JOURNAL OF SEMANTICS
AN INTERNATIONAL JoURNAL FOR THE INTERDISCIPLINARY STUDY OF THE
SEMANTICS OF NATURAL LANGUAGE
MANA GING EDITOR: PETER BoscH (IBM Scientific Centre, Heidelberg and Universiry of Osnabriick) ASSOCIATE EDITOR S: NICHOLAS AsHER (Universiry of Texas, Austin) R oB VAN DER SANDT (Univcrsiry of Nijmcgen) R EVIEW EDITOR: ANKE LODELING (Univcrsiry of Tiibingen) ASSISTANT EDITOR: ANKE LODELING (Univcrsiry of Tiibingen)
EDITORIAL BOAR D : MANFRED BIERW!SCH (MPG and Humboldt
HANS KAMP (Univcrsiry of Stuttgart)
Universiry Berlin)
BRANIMIR BocURAEV ( Apple Computer Inc) MARIO BoRILLO (Universiry of T oulouse)
KEITH BROWN (Universiry of Essex)
GENNARO CHI ERCHIA (Univcrsiry of Milan) AN N CoPESTAKE (Stanford Universiry) OsTEN DAHL (Universiry of Stockholm) PAUL D EKKER (Universiry of Amsterdam) CLAIRE GARDENT (Univcrsiry of Saarbriickcn) BART GEURTS (Universiry of Osnabriick) MICHAEl HERWEG (IBM Scientific Centre, Heidelberg) LAURENCE R. HoRN (Yale Universiry) JoACHIM JACOBS (Univcrsiry of Wuppcrtal) PHILIP N. JoHNSON-LAIRD ( Princeton Universiry) MEGUMI KAMEYAMA (SRI International Stanford)
SEBASTIAN LoBNER (Universiry of DiisscldorD Sm JoHN LYoNS (University of Cambridge) jAMES D. McCAWLEY (Universiry of Chicago) MARC MoENS (Universiry of Edinburgh) fRANCIS J PELLETIER (Universiry of Alberta) MANFRED PINKAL (Universiry of Saarbriicken) ANTONY S ANFORD (Universiry of Glasgow) AniNIM VON STECHOW (Universiry of Tiibingen) MARK STEEDM A N (Universiry of Pennsylvania) ANATOLI STRIGIN (Max Planck Gescllschaft, Berlin) HENRIEITE DE SWART (Universiry of Utrecht) BoNNIE WEBBER (Universiry of Pennsylvania) HENK ZEEVAT (Univcrsiry of Amsterdam) THOMAS E. ZIMMERMANN (Univcrsiry of Stuttgart)
EDITORIAL ADD RESS: Journal of Semantics, c/o Dr P. Bosch, IBM Germany Scientific Centre, Vangerowstr. 1R, D - 69 I 15 Heidelberg, Germany. Phone: (49-622 I-) 59-425 I I 4437- Tclcfax: ( 49-6221-) 59-3200. Email:
[email protected]
© Oxford Univcrsiry
Press
All rights reserved; no part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical. photocopying, recording, or otherwise.' without either the prior written permission of the Publishers, or a licence permitting restricted copying issued in the UK by the Copyright
Licensing
Agency
Ltd, go Tottcn.harn
Court Road,
London
W 1 P gHE. or in the USA by the Copyright Clearance
Center, 22.2 Rosewood Drive, Danvers, Massachusetts 01923. USA.
Joumal of Semarilics (ISSN 0107 jiJJ) is published quarterly in February, May. August and November by Oxford University Press, Oxford, UK. Annual subscription is US$qo per year. journal ofSemantiu is distributed by MAIL America, 2323 Randolph Avenue, Avenel, New Jersey 0700I, USA. Periodical postage paid at Rahway, New Jersey,
USA
and at additional entry points.
US POSTMASTER: send address corrections to Joumal ofSemantics, c/o MAIL America, 2323 Randolph Avenue, Avenel. New Jersey 07001,
USA.
For subsaiptio11 irformatiot� please see i11side back cover.
.
.
.
...
.. .
JOURNAL OF SEMANTICS Volume 14 Number 3
CONTENTS
ALFoNs MAEs Referent Ontology and Centering in Discourse
207
MARGA REis AND INGER RosENGREN A Modular Approach to the Grammar of Additive Particles: the Case of German Auch
237
Book Review
31 I
Please visit the journal's world wide web site at http://www.oup.eo.uk/semant
Editor-in-Chief: Joan Maling
Brandeis University, Waltham, MA, USA Natural Language and Unguistic Theory provides a forum
for the discussion of theoretical research that pays close attention to natural language data, so as to provide a channel of communication between researchers of a variety of points of view. The journal actively seeks to bridge the gap between descriptive work and work of a highly theoretical, less empirically oriented nature. In attempting to strike this balance, a primary goal of the journal is to encourage work which makes complex language data accessible to those unfamiliar with the language area being studied, and work which makes complex theoretical positions more accessible to those working outside the theoretical framework under review. The journal is structured to include: - theoretically oriented work on the syntax, semantics, phonology and the lexicon of natural language; - interdisciplinary contributions written with theoretical linguists in mind; - surveys of recent theoretical developments which facilitate accessibility for a graduate student readership; - reactions/replies to recent papers; - brief notes, comments or observations concerning languages or theoretical points; - letters to the editors; - book reviews of important linguistics titles; - special topic issues. Subscription lnfonnation:
ISSN 0167-806X
1998, Volume 16 (4 issues) S ubscription Rate: NLG 600.00/USD 308.00, including postage and handling. Special rate for graduate students: USD 45.00, incl.p/h. Please contact: Benjamin Bruening MIT Working Papers in Linguistics 200-219 MIT , Cambridge, MA 02139
and
Ph-ilosophy Editor-in-Chief: Manfred Krifka Dept. of Linguistics, University of Texas at Austin, USA Linguistics and Philosophy is a journal for studies focused on natural language, and is of interest to practitioners in the disciplines covered in the title. Although the field thus described is so extensive that a complete listing of relevant topics is precluded, at least the following specifically fall within it: - traditional areas in the philosophy of language such as meaning, truth, reference, description, entailment, speech acts - traditional areas of linguistics such as syntax, semantics and pragmatics (when the studies are of sufficient explicitness and generality to be also of philosophical interest) - aspects of artificial intelligence concerned with language such as computational linguistics and natural language processing - systems of logic with strong connections to natural language: modal logic, tense logic, epistemic logic, intensional logic - philosophical questions raised by linguistics as a science: linguistics methodology, the status of linguistics theories, the nature of linguistic universals - philosophically interesting problems at the intersection of linguistics and other disciplines: language acquisition, language and perception, language as a social convention. ·
Subscription Information:
ISSN 0165-0157
1998, Volume 21 (6 issues) Subscription Rate: NLG 630.00/US D 323.50, postage and handling.
