test-Srch-info-UNI is done upon stereotypic informatiotL Rule (L4), contrary to Heim's accommodanon, 1 5 when working in recogmtion, makes it possible to account for all these effects. We thus experience a pleasant surprise when we realize that the same algorithm allows us to carry out direct and indirect specifications: direct and associative anaphora (sections J.J.2.2a and b), as well as what we have called direct and indirect deictic identification (sectioru J.p.ra and b). This result can be understood as a consequence of the fact that in the direct case the uniqueness hypothesis contains only one existential variable to instantiate, whereas in the mdtrect case it contairu two: one vanable instantiating the referent and the other, non referential, irutantiating a noun-complement relatioiL 1 6 Moreover, when one works in generation, the availability of stereocypic mformanon is not even enough to guarantee that associative anaphora with 'le' will be possible. A5 a matter of fact the anaphoric link can be established by the definite article as well as with the clitic 'en' or the possessive 'son/sa/ses'. Fradin (1 984) and Kleiber (r993a. b) have made detailed studies of this problem, while working mostly on the syntactic surface and from the standpoint of generation rather than recognitiotL By definition, we limit associative anaphora to those cases where the search for a double instantiation (of referential and non referential arguments) through the textual stratum succeeds. Although. very often, the specification can be brought by a noun complement, we prefer to avoid any syntactic delimtion.
Francis Renaud
I7I
activation of lexical memory carried out during the analysis of monoreferent is a matter of long-term memory.
We are now going to see ocher cases of recourse to this long-term memory: 24) and general knowledge of the world (2 5). Let us look
episodic memory (23,
at the following examples:
(23)
(Le Monde) L'accord de libre-&hange nord-americain doit etre examine par Ia Chambre des Representants mercredi prochain.
[(Le Monde) The North-American free-trade agreement will be examined
(24)
by the House of Representatives next Wednesday]
Un accord de libre-&hange entre l'Indonesie et la Malaisie est en cours de [A free-trade agreement between Indonesia and Malaysia is under discus-
sion]
It seems to us that If the Le Monde JOUrnalist believes he is allowed to introduce
an 'accord' by means of definite article in the first sentence ofhis article, he will also believe that his readers already know of the existence of this agreement. On the other hand, brand new information, as in
indefinite article.
The definite article
in (2 3)
(24), must be
introduced by an
seems to function here in the same manner as
with direct anaphora, but Instead of searching through the cotext the algorithm
Srch-info-UNI searches the long-term memory Iffix.T· If it finds an identifying
constant c1, 1t bnngs it back to the sentence under analysis. But if it finds no
speofying information, 1t mtroduces a new constant by means of the existential
quantifier, exactly as the indefinite article does in
(24).
In this case, then, the
'familiarity' effect of the definite article comes from the activanon of long-term
memory.
As a matter of fact,
regardless of whether the activated information comes
from the speech situation, from a previous text or from older knowledge the
analysis process will work, in every case, in the same manner. It looks as though
there were
only one phenomenon , with subcases whose boundaries are indeter
minate, as in the following example:
(25)
II faut que j'emmene rna voiture �/dans un) garage. [I have to take my car to the/a garage]
The speaker will only use the indefinite article when he has an occasional garage in mind (on holiday for example), and he will use the definite article in
the other cases. But from the listener's point of view, the analysis of 'le garage'
will resort to: (a) deictic restrictions, if he shares some very specific knowledge with the speaker, enabling him to uniquely identify the garage; (b) anaphoric restrictions, if a garage has been introduced in the previous text; (c) general
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
discussion.
1 72 The Defirute Amcle: Code and Context knowledge, telling him 'every car owner usually takes his car to the same
garage, near his home'.
Any occurrence ofthe definite article thus triggers a search procedure through the base. The search strategy we have sketched consists in reviewing the pragmatic, textual, or general strata oft� base (also taking stereotypic knowledge into account) until the time when the right stratum isfound. Ifall the search processesJail and ifno specifications are given by noun modifiers, there remains only the existential valuepossibly associated with a stylistic effect (as we shall see later in section J.J.J.I) or the principle ofconservation of unity oftime and place {by a principle ofdiscursive coherence which requires that, in the absence ofcontrary information, spatio-temporal referring points are retainedftom one sentence to the next}. All the processes of search for information have failed
In section J . J . I we saw that the uniqueness hypothesiS may be guaranteed by
the lexicon, and m section 3 -3 .2 we saw that it can be obtained by searching
through short-term or long-term memory. We snll have to look at the case when all these methods fail. We will distinguish two possible situanons:
(ti)
(i) we have a stylistic effect; the specifying information 1s given by a modifier (noun complement, relative clause).
