JOURNAL OF SEMANTICS AN INTERNATIONAL jOURNAL FOR THE INTERDISCIPLINARY STUDY OF THE SEMANTICS OF NATURAL LANGUAGE PETER BoscH (IBMScienrific Centre, Heidelberg and Umversiry of Osnabriick) REVIEW EDITOR: T1BOR K1ss (IBMScientific Centre, Heidelberg}
MANAGING EDITOR:
EDITORIAL BOARD: N. ASHER (Universiry of Texas, Ausrin) R. BARTSCH (Universiry of Amsterdam) J. VANBENTHEM (Universiry of Amsterdam) M. BIERWISCH (MPG and Humboldt University Berlin} B. BocuRAEV (Apple Computer Inc) M. BORILLO (University of Toulouse) G. BROWN (University of Cambridge} 0. DAHL (Universiry ofStockholm) S. C. GARROD (Universiry of Glasgow) B. GEURTS (Universiry of Osnabriick) M. HERWEG (IBMScienrific Centre, Heidelberg} L. R HoRN (Yale Universiry) P. N. J oHNSON-LAIRD (Princeton University) H. KAMP (Universiry ofStuttgart) S. LEVINSON (MPI Nijmegen) S. L6BNER (Universiry of Dusseldorf}
SIR J oHN LYONS (Univcrsiry of Cambridge) A. MANASTER-RAMER (WayneStare Universiry) W. MARSLEN-WILSON (MRC, Cambndge) J. McCAWLEY (Universiry of Ch�eago) M. MOENS (Universiry of Edinburgh) F. J. PELLETIER (Universiry of Alberta) M. PINKAL (University ofSaarbriicken) R. A. VAN DER SANDT (Univers�ry of Nijmegen) T. SANFORD (University of Glasgow) R. ScHA (University ofAmsterdam} H.ScHNELLE (Universiry ofBochum) A. VON S TECHOW (University ofTiibingen) M.STEEDMAN (University of Pennsylvania) W. W AHLSTE R (DFKI,Saarbnicken) B. WEBBER (University of Pennsylvania) H. ZEEVAT (Universiry of Amsterdam) T. E. ZIMMERMANN (Umversiry ofSrurrgart)
EDITORIAL ADDRESS: Journal ofSemanrics, c/ o Dr P.Bosch, lBM GermanyScientific Centre, Vangerowsrr. 18, D-69115 Heidelberg, Germany. Phone: (49"-6221-) 59-4251/4483.Telefax: (49"-6221-) 59-3200. Email:
[email protected],
[email protected] New Subscribers to the Journal of Semanrics should apply ro rhe Journals Subscriprion Department, Oxford University Press, Walton Street, Oxford, OX2 6DP. For further informacion see the inside back cover. Volumes 1-6 are available fromSwers and Zeirlinger, POBox 830,216 0 SZ Lisse, The Netherlands. ©Oxford Universiry Press All rights reserved; no parr of this publication may be reproduced, scored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, or orherwtse Without either the prior written permission of the Publishers, or a licence permitting resrncred copymg 1ssued in the UK by the Copyright Licensing
1 9HE, or in the USA by the Copyright Clearance Center, 27
Agency Led, 90 Torrenham Court Road, London W P Congress Street, Salem, Mass o1970.
The Journal of Semantics is published quarterly in February, April, August and November by Oxford University Press Subscripuon is ; I 33·
$119 per year. Second class postage pa1d at Newark NJ and at addmonal mailmg offices. ISSN 0167-
POSTMASTER: send address corrections to The Journal of Semantics, c/o Virgin Mailing and Distribution, Cargo Atlantic, 10 Camptown Road, Irvington NJ 07111-IIOS. USA.