LANGUAGE AND UTERATURE JOURNAL OF THE POETICS AND LINGUIS TICS ASSOCIATION Edited by Katie Wales
University ofLeeds
"Language and Literature is a pointer to thefuture. -1imes Higher Education Supplement n
... covers the latest developments in stylistic analysis, the linguistic analysis of.literature and related areas. With its uniquely broad coverage, the · journal offers readers easy access to all the important new research relevant to stylistics. Topics covered include: literary and non-literary stylistics, the connection between stylistics, critical theory, linguistics and literary criticism, and their applications to teaching.
Language and Literature also includes book reviews, examining new publications in the field.
Published three times a year • ISSN: 0963-9470
VIsit tbe SAGB website: http://UIUIUI.sagepub.co. ull/ OIUlEI{ FOIUI FOI{ NEW SUHSCHIBEHS -Subscribe at the Introductory Hale
� SAGE Publications, 6 Bonhill Street, London EC2A 4PU, UK � Call our Subscription Hotline on +44 (0)171 330 1266 I Email:
[email protected] USA orders to be sent to: PO Box 5096, lbousand Oaks, CA 91359 Nrune
METIIOIIS 01' 11.\\'t\IENT
__________________________
Ad� -------------
7125
1
Q I enclose a cheque (made payable to SAGE Publications Ltd) for: '· ___, Please Invoice my credit card
r:J Mastercard r:J Visa
I
· ___, Amount: '-
r:J Introductory Rate for Individuals .S.27/USS43 (Usual Rate.£34/US$54)
I II I I I I I I II III II I Expiry Date:1�...----. -'
r:J Please send me a brochure on the journal
Signature: ______ Date:
r:J Yes! I want to subscribe to Language and literature starting with Vol7 (1998)
Card No:
r:J Instftutfonal Rate.£100 /US$160
I I
IC,
Language and Information Managing Editor: Dag Westerstahl
Dept. of Philosophy, Stockholm University, Sweden Editorial Assistant: Simon Winter
This is the official publication of the European Association for Logic, Language, and Information. The scope of the journal is the iogical and computational foundations of natural, formal, and programming languages, as well as the different forms of human and mechanized inference. It covers the logical, linguistic, and information-theoretic parts of the cognitive sciences. Examples of main subareas are Intensional Logics including Dynamic Logic; Nonmonotonic Logic and Belief Revision; Constructive Logics; Complexity Issues in Logic and Linguistics; ·Theoretical Problems of Logic Programming and Resolution; Categorial Grammar and Type Theory; Generalized Quantification; Information Oriented Theories of Semantic Structure like S ituation Semantics, Discourse Representation Theory, and Dynamic Semantics; Connectionist Models of Logical and Linguistic Structures. The emphasis is on the theoretical aspects of these areas. The purpose of the journal is to act as a forum for researchers interested in the theoretical foundations of the above subjects and their interdisciplinary connections, with an emphasis on general ideas increasing coherence. Subscription Information:
ISSN 0925-8531
1998, Volume 7 (4 issues) Subscription Rate: NLG. 590.00/USD 303.00, including postage and handling. Special rate for members of the European Foundation for Logic, Language and Information. For more information, please write to the FoLLI Bureau, FWI, University of Amsterdam, Plantage Muidergracht 24, 1018 TV Amsterdam, The Netherlands
Introduce a frien colleague to
journal of Semantics
Please send the person named below a free copy of PLEASE PRINT DETAILS
Name Address............... City/County...... Postcode..............
............................ ..
.
·
.
·
··' :...._
on adverti this journ please c
tel/fax:
+44 (0) 1235 201904
e-mai I:
[email protected]. uk
Scope of this Journal The JOURNAL OF SEMANTICS publishes articles, notes, discussions, and book reviews in the area of natural language semantics. It is explicitly interdisciplinary, in that it aims at an integration of philosophical, psychological, and linguistic semantics as well as semantic work done in artificial intelligence and anthropology. Contributions must be of good quality (to be judged by at least two referees) and should relate to questions of comprehension and interpretation of sentences or texts in natural language. The editors welcome not only papers that cross traditional discipline boundaries, but also more specialized contributions, provided they arc accessible to and interesting for a wider readership. Empirical relevance and formal correctness are paramount among the criteria of acceptance for publication.
Information for Authors:
Papers for publication should be submitted in 3 copies to the
managing editor. They should be typed on A4 (or similar format), one-sided, double-spaced, and with a wide margin and must be accompanied by an approx. 200 word summary. Notes and bibliographical references must appear at the end of the typescript. All bibliographical references in the text by author's surname and year of publication. Diagrams must be submitted camera-ready. All papers submitted arc subject to anonymous refereeing and are considered for publication only on the assumption that the paper has neither as a whole or in part already been published elsewhere nor has elsewhere been submitted or accepted for publication. Unless special arrangements have been made, copyright rests with Oxford University Press. Authors
receive 20 offprint� of their published articles and 1 o offprints of their published reviews, free of
charge. Larger numbers can be supplied at cost price by advance arrangement.