J.J.J. I
Stylistic device and the construction of a mental model
Let us consider, for example, the following sentence that began an article in the
Ccumer International: (26) Six heures, k reveil sonne dans � chambre de Peggy Schnetder. Elle se leve, prepare k cafe, pose k pain, k rniel et � con£ture sur � table de � cuisine.
[At six o'clock, the alarm clock rings in the bedroom of Peggy Schneider. She gets up, makes coffee, puts the bread, the honey and the jam on the kitchen table] The article goes on by making us attend to the morning activities ofPeggy. The author uses the direct style generally used in television reports. He thus invites the reader to build a
mental image of the scene for himself One can assume that
this image works as a sequence of TV pictures containing an alarm clock ringing, a woman making coffee, etc., accompanied by a commentary that we shall interpret accordtng to the principles of deictic restrictions seen in section J .J.2 . I a.
This styllSnc device, which does not date only from the period of television, is also abundantly used by novelists, as in the narrative:
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
3·3·3
Franos Re112ud 1 7 3
(27) Max posa la mai}l sur le genou de Sophie. [Max put his hand on Sophie's knee]
Although. Max has two hands, a s well as Sophie's two knees, by using this direct report style the author makes us see one single hand laid on one single knee. It remains necessary to define a new test in rule (L4) to direct this case towards the existential value (since (L4) would lead us to ask the question: which hand, which knee?). Might it be that (27) could be represented by the less natural sentence:
(28) Max posa une-main sur un genou de Sophie. [Max la1d � hand on � knee of Sophie.]
3·3·3·2 Specifying noun modifier Since giving a general analysis of noun complements and of relatives is too complex to be taken up here, we shall content ourselves with semng out some simple cases.
(a) Modification by an attributive complement
Examples of attributive complements are: 'L'ile de Krakatau', 'le port de Mombassa', 'la dynastie des Tang' (Krakatoa island, etc.). These kinds of definite noun phrases can be used without difficulty at first mention , even with a hearer supposed to know nothing about them. These complements are sometimes labelled 'attributive complements' since they can be associated With structures such as 'Krakatau est une ile (K. is an island)'. Let us again take the representation of'l'ile de Krakatau' given in section 3: /.. .9 3x(ile.x 1\ name.(Q.x).'Krakatau' 1\ .9.(Q.(Q.x))) We shall start by represennng17 the phrase: 'ile de Krakatau' by A.x(ile.x 1\ name.(Q.x).'Krakatau') then by applying (L4). In general, the uniqueness hypothesis 31x(ile.x 1\ name�Q.x).'Krakatau') will return nothing and thus will lead to the existential value 'l'ile de Krakatau' - A..93x(ile.x 1\ name�Q.x).'Krakatau' 1\ .9�Q�Q.x))).
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Since for us semantic representations are used to control all possible mferences, we do not see from this point of view what could distinguish (27) from (28). In any case, if we had to take account of this stylistic effect, it would be better to save this kind of information m a special register.
1 74 The Definite Article: Code and Context We could seek to guarantee uniqueness by a default rule which would say that, exceptions aside, all entities With the same name are identical: VX"i/YVA{.nameXA 1\ name.YA ::> X - Y) But we can verify that Srch-info-UM still Jails in this case, since
this rule does
not guarantee the existence of an island with the name of Krakatoa.
(b) Modification of an extractable noun complement
(29)
La porte de man bureau ne ferme pas [The door of my office does not lock]
a de.
La porte ne ferme pas a de.
whereas
(30)
[The door does not lock]
(3 I )
requires a specifying context. Or consider:
a 3 s m sous l'eau.