JOURNAL OF SEMANTICS Volume
II
Number 3
SPECIAL ISSUE: SEMANTICS AND PSYCHOLOGY PART I: THE SEMAN TICS OF QUANTIFICATION Guesr Ed1rors: Amhony J. Sanford and Linda M. Moxey
CONTENTS
LINDA M. MoxEY AND ANTHONY J. SANFORD EDITORIAL INTRODUCTION
I
49
ANTHONY J. SANFORD, LINDA M. MOXEY AND KEVIN PATERSON Psychological Studies of Quantifiers
153
ROBERT J . JARVELLA AND LITA LUNDQUIST Scales in the Interpretation of Words, Sentences and Texts
I7
DAviD A. RoUTH On Representations of Quantifiers
I99
STEPHEN E. NEWSTEAD Do Verbs Act as Implicit Quantifiers?
2 15
I
Journal of Logic, Language and Information Editor Peter Gardenfors Section Editors D. Gabbay, Daniele Mundlcl, Stanley Peters, R. Stalnaker, Johan F.A.K. van Benthem This is the official publication of the European Foun dation for Logtc. Language, and Information. The scope of the JOurnal is the logical and computa tional foundations of natural, formal. and program ming languages. as well as the different forms of human and mechanized Inference. It covers the logical, linguistic, and information-theoretic parts of the cognitive sc1ences Examples of main subareas are Intensional Logics including Dynamic Logic; Nonmonotonic Logic and Belief Revision; Constructtve Logics: Complexity Issues in Logic and Ungutstics; Theoretical Prob lems of Logic Programming and Resolution; Cat egorial Grammar and Type Theory: Generalized Quantification: lnformation-Ortented Theories of Semantic Structure like Situation Semantics. Dis course Representation Theory. and Dynamtc Se mantics: Connectionist Models of Logical and Lin gutstic Structures. The emphasts will be on the theoretical aspects or these areas. The purpose of the JOUrnal is to act as a forum for researchers Interested in the theoretical founda tions of the above subJeCts and their interdisci pltnary connections. with an emphasis on general ideas increasing coherence. It is hoped that the JOurnal will meet a need of an tnterdisciplinary com muntty within the cognitiVe sciences whose most tmportant publications have been scattered across JOurnals in different areas so far. The journal IS abstracted andlor mdexed In SOCiofogrcaJ Ab· stracts: Zentralblatt fur Mathematik; tnformabOn Tochnology and the Law· INSPEC
ISSN 0925-8531 Subscription Information 1994, Volume 3 (4 issues) Subscription rate. Oft. 353.00/US$184.50 incl. postage and handlmg
Spectat rate lor members of the European Foundatton tor LogiC. Language and /nlormatton available F01 mon11nformatton. please wnts to the FoLLI Bureau. FWI. UfiiV91'�Jty ofAmsterdam, Pfantage Mutdergraeht 24, t018 TVAmsterdam, The Netherlands PO 8oJt 322. 3300 AH Dordrochl, The Netherlands PO Box 358 Accord Station. Hongham, MA 02018-{)358, USA
1111111111111111
11"'
Journal
Highlight
KWWER ACADEMIC PUBLISHERS
*
Natural Language and Linguistic Theory
Editor Joan Mating Associate Ed1tors Ellen Broselow, Carol J. Georgopoulos, Michael Kenstowlcz. Alec Marantz, Frederick J. Newmeyer
Natural Language and Linguistic Theory provides a
forum for the discussion of theoret1cal research that pays close attention to natural language data. so as to prov1de a channel of communicauon between re searchers of a vanety of pornts of v1ew The JOurnal actively seeks to bridge the gap between descnpt1ve work and work of a highly theoretical, less empirical ly oriented nature. In attempting to strike th1s bal ance, a pnmary goal of the JOUrnal is to encourage work wh1ch makes complex language data acces Sible to those unfamiliar with the language area bemg studied, and work which makes complex theo retical positions more accessible to those working outside the theoretical framework under review.