Subscriptions: The Journal of Semantics is published quarterly.
Institutional: UK and Europe £78.oo; USA and Rest ofWorld US$140.00. (Single issues: UK and
Europe £23.00; USA and Rest of World US$4o.oo.)
Personal:* UK and Europe £38.oo; USA and Rest of World US$71.00. (Single issue: UK and
Europe £! 1.00; USA and Rest of World US$21.oo.)
*Personal rates apply only when copies are sent to a private address and payment is made by personal cheque/credit card.
Prices include postage by surface mail or, for subscribers in the USA and Canada by Airfreight or in Japan, Australia, New Zealand and india by Air Speeded Post. Airmail rates are available on request. Back Issues. The current plus two back volumes arc available from the Oxford University Press,
Great Clarendon Street, Oxford OX2 6DP. Previous volumes can be obtained from Dawsons Back Issues, Cannon House, Park Farm Road, Folkestonc, Kent CT19
sEE.
tel +44 (o)1303
850101, fax +44 (o)1303 850440. Volumes 1-6 are available from Swcts and Zcitlingcr, PO Box 830, 2160 SZ Lisse, The Netherlands. Payment is required with all orders and subscriptions arc accepted and entered by the volume.
Payment may be made by cheque or Eurochcquc (made payable to Oxford University Press), National Girobank (account 500 1 056), Credit cards (Access, Visa, American Express, Diners Club), or UNESCO coupons. Please send orders and requests for sample copies to the Journals Subscriptions Department, Oxford University Press, Great Clarendon Street, Oxford OX2 6DP, UK. Telex 837330 OXPRES, tel +44 (o)1865 267907, fax +44 (o)1865 267485.
Copyright: It
is a condition of publication in the Journal that authors assign copyright to
Oxford University Press. This ensures that requests from third parties to reproduce articles arc
handled efficiently and consistently and will also allow the article to be as widely disseminated as possible. In assigning copyright, authors may usc their own material in other publications provided that the Journal is acknowledged as the original place of publication, and Oxford University Press is notified in writing and in advance.
Advertising:
Advertisements arc welcome and rates will be quoted on request. Enquiries
should be addressed to Helen Pearson, Oxford Journals Advertising, PO Box 347, Abingdon SO, OXq
s)O<.,
UK. Tel/fax: +44 (o)1235 201904. Email:
[email protected].
journal oJS
14: 207-23)
© Oxford University Press
1997
Referent Ontology and Centering in Discourse ALFONS M A E S Tilburg University
Abstract This article describes three production experiments in which the role of the ontology of referents is investigated with respect to establishing and maintaining local coherence in
discourse. The experiments elaborate on the basic claims put forward by Centering device in two- or three-sentence seque�ces in which concrete and abstract referents were
manipulated along the lines of variables which in centering theory are claimed to determine the type of center transition and hence the accessibility of referents. The
distribution of different types of anaphoric expressions indicated that-all other things
being equal-abstract referents require more 'referential force' than concrete referents, which is claimed to be due to their being less accessible by their very nature. Furthermore,
demonstrative nominal and demonstrative pronominal anaphors appeared to function quite differently in expressing differences in transition stages of discourse referents.
There are two facts about referential coherence in discourse which are hardly controversial among cognitively oriented linguists: one is that the form of referential expressions is indicative of the . accessibility of the underlying discourse referent; the second is that the accessibility of discourse referents is to be presented on a scale indicating the degree of mental activation of the discourse referent. A way of presenting these two convictions in an elegant fashion is by relating different degrees of mental activation of discourse referents to different types of referential expressions. Although this is by no means the whole story about discourse anaphora, much of the recent work on discourse reference is based implicitly or explicitly on such a theoretical vehicle (e.g. Ariel 1994; Bosch 1 988; Bosch & Geurts 1990; Chafe 1 987; Garrod & Sanford 1982; Gernsbacher 1989; Givan 1992; Grosz, Joshi, & Weinstein 1995; Gundel, Hedberg, & Zacharski 1 993; Maes 1996a; Tomlin 1987). Within this line of thought, one might distinguish as a starting poi�t two degrees of mental activation of discourse referents, i.e. activated and non activated referents. This dichotomy is intuitively satisfying if we take it to reflect, on the one hand, the two major stages in the life cycle of a discourse referent, i.e. introduction and maintenance and, on the other, the corre sponding two major functional classes of referential expressions, i.e. non-anaphoric and anaphoric expressions. The cognitive relevance of this
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Theory. Subjects were asked to complete sentences or fill in the most adequate anaphoric
208
Referent Ontology and Centering in Discourse
dichotomy in terms of differences in processing effort and strategies is sufficiently accounted for {see e.g. Murphy I984; Clark & Haviland I 977; Garrod & Sanford I 977) Such a neat correlation between referent activation and anaphoricity is far too rough, however, both in that it wrongly assumes that previous mention is the only source of referent activation and it underdetermines the variety, the distribution, and the fluctuations of anaphoric as well as non anaphoric expressions in discourse. As far as 'new' discourse referents are concerned, a variety of expression types, such as indefinite NPs, proper nouns, or definite NPs, reflect different degrees of referential 'givenness'. In the first place, this variation is due to different 'initial' degrees of assumed familiarity or givenness of the referent {the way it is implemented e.g. in the familiarity scale of Prince I 98 I or the givenness hierarchy in Gundel et al. I 993). In the second place, it is influenced by the intended semantic interpretation of discourse referents, as it is expressed in terms of dichotomies such as generic vs. non-generic, specific vs. non-specific instantiated vs. not instantiated, referential vs. attributive. Finally, the linguistic variation of referents reflect� inherent conceptual characteristics and categories of discourse referents. For example, following the tripartite cognitive ontology proposed by Fraurud {1996), the main expression types of referents can be argued to be typically related to the main ontological classes of referents, i.e. individuals {proper names),J nctiona ls (definite NPs), and instances (indefinite NPs). Anaphoric expressions referring to 'old' discourse referents, show a similar variation mainly due to their having different degrees of pro minence or focus of attention. In recent years, many experimental and corpus-based studies have resulted in a number of contextual variables which are claimed to affect the attentional status of referents (for a short survey, see Maes I996a: 7-8). Thus, different types of anaphoric expres sions (e.g. nominal vs. pronominal, demonstrative vs. non-demonstrative) reflect fluctuations in the attentional state of referents, the most common cover terms for which are maintenance, continuation, retention, shift, and re-establishment of referents. These considerations result in a view of a referential expression as being moulded by two kinds of variables, which are inextricably inter twined: on the one hand, a referential expression exhibits the initial assumed familiarity, the semantic interpretation and ontological status of the referent-.,.characteristics which are pre-eminently manifest in intro ductory expressions-and, on the other, it reflects the fluctuations of the attentional state of the referent, which is typical of subsequent {i.e. anaphoric) mentions of the referent. In this article, I will concentrate on the intersection and the interaction .