(Cosquer a decouvert une grotte dans les Calanques) L'entree de la grotte se trouve
[(Cosquer has discovered a cave in the Calanques) The entrance to the cave is
3S
m below sea level]
The phrase 'la grotte (the cave)' can be analysed by anaphora according to the method seen in section
] .].2.2. Let us suppose that Srch-info-UM brings back
the constant c33 that has been introduced by the skolernization of 'une grotte'. In this case, if entrEe
(entrance) is considered as a relational noun, the set of all entries of the cave will be A.x entrEe. ( Q.cJJ).x. If the uniqueness hypothesis is not able to bring anything back, the phrase '}'entree de la grotte (the entrance of the cave)' will be represented by the existential value: A. � (grotte.c33 1\ entree.(Q.c33).x 1\ !JJ .(Q�Q.x))) It is because a cave can have one or several entrances and because we do not have any data on that subject
(Srch-info-NU.(entree. (QW.c11))
consequently
fails) that the noun specification will pass through the assertion of the uniqueness hypothesis of the noun head:
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
'Le fils de Sophie (the son of Sophie)', from which we started, but which we still have not classified, is an example of an extractable18 noun complement. As we have seen, to be identifying, a noun without a modifier, such as the noun in 'la porte (the door)', needs an anaphoric or a deictic context (the door about which we have spoken or the door here present in situation). On the other hand, m 'la porte de mon bureau [the door of my office)', the presence of a modifier will translate the task of finding a specification to the complement. For mstance, the following sentence can be uttered out of the blue:
Francis Renaud
175
31x(grotte.c33 A entree�Q.c33).x).
So we see that the definite character of'l'entree de la grotte' does not come here &om the Identifying value of (L4) but &om its existential value!
In this case, the two checks of rule (L4) guarantee that we can add the uniqueness condition to the KB without ending up in inconsistency (noted
by
o)
KB v 3 1 x{grotte.cn A entree.(Q.c33).x) 1-/- o On the contrary, with a functional noun such as 'pere', which by definition is
constrained by the property 'tout individu n'a qu'un seul pere (every person has
will return an identifying value as seen in section J . J.2. I a. But then it is this kind of uniqueness specificity which makes 'pere de Sophie' incompatible with the indefinite article 'un'. one single father)', the algorithm
Srch-info-UNI will
succeed every time and
We thus see that what is sometimes called
complete definite description
can be
obtained by the idennfyng as well as the existential value of (L4). The analysis of 'le roi de France (the king of France)' cannot be reduced to one of these cases, since 'roi' is a functional noun, 1 9 albeit one with a temporal dimens10n. In the sentence 'le roi de France est malade (the king of France is ill)', the algorithm
Test-Srch-info-UNI
in rule (L4) would activate the knowledge
that France is now a repubhc with no king. Then
Test-Srch-info-NU
succeeds (card N
-
Test-Srch-info. UNI fails and
o), so (L4) rejects the sentence by
replying 'which king of France?' Modification by a relative clause gives the same type of restrictions as
We have seen in this long section that (L4) is very useful for unifying a lot of
modification by a noun complement. We shall not consider it here.
data into a single formula which were scattered until now. This rule points out
tries to identifi a refermt in the knowledgt base. It does not transmit an assemon taken in the lexicon, but runs an
that the article 'le' in all its specific uses firstly
algorithm that searches among items of pragmatic knowledge on speech situations ('ferme la porte, le gouvernement accorde une aide' . . .), contextual, general, or episodic shared knowledge (respectively, anaphora, 'le pere de Sophie', 'le boulanger) to return an identifying constant (and more generally a ground term). The uniqueness hypothesis that guides the processing of information works well in so far as it applies to a well-delimited stratum of the knowledge base.
When the search succeeds, one gets a 'familiarity' effect, an effect which has been known for a long time by traditional grammar. But when the search fails, the hearer is forced to introduce a new referent, while possibly adding the
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(*un!le)(pere de Sophie/nez de Sophie/n!gne de Koubilai! . . .)