.
ll
lllll lllllllllll
lllllll lll
lll'"''
Journal
Highlight
- theoretically oriented work on the syntax. seman tics. phonology and the lexicon of natural lan guage; - mterdisciplinary contributions wntten with theo· relical llngUJsls in m1nd; - surveys of recent theoretical developments which facthtate accessibility for a graduate student read ership - reactions/replies to recent papers. - brief notes, comments or observations concerning languages or theoretical po1nts; - letters to the ed1tors; - book reviews of important linguistiCS titles; - spec1al topic issues. The JOurnal is structured to mclude;
Natural Language ana Ungwsl/C Theory Is surveyed tJy MLA lntemat•onal B•bflography. Lingu•siiC & Language BeflaVIOt Ab srraC1S. Current Q>ntenrs/Arts & Humaml•es. CCISoc•al & 88haVJoral Scumces. ASCA Arts & Humamt•es C•ratlon /nd(l)(. Socml SCienCliS C•tB11on /ndtnc, LtngUist•c Abstracts. Lingu•stJc B•bllography IBZIIBR. Computer & lnforrrw10 t n SystemiS Ab stracts Journal Socsolog,ca/ Abstracts. C8mbfldge sc,enr,flc • Technology and the Law AbstraeU. Rttsesrch Alttrt. Informat()(!
Subscription Information ISSN 0167-806x 1994. Volume 12 (4 issues) SubscnpiiOn rate: Dfl. 383.00/US$200.00 incl. post Private rate. 011.156.00/US$59.00 incl. postage and age and handling
handling P.O Bole 322. 3300 AH Oordrechl. The Netnerlanos P.O Boll 358. Accord Stauon, H•ngham. MA 02018-0358. USA
KWWER ACADEMIC PUBLISHERS
Comp utational
Linguistics
Compmational Linguisrirs �� rhc foremost joum�l devoted exclusively to compuurlooal analyses of n;arural language.
lr encompasses AI research in language, linguisrics, and rhe psychology of language processi ng and perfOrmance.
Julia Hirschberg
Wirh each 1ssue prov1ding applied and
Editor
rheo�tical papers. book reviews. technical correspondence, and leners to the editor, the JOurnal prcsenTh a
• YC2d) lngt •ncl hmdl1ngou"'dc
1deas and uc:nds in compmanonal
U.\.A. C.....dun• odd ..Wiuon.l 7�o
language.
G�T Pr17-lSII-677')
part of evc:ry research library collecuon.
JOUnul...,rdcrWlm•tmbmhip rn the .Mooci.ltion li>c C"A>mpu,.ri60·1'110 U.S.A.
Linguistics and Philosophy A Journal of Natural Language Syntax, Semantics, Logic, Pragmatics, and Processing Edttor-in-Chlef Gregory N. Carlson
Lingutsttcs
and Philosophy IS a journal for studtes focused on natural language. and is of rnteresr ro practtltoners in the diSCiplines covered '" the t1Ue. Al though the field thus descnbed is so extens1ve that a complete listing of relevant topics is precluded, at least the following specifically fall Wlthtn II: - tradtonal areas In the phtlosophy of language such as meaning, truth, reference, descnption. entail ment. speech acts - traditional areas of linguistics such as syntax. semantics and pragmaucs (when the studies are of sufficient explicitness and generality to be also of philosophical interest) - aspects of artificial intelligence concerned w1th language such as computatJonal linguistics and natural language processing - systems or logic with strong connections to natural language: modal logic, tense logtc, eptstemic logic, mtenstonal logic - philosophical quesllons raJsed by linguisttcs as a sc1ence: hngu1st1cs methodology, the status of lin guistics theories, the nature of linguistic universals - phllosophtcally intereshng problems at the inter section of hngutsrics and other disciplines: lan guage acquis1tion. language and perception, language as a soc1al convention. LlrlgUIStoCS and PtriiOSJOphy as abtlrac!ed/tndexod In Bll/14ttm S•g nalet,que, Commumry OeveloQtrlent ADs tracts, CAJrrent Cnt�ntsl Arts & HurMnlllfiS:Ants & HU/MMJit$ Cltai•OII lncJOII,I.JngU•SIIC$ & LanQu/lge BMav1or Abstracts; L8/Jguage Teachmg & LlflgviStiCS. MLA lntemattOnltl B•CIIOfJrBPhy. The Ph•losopher's lnd8J( Refire· ttvnyt Zhurnal. 5ocioiogfC4f Abstracts. 18ZIIBR Lingu,stfC$ AC· stracts Lmgu•stlcs BibiiOfJraphy, ZentraiO/att fur Mathemat•lc, RevUfl Phi/OM>ph,q1J8 de LO . 1 , but positive verbs produced significantly higher ratings than negative ones, F(I , 1 4) 1 4- 3 5. p < 0.0 1 There was no significant interaction between these =
=
=
.