.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
u
Alfons Maes
209
of both types of variables. In particular, I will take the following perspective: if we observe that the variation of referential expressions used to introduce a referent reflects different 'initial' and 'inherent' properties of discourse referents, and if we observe that the variation of referential expressions used to maintain a referent reflects different attentional fluctuations of a referent in discourse, we might ask whether the initial conceptualization of a referent has an influence on the way the referent is maintained in subsequent discourse. When we start from the accessibility view on discourse reference, the answer to this question might seem simple: the very fact that the referent is introduced seems to neutralise initial or intrinsic activation differences of referents, as can be validated by observations as in (I). Whether the introduced referents are previously known by name, identifiable as a unique or generic entity, whether they are attached to another known referent ('inferrable') or not, once they are introduced, they obey the same rules governing the attentional fluctuations of referents, i.e. they can simply be resumed by a zero pronoun (Ia) or a simple pronoun {Ib), and they can be shifted to by a stressed pronoun or a definite {Ic). Similar observations can be made for other referent characteristics, such as generic vs. non-generic, specific vs. non-specific, or referential vs. attributive. ·
In this article, I shall go into one inherent aspect of discourse referents which I claim is relevant for the way in which attentional fluctuations of the referent in subsequent discourse are expressed, i.e. the ontological status of referents. I will focus on the difference between concrete and abstract referents. Without entering into the large body of fine-grained semantic and philosophical discussions on different ontological types of referents, we can safely refer to the common understanding of a distinction between referents which can be perceived as distinct 'concrete' entities (humans, animals, physical objects) and referents which express some kind of state of affairs or activity (states, events, situations, speech acts, facts, proposi tions). Concrete and abstract referents are conceptually different in that concrete referents have some kind of individual ontology (see Fraurud I996 for different degrees of individuation of concrete referents), which is hardly ever the natural conceptualization of an abstract referent. Likewise,
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(I) a Yesterday, Eddy Merckx;/the doctot;/the pope;/the boy I live with;Ia man;/ a man I never met before; saw Paul and ¢; told him that . . . b Yesterday, Eddy Merckx;/the doctor;/the pope;/the boy I live with;/a man;/ a man I never met before; saw Paul. He; told him that , . . c Yesterday Paul saw Eddy Merckx;/the doctor;/the pope;/the boy I live with;/a man;/a man I never met before;. HE;/ The LATter; saw Paul as well.
210
unlike abstract referents, concrete referents have natural conceptual categorical relations with other concrete referents (part/whole, hyponymic/ hyperonymic, set/member, etc.). Linguistically, concrete · and abstract referents differ as well; concrete referents are expressed conventionally by nouns and noun phrases, whereas abstract referents are realised by a rich repertory of sentential, verbal, and nominal constructions, such as full clauses, infinitival constructions, gerunds, that-clauses, bare infinitives, VPs, deverbal nouns, nominalized verbs, and simple nouns, the complexity of which is sufficiently demonstrated by Fraurud (1992), Asher (1993), Romijn (1996), Maes (1996b), and Hellmann & Dahl (I 994). Expressions referring to concrete referents entail either features of the referent (such as animateness or (fe)maleness) or linguistic features of the antecedent, such as morpho logical number and gender, whereas abstract expressions do not necessarily express either of these features. . Although these indications do not fully account for the concrete-abstract distinction (because, amongst other things, they give the wrong impression of concrete and abstract referents as being monolithic categories), they do suffice for current purposes in that they address two classes of referents for which a different 'inherent' degree of mental activation can be hypothesized on the basis of their different ontological status. In this article, I aim at testing this hypothesis by investigating whether this difference has an effect on the way in which both types of referents are maintained in subsequent discourse. In answering this. question, one could start from the intuitions behind the examples (ra-e). Once a referent (either concrete or abstract) is linguistically introduced, its life cycle as a discourse referent (maintenance, shift, retention, re-establishment, etc.) is controlled by principles of attentional change which are independent of ontological differences between referents. Recent theories on discourse reference implicitly adhere to this position (e.g. Prince I 98 I; Ariel I 994), in that they all consider the linguistic introduction of a referent (whatever its ontological status) as a sufficient condition for accessing specific levels of activation, associated with specific types of expressions. In this article, I will hypothesize that the concrete-abstract difference affects the behaviour ofa linguistically introduced referent as it is expressed in subsequent anaphoric expressions. In principle, two related motivations can be given for such a hypothesis, a transformational and an ontological one, roughly comparable to Hankamer and Sag's suiface and deep interpretation process respectively (Hankamer & Sag I976; Sag & Hankamer 1984). The first starts from the observation that abstract anaphors usually have a complex, verbal or sentential antecedent. Transforming such a complex structure into an anaphoric reference object requires a structural transformation resulting in the 'objectification' or 'hypostatization' of the referent (see Conte 1 996;
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
·
Referent Ontology and Centering in Discourse
Alfons Maes
2I I