1 76 The Definite Article: Code and Context
Rece:�ved: 25.07.94 ReVISed verswn recetved: 1 5.03.95
FRANCIS RENAUD CRLAO EHESS 54, Bd. Raspail
75170 Pans Odo: o6 Frana t"-Ttulil rma ud@ (hmfr
APPE N D I X A
first approxtmanon of the rules ts:
(RG1)
(RG2)
(RG�)
I I I
N2 - Det N U 1 ::acdU2:acc 1\ Uo:U 1 So :- S1 .S2
(RG5)
Vz - Vt Nz Uo:UI SO : A. 9tS2.[i.. � 'IY( � .Y :::) 3E3X( XX 1\ S1.EX.Y)) /1. VX( XX :::) 3E3Y( � .Y /1. S1.E.X.Y)))
(RG6)
N2 V2 SP :- !{ U I :{acc:nb:pl n cln.�grp) /1. U 2:clv:coll then S I .(A.:!t:.Y ,. X /1. plur. 9t} tfU2:chr.disrr then S I .{). :!t:.Y ,. X /1. diStr. X" else S 1 .S2
N2 - NPR UO:U 1 SO :- A..9(.9'N.(Q.S I )))
V2 - Vi Uo:U1 SO :- A. XVX( XX :::) 3E(S 1 .E.X))
-
PH -
In
the second line of the rules, we introduce feature terms (or attribute/value lists) which enable us to make rests and to mrunut
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
additional information given by the uniqueness hypothesis, which in such a case behaves as a specifying assertion. Furthermore, we notice With pleasure that our analysis does not stop here. We observe that when the algorithm is run With a uniqueness hypothesis containing several existential variables, it is able to extract indtrect specification (deictic indirect Identification and associative anaphora) from the siruational or contexrual stratum of the base, possibly complemented with some stereotyptc knowledge. Finally, we have noted that the traditional opposition between the deictic and anaphoric interpretations has been relativized a little since in both cases we have the same kind of processing of memonzed information. This unitary approach is moreover empirically corroborated by the fact that m French no linguistic expression IS specialized in the anaphoric function and that all the anaphoric expressions can also have deictic uses (Kleiber 1 991). Rule (L4) allows us to give a unitary presentation ofanaphoric as well as deictic values with direct as well as indirect referential identification. It sheds light on the first-mention use as well as on anaphonc recall.
Franru Renaud I 77 information over long distances (Ui repre sents the fearure term at category 1 of the rule, startmg from the left). Thus in (RGI), U I :acciU2:acc tells us that the term UI of the Det and U2 of the N will unify for the agreement fearure ace (this is the notation of Smolka I 992 but It IS not essennal for the rest). The lextcal enmes are: LEX:
le, Det, acc:(nb:sg A gr:masc),
AN/..9' 3ix(N.x) A 3x(N.x A .9'N.(Q.x)))
AN/...9 3 ix(N.x A .9 N�Q.x))) une, Det, acc(nb:sg A gr:fem).
AN/..9' 3 i x(N.x A troiS,
.9 N.Q.x)))
Det, acc:nb:pl, AN/...9 351 3X(X c N A cardX - 3 A X - u.51 A .9.51 ) Ia plupart des, Det, accnb:pl, AN/...9 3.5t' 3X(X c N A card.X > II 2card.N A X - u.51 A .9 .51 ) max, NPR, acc(nb:sg A gr:masc), m Soplue, NPR, acc:(nb:sg A gr.fem), s eleve, N, acc:nb:sg, Cleve These rules only apply to count nouns wtth therr sp«rfic ��alue. On a trial basiS, let us see how our rules parse the example Us ll� se sont riunis (Tk students met). With (RGI), we have les Cleves - A..9 351 (eleve - u5l A .9.51 ) And With (RG4), we get (in the first step, we disregard the auxiliary): se sont reUUIS - )... stVX( XX ::::> 3E sereunrr.E.X) A5
the leXIcal entry of� riunir has the fearure clv-.coll and as the subJect is plural (accnb:pl) and has no group fearure, rule (RG6) Will add the plunliry constraint to the cover 3.91 (eleve
- u.51 A plur.51 A VX(51X ::::> 3E se-reunir.E.X)
We thus get the formula we were looking for.