.
,
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The same verbs were used in this srudy as had been used in Experiment 1 . The subject term was always people and the object term was always magazines ; these were chosen as fairly neutral terms which made reasonable sense with each of the verbs used in the experiment. Booklets were prepared containing the test items. The instructions to the booklet asked participants to consider only the information contained in the statements, and asked them to assume that they did not know the people to whom the sentences referred, nor anything about what they did to magazines. Each test item was similar to the following example:
Srephen E. Newsread 223
Table
2
Mean values, expressed as percenrages, given in Experimenr 2 Subjecr Typical value
Minimum value
Typical value
Mmimum value
78.6 65.2 74·9 23 · 3 6o.5
4 1.8 26.2 36.6 4·8 27·3
68.3 59·7 56.2 58.2 6o.6
39·3 22.0 1 8.o 41.5 30.2
69.6 46.8 8o. I n8 68.5
J2.6 20.6 4 1 .9 42.8 34·5
6o.8 50.8 84.6 79·4 68.9
22.4 1 7·3 45·I 37·5 30.6
5 1 .9 30.8 46.6 32·3 40·4
25.8 I 1.7 I 9·9 I 0.5 I 7.0
63·4 36.2 48.8 46·3 48.0
3J.I I 3.8 22.7 2 1.8 22.9
3J.I 2I.I J4. I 45·8 33-5
I 6.9 8.6 1 0.4 I 6.2 I 3.0
4 1 .3 2 1 .4 � t ·9 59· I 43· I
I 5.6 I 1.9 28.4 3 1 .4 2 1 .8
two factors, F( I , I 4) 1 .26 p < . I . Visual inspection ofTable 2 reveals that this pattern of results clearly holds both for sentence subjects and for sentence objects. A similar, but not identical, picture emerges for the typical ratings. There was no main effect of subject/object, F( I , J 4) 2,60, p > o. I , nor, in this case, any significant interactions involving this factor. There was no overall difference between manifest and subjective verbs F( I , I 4) o.o I , p > . I , but positive verbs gave significantly higher ratings than negative ones, F( I , J 4) 24.76, p < o.oo t . There was also an interaction between these two factors, F( I , 1 4) S·77• p < o.os, with subjective verbs producing higher ratings than manifest when the verbs were positive but the opposite way round when the verbs were negative. This weak interaction has no easy explanation and has not been found in any previous studies nor in the present Experiment I . �
=
=
=
=
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Manifesr positive Buy Use Have Make Mean Subjective posi rive Love Admire Undersrand Enjoy Mean Manifesr negative Destroy Harm Damage Abuse Mean Subjective negative Hare Fear Avoid Ignore Mean
Objecr
22�
Do Verbs Act as Implicit Quantifiers?