Fraurud 1992; and for a similar view Webber 1991), whereas concrete antecedents are usually syntactically simple and can be transformed into an anaphor by simple rules of attenuation and congruence. (2) The little pin causes the increase in pressure;. This; generates an increase of the compression of the motor. 1
THE RELEVANCE O F THE ONTOLOGY O F REFERENTS: C ORPUS - BASED EVIDENCE Not only motivations, but also empirical indications can be put forward in favour of the relevance of the ontology of referents. First, it is plausible to
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Applied to (2), it is plausible to assume that the processing of the abstract anaphor this requires an extra transformation from a predicate/proposition like representation of the antecedent into a discourse referent representation ofthe anaphor. The abstract antecedent in (2) will be processed and represented as a predicate-like information structure. The processing of the abstract anaphor requires a transformation into an entity-like structure, which in turn has to be plugged into the predicate of the second sentence. This transforma tion might result in a lower accessibility ofthe abstract referent, might require extra cognitive effort in maintaining the referent, which might be reflected in the choice of stronger expressions (such as nominals as opposed to pro nominals, or demonstrative as opposed to non-demonstrative expressions). The transformation becomes more and more plausible as abstract anaphors require more complicated transformations, in which whole clauses, sentences, or stretches ofdiscourse are concerned, instead ofsimple argument positions in a dame, and in which not only the event itself is concerned but also the expression or proposition of the event or speech act associated with the event. For this explanation to be valid, the different degrees of transformation or ·hypostatization as exhibited by the different degrees of structural complexity (going from simple nouns to stretches of discourse) should be reflected in different degrees of referent accessibility. Second, one could think of abstract referents as being by their very nature simply less accessible. It is plausible to assume that it is more difficult for human beings to conceptualize abstract events and processes than humans, animals, and physical objects, because concrete objects are in themselves distinct ontological categories which can be recognized, identified, named, pointed at, classified, discriminated, retrieved from the mental lexicon, etc., whereas all these cognitive actions are to a large extent �xcluded or more difficult in the case of abstract referents. This makes abstract referents more difficult to refer to, which again results in the use of stronger expressions.
212 Referent Ontology and Centering in Discourse
=
=
Table 1
The proportion of abstract and concrete 'first anaphors' (pronominal, definite NP, and demonstrative NP) proportion of 'first anaphors' pronominal anaphors
abstract referents concrete referents definite NP-anaphors
abstract referents concrete referents demonstrative NP-anaphors
abstract referents concrete referents total anaphors
abstract referents concrete referents
(n (n (n (n (n (n (n (n (n (n (n (n
= = = = = = = = = = = =
6r6) 43 7) 1 79) 62o) 297) 323) 453) 295) rs8) r689) 77 8) 3 76)
0.)2 0.41 0.88 0.7 3
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
assume that abstract referents are relatively less persistent, i.e. they have a relatively shorter life cycle than concrete referents. Support for this assump tion comes from the analysis of 'first anaphors' in a corpus of Dutch written texts? Within the life cycle of a discourse referent, first anaphors come in second position: they immediately follow the introduction of a discourse referent, and anaphorically refer to the referent for the first time. So, the higher the proportion of first anaphors, the lower the proportion of anaphors referring to a referent for the second, third, . . . time, and hence the lower the relative persistence of the underlying referents. Table I shows that abstract referents are relatively less persistent than concrete referents in that the proportion of first anaphors referring to abstract referents is higher than the proportion of concrete first anaphors. A second corpus-based indication can be found in the observation that abstract referents are associated more with marked anaphoric expressions than concrete anaphors are: in the same corpus, So% of all demonstrative pronominal anaphors (n 410) and 65% of all demonstrative nominal anaphors (n 45 3) refer to abstract referents.3 This could be looked upon as a support for the view that abstract referents demand more referential force. These indications, however, are by no means sufficient. They lack precision regarding the role of all contextual conditions determining the activation of abstract or concrete referents in discourse, and they cannot disentangle the role of and the interaction between referent ontology and antecedent complexity in interpreting abstract anaphors. Therefore, we need more methodological control over the relevant parameters determining the expression type of referents on the local level of discourse.4
Alfons Maes
21 3
CENTERING T HE O RY: A M ODEL O F L O CAL C O HE RE N CE IN D I SCO URSE
(3) a Terry really goofs sometimes. b Yesterday was a beautiful day and he was excited about trying out his new sailboat. c He wanted Tony to join hiin on a sailing expedition. d He called him at 6 a.m. e Tony was sick and furious at being woken up so early. f He told Terry to get lost and hung up. g Ofcourse, Terry hadn't intended toupsetTony. (Groszetal. 1995: 207-8) (4) a Terry really goofs sometimes. b Yesterday was a beautiful day and he was excited about trying out his new sailboat. c He wanted Tony to join him on a sailing expedition. d He called hiin at 6 a.m. ?e He was sick and furious at being woken up so early. (Grosz et al. 1995: 207) (5) a Terry really goofs sometimes. b Yesterday was a beautiful day and he was excited about trying out his new sailboat. c He wanted Tony to join him on a sailing expedition. d He called him at 6 a.m. e Tony was sick and furious at being woken up so early. f He told Terry to get lost and hung up. ?g Of course, he hadn't intended to upset Tony. (Grosz et al. 1 995: 207) Centering specifies an attentional mechanism which guides local coherence in discourse. The main starting. points are the claims that each
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