parsed by these rules m a Simple manner. 351 3 (X c Cleve A card.X > I 12 card.Cieve A X - u.51 A VY(51 Y ::::> 3Ese reurur.E.Y)) In conclusion, let us point out the formal relanonsh1p between /e and Ia. At first sight, therr representanons seem utterly different. Nevertheless we shall show that they are closely related. They only differ in thm presuppositions: 'le N' presupposes that N IS composed of only one mdlVldual, whereas 'les N' presupposes that N IS composed of several md1viduals. Let us start &om the folloWing representations: le - AN/..9' card.N -
I
A
A
351 (N - u.51
{L I ) .9' 51 ) les - Am9' card.N > I A 351 (N - u.51 {L2) A .9 .51 )
The condition card N - I amounts to sayrng that N IS a singleton 3x(N - Q.x), so {L I ) is ).N).g> 3x351 (N -
Q.x A u.51
but we know that the cover of[x}
A
.9 .51 )
IS
[{x}}
(Q.x - u.51 ) - (51 - Q.(Q.x)) thus
).N).g> 3x351 {N A
.9 .51 )
- Q.x A 5I - (QN.x))
and by subsnrunon
A.NA..9 3x(N - Q.x A .9 .(QN.x))) moreover one can demonstrate that f-
{N - Q.x) ::::> N.x
so )M.9' 3x(N.x A
.9 N�Q.x)))
and if the uniqueness condition IS re established, one finds the previous defirunon of /e given m the lexicon LEX.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Ia, Det, acc:(nb:sg A gr.fem), 3x(N.x A .9�QN.x))) les, Det, acc:(nb:pl). AN/...9 card.N > I A 3.91 (N - u.51 A .9.51 ) un, Det, acc:(nb:sg A gr:masc), AN/..9' 3 i x(N.x) A
Fmally, the sentence LA plupart da il� se sont riunis (Most students met), winch presents so many problems for Kamp & Reyle { I 993) (as they acknowledge in section 4-4-6), IS
I 78 The Definite Article: Code and Context NOTES
7
• • •
8
9
10 II
I2
I3 I4
descnption {I) from (2) and (3}. Tradmon would claim that in (5.I) the NP Ia portiere de rna v01rure presupposes ma 1101turt a unt stUit portiht (my car has only one door) and although I feel this presup position false, I don't feel clearly (5.I) truth-valueless. So I prefer to deal with It, fts,and un m thesarne manne rand say, like Caffi (I 994). that the difference between presupposition and irnphcarure '15 more a rnarter ofdegree than a dichotomy'. This operator can be defined by A.p L{A.n N.n 1\ n.p) With N the set of narural numbers and L the iota operator (c£ Andrews I986: 203, where the narural number n IS defined as the class of equipollent sets with n elements). But we assume that this formula, in theu view, is only a first sketch. Kadmon {I990) has given a DRT analysis of uniqueness whose formaliSm IS not quite clear to me. Opaaty effectS also obtain between two proper names (such as I=lort Ducasst and Lautriamont or Khanbalik and Nltm) or between a proper name and a mono refemng phrase (Lauis XIV and It Roi Solti0. For the method of refutation by resolu non 10 logtcs, c£ Chang & Lee (I 973); 10 type theory, cf. Wolfram (I993). Contrary to the process of 'accommoda non' 10 He1m (I988: p. 370) which, more over, does not explam the d1fference between (4.I) and (5. I). Of course, one can restrict the sJruanon even more by supposing that before pumng two of Sophie's sweaters (one blue and one pirlk) to dry, someone has said 'I fear the pink sweater will shrink.'. In this case, ( n ) becomes poss1ble again, 'le pull' refemng to the pink sweater, by an anaphoric process which we shall see later. and thus depending on the speech sirua non through the speaker-hearer link. Let us recall that our system of construe-
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
I For their help m Jevelopmg the present paper, I would like to thank Irene Tamba and an anonymous reviewer. 2 For Kamp & Reyle, the semannc construction can only begin after the syntacnc analysis has been completed. 3 The appeal to the nonon of fixed-point, as made by nonmonotonic logtc, seems to reflect properly this dependency of the parts on the whole. We have med to do Without this nonon m this amcle, but we are convmced that With regard to the plural definite article we cannot diSpense With it. 4 All the quannfiers are put before the Q.,x,F. Suppose Q.. IS an formula Q1x1 ex!Stennal quannfier. If no universal quannfier appears before Q.. we choose a new constant c and replace x,. by c. If Q.1, . . Q.., are all the uruversal quantifiers appeanng before Q.. we choose a new n-place funcoon f and replace x,. by f\x.1, . . . x.J. 5 In our Vlew, the argument of the predicate � riunir IS taken in the powerset of t'IM wh1ch is a Boolean algebra, hence a serrulamce. We agree With the cnncism of Lmk gtven by Landman (I989). Moreover, Kamp & Reyle {I 993) establiSh that each complete, atomic, free upper semilamce with zero is ISOmorphic tO the powerset of a set B: 'Theorem 3 shows that the choice between a lattice theorenc and a set-theorenc approach towards the model theory of srngular and plural count nouns IS not cruaal from a smctly formal po10t of view: models based on the one approach can be readily converted into equivalent models based on the other' (p. 405}. 6 As one referee points out to us, 'the failure of the condinons' 10 (4.2) does not result m a clear lack of truth-value. Would It not be berter to speak of 'conversational �mplicarure'? Perhaps. but tfwe do I don't see why we should take apart the definite
Francis Renaud 1 79 16 We should tmprove (Ls) to perform the double tnstantianon. 17 The rule system must make use of encyclopedic informanon if 1t IS to diS anguiSh between anribuove comple ments ('Ia VIlle de Bordeaux') and possessive ones ('Ia gare de Bordeaux'): the hearer can only make the diSnncnon tfhe knows that Bordeaux rs a town and that towns hallf! stations. I 8 The noun wtthtn the complement can be relanvized, rumed into a quesnon and chnCIZed. I9 'every kmgdom has one and only one king'.