EXPE RIMENT
3
The aim of this study was to investigate implicit quantification over the same set of verbs as used in the previous studies but using neutral, abstract material. Participants
There were 25 parnCtpants, all of whom were psychology students at the University ofPlymouth who took part for course credit. None had taken part in either of the two previous experiments. Materials
Booklets were prepared in which the 1 6 verbs were paired at random with letters of the alphabet as subjects and objects. Each test item involved
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
In general, the findings of this experiment indicate that the pattern of results obtained in Experiment 1 was not due to the fact that the quantification was over subjects, since an almost identical pattern of results emerged with ratings over objects. Furthermore, the use of a different subject term seemed to have minimal effects, as did the change from asking participants to give their estimates using percentages rather than absolute numbers. It seems reasonable to conclude that the results of Experiment 1 are robust, at least using this set of verbs, and very different from those obtained in previous studies using different techniques. There is a slight complication in that two interactions emerged in Experiment 2 which are not easy to explain; however, the fact that they emerged with only one of the types of ratings suggests that they may not be strong effects. Both Experiments 1 and 2 used a restricted number of subjects and objects. Experiment 1 used the subject term Arcadians throughout, while Experiment 2 used only the subject term people . There was more variety with the object terms, at least in Experiment 1 , but it remains a possibility that the obtained effects are attributable in part to the particular subject and object terms selected. It is important to know whether the verb effects obtained are more general than this. Hence in Experiment 3 completely arbitrary, abstract material was used: the subjects and objects were all randomly chosen letters of the alphabet. Ifthe estimates remain the same, this would indicate that the verbs are powerful determinants of quantification, albeit in a rather different way than might have been assumed on the basis of previous research. If, however, the estimates change, this would suggest that the verbs are not the only factor in quantification but that subjects and objects can also have an influence.
Srephen E. Newsread
225
participants giving a typical and minimum estimate of quantity, expressed as a percentage. The following is a typical test item: This is most likely to refer to As a minimum, this refers to
Qs buy Zs % of Qs and % of Qs and
__
__
% of Zs % of Zs
__
__
Method
Participants were run in small groups of varying sizes. They were allowed as much time as they required to complete the test items. Results
The results are summarized in Table 3· With respect to the minimum ratings, an analysis of variance involving the factors of subject/object, manifest/ subjective and positive/negative showed no significant effects on any factor, nor were there any interactions between factors. The three-way interaction between all the factors approached significance, F( 1 ,23) - 4. 1 0, p - 0.05 5, but no other effects were even marginally significant. With the typical ratings, there was just one significant effect, again the three way interaction involving all the factors, F( 1 ,23) < 5·54. p 0.005. The source of this interaction seemed to be a slightly lower rating for the manifest negative verbs when they were the objects of the sentences. It is not easy to devise any plausible explanation as to why this should be the case. The results of this experiment indicate that when abstract terms are used, differences between the verb types, and indeed between the individual verbs as well, tend to disappear. =
GENERAL D I SC U SS I O N Taken together, the results from these three studies point to a very clear conclusion but one which is surprising in the context of previous research: verbs do not dominate quantity assignment. The first two experiments
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Participants were told that the letters referred to two sets of entities (people, things or objects) but that they did not know which. They were asked to fill in the percentages that they thought were indicated by each statement. They were told that there were no right and wrong answers, and that it was their personal judgement that was of interest. Each verb occurred just once in the booklet, and the order in which the verbs appeared was varied.
226 Do Verbs Act as Implicit Quantifiers?
Table 3 Mean values percentages given in Experiment 3
Subject Typical value
Minimum value