An appealing starting point for the assessment of the role of the ontology of referents in establishing local referential coherence is Centering Theory. Centering offers a theoretical a<:count of the interaction between local referential coherence and choices of referring expressions. According to a recent survey article (Grosz et al. 1995: 204-5), centering starts from the idea 'that differences in perceived coherence correspond in part to the different demands for inference made by different types of referring expressions, given a particular attentional state.' Therefore, the form of referring expressions plays an important part in controlling the inference load associated with the processing of each new discourse unit. Adequacy differences and obvious differences in coherence-e.g. (3) being more adequate than (4) and (s)-are considered as a psychological reflex of differences in inference load associated with types of referring expressions:
214
Referent Ontology and Centering in Discourse
(6) a Susan gave Fred a pet hamster. b In his opinion, she shouldn't have done that. continue
c She/Susan just assumed that anyone would. love a hamster. shift
c' He/Fred doesn't have anywhere to put a hamster cage. d Giving a pet as a gift can be somewhat of an imposition. (Gordon et al. 1993: 3 37-8)
INVESTI GATING THE ROLE O F THE O NT OL O GY OF REFERENTS: THREE PRODUCT I O N EXPERIMENTS The main premises of Centering Theory can perfectly well be used to investigate the role of the ontological type of referents in establishing local
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
utterance has a most prominent entity, i.e. the backward-looking center (Cb), and an ordered list of candidates for the role of Cb in the next utterance, i.e. the so-called list of forward-looking centers (CD. Further more, centering defines three transition stages for centers and a preference among them: continuation is preferred to retention, which is preferred to shift. Finally there are rules and constraints on center movement and realization.5 Centering focuses on a limited set of discourse characteristics as determinants of the prominence of referents: the pronominal (as opposed to the nominal) form of the anaphor, subject position and initial position in the sentence. These characteristics are widely acknowledged to be associated with prominent referents, but centering offers the mechanism to unam biguously investigate the role of each of these variables on the attentional fluctuations of referents. Hence, reading experiments (e.g. Gordon & Scearce 1995; Gordon, Grosz, & Gilliam 1993) as well as production studies (Brennan 1995) provide us with evidence for the determining role of these variables. Brennan (1995) found that in discourse production, 'centered' entities, much more so than non-prominent entities, are associated with subject position and pronominal form. Gordon, Grosz & Gilliam found a so-called repeated noun penalty (which means a significantly longer reading time for literally repeated nominal anaphors) for referents which in the previous utterance occupied a subject position as well as an initial position. This means that the reading time for Susan in the continue condition (6c) as well as for Fred in the shift condition (6c ') is longer than for their respective pronominal variants, which is said to prove that subject position contributes to prominence independently of initial sentence position.
Alfons Maes
21
5
( 7) a The little pin causes the increase in pressure. b The increase in pressure is controlled by the little pin. This implementation of the concrete-abstract dichotomy is assumed to generate the most valid experimental results. If we assume that inanimate physical objects represent a low degree of referent individuation compared to other classes of concrete referents, such as humans and animals (Fraurud 1996), and if we observe that (at least some) inanimate physical objects are
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
coherence in discourse. The contextual variables which according to centering determine the transition stage of referents (expression type, syntactic position, and sentence position) enable us to define contexts within which the attentional state of referents and their ontology can be controlled and manipulated. This is what the experiments described in the remainder of this article are intended to do. The production experiments extend and modify the current empirical research on centering in three ways. First, methodologically, they focus on controlled production tasks as a means of investigating attentional state properties of referents. In these tasks, opting for a particular type of referential expression is taken to reflect assumptions on the cognitive effort needed to select a referent and hence on the attentional state of the referent (see also Madsen-Wilson, Levy, & Tyler 1982: 340; Fletcher 1984; and Garcia 1996). Hence, as in centering, adequacy differences based on differences in referential expressions are used as a method of assessing the mental activation of referents. Second, thus far, the centering experiments as well as most of the comprehension experiments on the processing of anaphoric devices in general-have concentrated on two major classes of anaphoric expressions, nominal (mainly proper names) and (unmarked) pronominal anaphors. In the production experiments, a wider range of expression types is taken into account: marked nominal anaphors (demon strative NPs), unmarked nominal anaphors (definite NPs), marked pro nominal anaphors (demonstrative pronouns), and unmarked pronominal anaphors. Finally, whereas most experimental work on discourse anaphora restricts itself mainly to concrete, preferably animate referents, my production experiments vary over concrete and abstract referents, both implemented in one specific way. Concrete referents were restricted to inanimate physical objects; abstract referents were events and processes as expressed in activity, accomplishment, and achievement verbs, and realized as an argument of a finite verb. Implementing the abstract-concrete distinction in this way offered the opportunity to develop syntactic and semantic environments within which abstract and concrete referents could be varied in a natural way along the lines of the centering variables, as is exemplified in ( 7).