RE FE R E N C E S
Andrews, P B . ( I986), An Introduction to Mathmuwcal Logic and T}'P( Tkory· To Truth Through Proof, Academic Press, New York. Barker, C. ( 1992), 'Group terms m EngliSh: represenong groups as atoms', journal of &mantics, 9, 69-93· Caffi, C. ( 1 994), 'Presu pposttion, pragmancs', in R E. Asher (ed.), Tk Encydoptdia of Lmgua� and Linguutics, Pergamon Press. Carpenter, B. (I 992), Tk Logtc of T� Ftaturt Structuu, Cambridge Uruverstty Press, Cambndge. Chang, C. & Lee, R (I 97 3� SymboilcLogic and Mtehanical 'Ikory Proving, Academic Press, New York. Christopherson, P. (I 939), Tk Art1clts: A Study of thrir 'Ikory and Uu in English, Munsgaard. Corblin, F. (I 987), Indifini, diflni tt dimon stratif, Droz. Davies, M. ( 1 989� ' TUJO txaminm marlml six scripts: interpretations of numerically quantified sentences', Linguistics and Philo sophy. u, 3, 293-323. Damourene & Pichon (I 950� Essai tk gram mairtfrat!ft!rst: dts mots a Ia �. d'Arrrey. Does,]. M. van der ( 1993� 'Sums and quann fiers', Lmguistics and Philosophy, 16, 5 , 509so.
DOlling, J. (1990), 'Plurals and collecnve nouns: an outlook on semanncs of group expressiOns', m IUlnUn & Polos (eds.), Papas from tk &cond Symposium on Log1c and LAnguagt, Akademiat Kudo, Budapest. EnJalbert, P. (I 989), 'Notes prelinunaires :i une theone operationnelle du sens', Intd kcllca , 8, I 09-59 Enplbert, P. & Vtctom, B. (1994), 'Du langage au modele', La./. {Traittmmt automatiqut dts /angUts), 35, I, 3 7-
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
cion of semantical representanon rakes into account, as early as possible, the contexrual information available. Thus one nught say that tlus noncomposinon aliry is more a rnaner of 'unerance' representanon than 'sentence' representa tion. although we doubt that the diStinc tion is a vahd one. 1 s These de6rute NP at first mennon do not need 'accommOihtion' of the KB (addi non of information to the file to remedy mfeliciry, as Heirn says). A5 Kletber ( 1 993) clearly argues, 1t IS the narure ofthe mfor manon acnvated wluch explaUIS the devuncy of(22).
1 80 The Definite Amcle: Code and Context
definiteness Pan
II', LmguistiS(M &rich�.
Fral!fOJS(, 71..
Kleiber, G. ( 1 991), 'Anaphore-
L1n
Kleiber, G . (1993a), 'Anaphore assoa.ative et relation paroe-tout condition d'ahenation et pnncrpe de congruence ontologtque', preprint. Kleiber, G ( 1 993h� 'Anaphore assocJaOve, poncage et stereocypie', Linguistica( /nii(Sti gation(S, 17, I , 3 5-82. Krifka, M. (1990:1), 'Four thousand ships passed through the lock: object-mduced measure funcnons on events', Linguistics and Philosophy, 1 3, 5, 487-520. Krifka, M. (1990b� 'Boolean and non Bool ean "and" ', in Kalman & Polos (eds� Papn1 from tM &cond Symponum on Logic and Langua�, Akademiai Kiado, Buchpesc.