Typical value
Minimum value
60.4 63.8 73· 1 62.8 6 5.2
36.6 34·0 45·5 36·7 38.o
53·6 63·9 63·9 68.8 62.2
36.8 37·9 42.2 45·2 40. 1
74·4 69.6 52.1 66.8 6q
45·6 43-4 28.6 J7.2 38·7
62.0 59·3 56.7 67.6 6 1 .4
36·9 39·5 3 4.2 40.0 37·7
58.6 65·3 60.5 53·8 59·5
3 1 .6 J7.0 34·9 28.9 33·1
50·7 52.6 5 8.8 44·6 5 1 .7
27. 5 26.6 36·9 22.6 28.4
62.8 58.8 59·0 58.1 59·7
38.1 36.6 3 J.2 32·9 3 5-2
70.2 63.0 59·2 57·7 62.5
48.2 43· 1 3 5 ·4 . p .2 42.0
indicated a small but nonsignificant difference between manifest and subjective verbs, in the direction predicted on the basis of previous research, and a large and highly significant difference between positive and negative verbs but in exactly the opposite direction to what had been predicted. Most dramatically, when completely abstract material was used in Experiment 3, there were no differences whatsoever between verbs of these two types. It seems clear that the subjects and objects are exerting an influence on the ratings given in Experiments 1 and 2. The most likely explanation seems to be that participants were responding on the basis of the expected frequency of the complete scenario depicted by the subject-verb-object combination. It is easy to see how people might respond in accordance with their prior expectancies based on their own experience. Experiment 1 required partici pants to indicate the number of inhabitants of an imaginary town who behaved
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Manifest positive Buy Use Have Make Mean Subjective positive Love Admire Understand Enjoy Mean Manifest negative Destroy Harm Damage Abuse Mean Subjective negative Hate Fear Avoid Ignore Mean
Object
Srephen E. Newsread
227
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
in various ways. It seems likely that participants brought with them to this task expectancies as to how many of the inhabitants of any town would behave, for example, how many of them might buy television sets. And, if this is a fairly typical British town, this proportion is likely to be quite high. Similarly, participants might have expectancies concerning the number of inhabitants of a town who would destroy television sets and this number might intuitively be expected to be somewhat lower. If these expectancies are similar with other entities (such as the number of people who might be expected to buy or destroy chairs, cars, or shirts), this could readily explain the result that was obtained, that buy gave consistently higher ratings than destroy. Subsidiary findings lend some support to this explanation. It is clear from inspection of Tabies 1 and 2 that the verbs in each category are not a homogeneous group. This is perhaps best illustrated with the manifest positive verbs. The verb make seems very different to the other three verbs, consistently producing lower quantities than them. For example, in Experiment 1 the typical value accorded to make was 25.9%, while the lowest value accorded to any of the other manifest positive verbs was 6 J. l %. Perhaps, with the decline of manufacturing industry in the UK, there is an expectancy that very few things are likely to be made by anyone. In addition, there was also some evidence that the quantities implied varied as a function of the object term used. This could only be investigated in Experiment 1 , since this was the only study to use different object terms, and even here each verb was used with only four different objects. Nevertheless, there did seem to be some interesting findings. For example, the typical value chosen for use was 5 1 % for pottery, 8 1 % for cars; the typical value for have was 42% for cameras, 84% for chairs. Intuitively, these differences seem to correspond to differences in expected frequency, since we would probably expect more people to use cars than pottery in a typical town, and more to have chairs than cameras. These differences cannot be properly tested for statistical significance, and it would be wrong to read too much into them, but they are at least suggestive. However, perhaps the most compelling argument for the role of prior expectancies is that the use of abstract material in Experiment 3 completely removed any difference between different verb categories or indeed beween individual verbs. The most plausible explanation for this is surely that this abstract material led to no prior expectations and that in the absence of these there were no differences between verbs. If the present explanation is correct, then the quantities attached to subjects and objects derive from prior expectancies about the scenarios involving the combination of subject, verb, and object in the sentence. The verb might be especially influential in determining the scenario, but it is by no means the only factor.