2r6
Referent Ontology and Centering in Discourse
EXPER I MENT I: CLOZE EXPERIMENT I The first experiment focuses on the relationship between the markedness of pronominal devices (unmarked it vs. marked that/this), the transition stage of the referent (continuation or shift) and its ontological type (concrete vs. abstract). Twelve two-sentence sequences (together with 4 filler sequences) were presented to Is 2 subjects, who simply had to fill in a marked or unmarked pronominal device as a first word of the second sentence of each sequence. Twelve items were constructed in 32 conditions each, I6 conditions for the first sentence combined each with two conditions for the second sentence.7 In (8), four conditions of one experimental item are illustrated: two conditions of the first sentence (a and b) followed each by the two possible versions of sentence 2. (8) ai [continuing a concrete riferent] Het kleine pinnetje zorgt voor het verhogen van de druk in de cylinder. kan gemakkelijk afbreken als de luchtfilter van de motor wordt gehaald. The little pin causes the increase in pressure in the cylinder. can break easily when the air filter is removed from the motor. a2 [shifting to an abstract riferent] Het kleine pinnetje zorgt voor het verhogen van de druk in de zorgt ervoor dat de compressie van de motor grater· cylinder. wordt. __
__
__
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
associated with 'abstract' notions in a mass or a generic interpretation, we might look at inanimate physical objects as being the least 'concrete' of all types of concrete referents. Likewise, event arguments can be regarded as being the least 'abstract' of all abstract referents, when compared to e.g. speech acts or propositions. In other words, on a scale representing different types of concrete and abstract referents arranged according to their degree · of abstractness, inanimate physical objects and events could be seen as being placed next to each other. The sentences used in the experiments were presented as coming from a technical manual. In each experiment a different set of subjects participated. Subjects were asked to consider the experiment as an exercise in textual coherence. They had to imagine that they were reading a technical manual, they were asked to figure out what each discourse chunk was about, and they had to try to preserve the optimal level of coherence by opting for the most adequate anaphoric device or completion.6
Alfons Maes
217
The little pin causes the increase in pressure in the cylinder. generates · an increase of the compression of the motor. hi [continuing an abstract r4frent] Het verhogen van de druk in de cylinder wordt geregeld door het kleine pinne*. zorgt ervoor dat de compressie van de motor grater wordt. The increase in pressure in the cylinder is controlled by the little generates an increase of the compression of the motor. pin. h2 [shifting to a concrete n:forent] Het verhogen van de druk in de cylinder wordt geregeld door het kleine pinnetje. kan·gemakkelij k afbreken als de luchtfilter van de motor wordt gehaald. The increase in pressure in the cylinder is controlled by the little pin. can break easily when the air filter is removed from the motor. __
__
·
__
__
Each first sentence contained a concrete and an abstract referent. The order was initial concrete subject vs. non-initial abstract non-subject or vice versa: initial abstract subject vs. non-initial concrete non-subject. Within this basic variation, the expression type of the referents varied. Concrete referents were varied over definite and indefinite and over neuter and masculine/feminine nouns. Abstract referents took the form of a clause (9a), a neuter verbal noun identical in form to the corresponding verb (9b), or a masculine/feminine verbal noun with a nominaliiing sufftx (9c).8 The different conditions of the first sentence are given in the appendix. (9) a. clause b. neuter
N
c. mas/fern
N
dat de druk verhoogt 'that the pressure increases' het verhogen (n ) van de druk 'the increase in pressure' de verhoging ( f ) van de druk 'the increase in pressure'
The second sentence either continued the subject · referent of the first sentence as in (8a1) and (8b1), or shifted to the PP (non-subject) referent, as in (8a2) and (8b2). Alternatively, the second sentenc;e could either go on with the concrete referent, as in (8a1) and (8b2) or with the abstract referent, as in (8a2) and (8b1). The implementation of the centering variables in the experimental material is not indisputable, especially in the following two respects. Firstly, varying concrete and abstract referents over syntactic functions the way it is illustrated in (8) implies that the semantic role of abstract and concrete antecedents can differ. Although this might well be a relevant parameter, the difference will have to be accepted if we want to guarantee the different conditions to be perfectly comparable on the central centering variables. Secondly, for concn!te referents it proved .to be easy to construct second
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
__
2I 8
Referent Ontology and Centering in Discourse
=
=
=
=
Experiment I: proportion of unmarked and marked pronouns as a function of transition type (continuation-shift) and referent type (concrete-abstract)
Table 2
continuation of concrete referent n 449
abstract refere�t n 448
concrete referent n 443
abstract referent n 457
0.86 O.I 4
O. I 8
o.8 2
0.48 0.52
0.09 0.9I
=
unmarked pronoun marked pronoun
shift to
=
=
=
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
sentence predicates which unambiguously and exclusively referred to the concrete antecedent NP, as in (8ar) and (8b2). For abstract referents, however, this was much more problematic: the fact is that in (8a2) and (8br) the predicate of the second sentence most plausibly applies to the abstract antecedent NP ('the increase in pressure'), but a propositional reading in which the predicate applies to the whole antecedent sentence ('the situation that the little pin causes the increase in pressure') is not excluded. This might affect part of the results and is one of the reasons for altering the completion methods in the next two experiments (see also note ro). The results in Table 2 show first of all the basic effect predicted by Centering Theory: shifting the center to a referent invites subjects signifi cantly more often to opt for a marked pronoun than simply continuing a referent, both in the case of concrete (X2(r) r44.637, p < .oar) and abstract (X2(r) r5.03 3, p < .oar) referents. Apart from that, there is a significant interaction between expression type and referent type both in the continuation (X2(r) 408.695, p < .oar) and the shift (X2(r) r6 3.747, p < .oar) condition. Continuations of concrete referents are realized predominantly by unmarked pronouns, for shifts to concrete referents both unmarked and marked pronouns appear to be available. On the other hand, both in continuing and in shifting to an abstract referent, marked pronouns are disproportionately high. The use of different determiners in the antecedent expressions (definite vs. indefinite) did not result in significantly different proportions of marked and unmarked devices. Two of these results fall outside the predictions of centering: the high proportion of marked pronouns used to continue an abstract referent-as in (8br)-and the high proportion of unmarked pronouns used to shift to a concrete referent-as in (8b2). The first result can be seen as a psychological reflex caused by the fact that the processing of abstract anaphors requires more cognitive effort and a higher degree of inference load compared to concrete anaphors. The high proportion of unmarked pronouns used to shift. to a concrete referent might be caused by the fact that, apparently, the
Alfon s Mae s 2 I 9
=
=
=
=
=
=
Table 3 Ex pe rimen t I: proportion of abstract marke d pronouns as a functi on of ex pressi on ty pe (cl ause , ve rbal ne ute r N, ve rbal m/ f N) and transi ti on ty pe (con tin uati on -shift)
marke d pron oun
con tin uati on of an abstract re feren t whe re the an te ce dent is a
shi ft to an abstract refe ren t w here the an te ce dent i s a
clause n=I I 4
n-noun n=I I 4
clause n = I I4
o.8o
0.88
m/ f-noun n = 228
0.88
n -n oun n =II4
m/ f-noun n = 22 8
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
abstract referent is very unlikely as potential focus even if it is realized as the subject of the previous sentence. The concrete referent in non-subject position can be shifted to pretty easily by an unmarked device because it is not in competition with the abstract referent. The doze experiment cannot discriminate in an absolutely conclusive way between the transformational and the ontological hypothesis described above. However, it contains an indication in favor of the second hypothesis. The fact is that the expression type of the abstract referent was varied: abstract referents in the first sentences were realized as clauses (see 9a) or as NPs (see 9b-c), both taking up an argument position within the antecedent sentence. If the processing of abstract referents were hampered by the transformation of predicate-like structures into an entity-like concept, one would expect subjects to opt significantly more often for marked pronouns in the case of a clause-like antecedent, which is the natur.al 'packaging' for a predicate. Conversely, we would expect the noun-format, as in (9b) and (9c), to reduce the transformation penalty of abstract anaphors, because it would enhance or trigger the construction of entity-like representations, or even prevent the construction of a predicate-like representation of the ante cedent. However, as Table 3 shows, the proportions do not differ significantly, neither in the continuation condition (X2(2) o.8o9, n.s., p .667), nor in the shift condition (X2(2 ) 3·795, n.s., p . 1 49). The irrelevance of the form of the antecedent is further supported by the different proportions of 'illegal' concrete and abstract anaphors, i.e. neuter anaphors referring to non-neuter abstract and concrete antecedent nouns (as is illustrated in (ro) and ( 11 ) respectively). Subjects overwhelmingly used neuter pronouns as abstract anaphors, despite the morphologically mascu line or feminine gender of the antecedent (97% of n 6o8), whereas 'illegal' concrete anaphors were less frequent (so% of n 304). 9 This is an additional indication of the fact that abstract referents do not depend on surface form the way concrete referents do.