Landman, F. (1 989:1), 'Groups 1', LinguJStus and Philosophy, I l, 5. 5 59-600. Landman, F. (1989b), 'Groups II', Linguistics and Philosophy, 12, 6, 723-744Lmk, G. (1983), 'The logical analysis of plurals and mass terms. a lamce cheorencal approach', in Bauerle (t a/. (eds� M(Qmng, Uu and Inurpmation in Langua�, De Gruycer, Berhn. Lmk, G. (1 984� 'Hydras: on the logtc of relative conscrucnons with mulnple heads', in F. Landman & Velrman (eds), Van(tia of Formal &mantles, GRASS 3. Fo ris, Dordrecht. Lmk, G. ( 1 987), 'Generalized quannfiers and plurals', m P. G.irdenfors (ed.) ( 1 987). LOhner, S. (1985), 'Definites',jouma l of&man tics, 4. 2 79-3 26. Pollard, C. & Sag, I. A. (1 994), H(ad-Dri1J(11 Phr� Structuu Grammar, Uruvers1ty of Ch1cago Press, Clucago. Reiter, R. ( 1 980), 'A logtc for default reason ' mg , ArtlfiCJal /n{(l/i�u. 1 3
Reiter, R. (1 984), 'Towards a logtcal recon struction of relanonal database theory, m Brod1es (t al. (eds) On Conuptual Moddl ing, Spnnger-Verlag, Berh!L Renaud, F. ( 1 988� Pour un( simantlqu( opira tOJT"(, Langages Cro!SCs. Renaud, F. (forthcoming), Simanttq�Udu umps (/ lambda-calcul� PUF,
Pans
Schwarzschild, R. ( 1 992), 'Types of plural individuals', Linguistics and Phtlosophy. 15, 6, 641-76. Smolb, G. (1 992), 'Fearure consrramt logtcs for urnficanon grammars', journal ofLogic Programming, u, 1 -2. Sombe, L. (1 989), Raisonn(mrnts sur ks mfor mations incomp/}{(s rn LA , Teknea. Sweerman, D. ( 1 990), Un( Vt( d( Vtncrnt Van Gogh ,
L1vre de Poche.
Wilson, E. 0. (1 993), La Divmiti d( Ia vi(, ed. Od!le Jacob. Wolfram, D A. (1 993), Th( Clausol T�ory of Typa, Cambridge Uruversity Press, Ca m bndge.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
.s. 1 -27. Hawkins,]. A. (199 1 ), 'On (in)defirute amcle', journal oJLinguistics, 1.7, 405-42. Heim, I. (1988), Tk Smrantics oJikflm� and Indrfini� Noun Phra=, Garland Pubhslung (&om Ph.D. 1 982). Johnson-Laird, P. N. & Garnlum, A. (1 980), 'Descripnons and discourse-models', Lin guiStics and Philosophy, J, 3. 37 H)4Kadmon, N. ( 1 990), 'Uniqueness', Linguistics and Philosophy, 1 3, 3 . 273 - 324. Kamp, H. & Reyle, U. (1993� From Discoum to Logic, Kluwer, Dordrecht. Kempson, R. & Cormak, A. ( 1 98 1), 'Ambigu Ity and quantification', Linguistics and Philosophy. 4. 2, 259-3 1 0. Kleiber, G. ( 1 98 1), Problbna d( rifimtu, d(S(riptions, dlfinta n noms propr(S, Klinck sieck. Kleiber, G. (1983), 'Article defini, throne de Ia localisation et presupposinon existen tielle', Lan� Fral!filis(, 57, 87-105. Kle1ber, G (1986), 'AdJectif demonsrratif et arocle de6ru en anaphore fidele', m DaVId & G. Kleiber (eds), Ditmninanl$. synta.u n simanttqu(, Recherches Lmguisnques, Khncksieck. Kleiber, G. ( 1 986), 'Pour une exphcanon du paradoxe de Ia reprise Immed�:�te', Lang�U