228 Do Verbs Act as Implicit Quantifiers?
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
There are instances where the same verb seems to lead to very different expectancies depending upon the subject and object it is used with. Examples of this have been presented already, and there are likely to be even more dramatic differences in the expected frequency attached to the verb make in People make ice cubes (high frequency) as opposed to People make television sets (low frequency). Hence one would expect participants to attach very different quantities to both the subject and object in such sentences. It is of course true that the present studies differed in a number of ways from others which have been used to investigate implicit quantification. The present research used direct numerical estimates of the quantity indicated in the sentence subjects and objects, while previous studies (e.g. Kanouse I 972) have asked participants to assign verbal quantifiers. Since quantifiers can change their meaning as a function of context, while numerical values presumably do not, it seems likely that the present studies are less flawed than previous ones; hence it would be ill advised simply to dismiss the present results as being due to some experimental confound. In fact, it seems likely that studies using quantifiers rather than numerical estimates also reveal more about expected frequencies than about implicit quantification. In other words, the reason that people think that Helen likes dresses means that Helen likes MANY dresses (Kanouse I 972) may derive from prior expectations. It may be that people expect women to have a liking for dresses in general. Alternatively, it may be that the statement is interpreted as meaning that Helen has a greater than expected liking for dresses, and hence that she will like more dresses than the average person. Either way, the quantity derives from prior expectancy rather than any inherent properties of the verb. Helen likes snakes might lead to a very different estimate. This same line of reasoning can also explain other research on implicit quantification which has used the Gilson & Abelson ( I 96 5) technique rather than direct estimates of the quantifier or quantity. It will be remembered that Gilson & Abelson found that, in the light ofjust one-third positive instances, participants would much more readily accept generalizations using verbs such as have than ones using verbs such as understand . The differences between verbs were seemingly consistent, being found across different subject and object terms. According to the expected frequency explanation one might expect that variations in the terms used would lead to changes in expected frequency and hence to variations in generalizability. It is, however, possible that such variations did occur but did not show up in Gilson & Abelson's analysis, since they did not investigate interactions between particular verbs and subjects/ objects, only main effects. Since the verb is obviously an important determinant of the overall scenario it is also possible that the differences between expected frequencies were relatively small. It thus remains a real possibility that expected frequency can provide an explanation for the findings.
Stephen E. Newstead 229
Acknowledgements The author would like to thank Penny Armstead for her assistance in collecting and analysing the data reported m this paper, and Tony Sanford and rwo anonymous referees for their helpful comments on an earlier version of this paper. STEPHEN E. NEWSTEAD Department ofPsychology University ofPlymouth Drake Circus Plymouth PL4 8AA UK
Received: o6.1 2.93 Revised version received: 10.02.94
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
In recent years, the work of Gilson & Abelson has been followed up principally in research on implicit causality. It has been claimed that some verbs (e.g. hit) lead to the inference that the subject of the sentence is the causal agent, while other verbs (such as like) put the causality in the object. To illustrate, john hitsJim suggests that the John is the cause of the hitting, while John likesJim suggests that Jim is the cause of the liking (see Semin & Fiedler, 1 992, for a recent review). Categorization of verbs into different types has become more and more refined but the divisions bear much in common with the categoriza tion schemes attempted with respect to implicit quantification, as the authors of such schemes readily acknowledge. Nor is this similarity coincidental since, as Kanouse ( 1 972) and Manetti & de Grada ( 1 99 1 ) point out, implicit quantification is one possible explanation of causal attribution. A verb which suggests that the activity is carried out by only a small number of people might lead to a greater tendency to accept the subject as the causal agent, as this will make the subject more distinctive as a potential causal agent. This raises the question, then, as to whether these classificatory systems, too, owe more to prior expectations than to inherent properties of the verbs themselves. Context can undoubtedly have an effect, since the extent to which we believe that Jim is the causal agent in john likesjim will depend in part on how well liked Jim is by other people and also on how inclined John is to like people in general. What is not known, however, is the extent to which implicit causality is determined by such context-dependent expectations rather than by properties of the verbs themselves. The general tenor of the arguments put forward in this paper bears more in common with writers such as Corrigan ( 1988) who have argued that verb effects can vary as a function of context. If implicit causality is indeed primarily determined by the scenario created by the entire sentence, it will be necessary to re-evaluate the extensive research which has focused almost exclusively on the verb.
230 Do Verbs Act as Implicit Quantifiers?
RE F E R E N C E S Maxey, L. M. & Sanford, A. J. (1 993), 'Prior expectation and the imerpretation of natural language quamifiers', Eun,pean Journal ofCoxnitive Psychology, 5, 7 3-9 1 . Newstead, S. E. & Collis, J. M. ( 1 9X7), 'Con text and the imerpretation of quamifiers of frequency', Erxonomics, 30, 1 44 7-62. Peeters, G. ( 1 97 1 ) 'The positive-negative asymmetry: on cognitive consistency and positivity bias', European Journal of Social Psychology, I, 45 5-74. Pepper, S. & Pryrulak, L. S. ( 1 974), 'Some times frequently means seldom: context effects in the interpretation of quantitanve expressions', journal