220
Refe rent Ontol ogy and Cen te ring in Discourse
(ro) De verhoging (£) van de druk in de cylinder wordt geregeld door het kleine pinnetje. Ze (£)/*Het (n.) zorgt ervoor dat de compressie van de motor groter wordt. The increase {f) in pressure in the cylinder is controlled by the little pin. She (f)/*It (n.) generates the compression of the motor to increase. (I I ) De pin (£) zorgt voor het verhogen van de druk in de cylinder. Ze (£)/ *It (n.) kan gemakkelijk afbreken als de luchtfilter van de motor wordt gehaald. The pin (f) causes the increase in pressure in the cylinder. She {f)/*It (n.) can break easily when the air filter is removed from the motor.
In the second experiment, I 5 9 (other) subjects had to complete the second sentence of I 2 two-sentence sequences, obligatorily beginning with an unmarked (het) or a marked (dat) anaphoric device. The first sentence contained exactly the same I6 conditions as in experiment I , the second started either with the unmarked het ('it/ she') or with the marked dat ('that'), as is illustrated in (I 2). (I2) ai The little pin causes the increase in pressure. It/The a2 The little pin causes the increase in pressure. This __
__
bi The increase in pressure is controlled by the little pin. It/ The b2 The increase in pressure is controlled by the little pin. This __
So, compared with Experiment I, the independent and dependent variables have changed places: in the second experiment, the expression type was the independent variable, whereas the continuation or shift was the dependent variable. This set-up created two additional ways of assessing the referential behavior of subjects. First, subjects were not forced to go on with the referents introduced in the first sentence: they could decide to use the trigger words het ('it/the') and dat ('that/this') either anaphorically or non anaphorically. This enabled us to investigate whether and, if so, to what extent this decision depended on the type of referent. Second, in completing the trigger word, they were given a choice between a pronomial and a nominal referential expression. The fact is that, unlike English it, Dutch het can be either a pronoun or a determiner. For Dutch dat the situation is similar to English this/that. This way, the results show a variety of marked and unmarked, non-anaphoric and anaphoric, pronominal and nominal express10ns.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
EXPERIMENT 2: C OMPLETI O N EXPERIMENT
Alfons Maes 22 I
The results show that in only about half (n· 997) of the completion sentences (n I9o8) were the trigger words het and dat anaphorically related to the concrete or abstract referent in the first sentence (52.25%). 10 The proportions of anaphoric completions of dat (n 48 I, i.e. 0.50) and het (n 5 I 6, i.e. 0.54) do not differ significantly. The proportions of het and dat distributed over the four main conditions (continuing a concrete/ abstract referent, shifting to a concrete/abstract referent) do not differ significantly either (X2 (J) 6. I 07, n.s., p . Io7). Finally, het does not show a significant interaction between the . two main variables, referent types and transition type (het: X2 ( I ) 1 .9 I 5 n.s., p . I 66); only dat shows a slightly significant interaction (X2 (J) 3 .9 1 3 n.s., p .048) , which means that the effect of referent type might be partly caused by the effect of transition type or vice versa. The most important result clearly corroborates the results of Experiment 1. Subjects significantly more often used het (o.89) and dat (o.89) to continue or to shift to a concrete referent (X2(J) 56.746, p < .oo1). They clearly avoided continuing the abstract referent or shifting to the abstract referent. Apparently, abstract referents were not considered to be valid candidate centers for subsequent discourse. Additional conclusions can be drawn from the distribution of nominal and pronominal anaphoric devices in Table 4· The total numbers of each of the different types of anaphoric expressions do not allow for valid statistical comparisons, as they all have to be regarded as proportions of the total number of anaphoric and non-anaphoric completions. Still, the very distribution itself offers additional indications as for the role of the referent type as well as the function of markedness of nominal and pronominal devices. The distribution of pronominal anaphors referring to concrete referents =
=
=
=
=
=
=
=
=
=
Experiment 2: number of pronominal and nominal anaphors (n 997) function of transition type (continuation-shift) and referent type (concrete-abstract)
Table 4
=
concrete referent 384 n
abstract referent n = 36
concrete referent n soo
abstract referent n 77
r8s
IS
187
40
=
unmarked marked nominal anaphors
unmarked marked
a
shift to
continuation of
pronominal anaphors
as
1 32 53 199
82 1 17
8 7
21
14 7
=
89 98
313
I S4 1 59
=
4
36
37
4
33
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
=
222 Re fe re nt Ontology and Ce nte ri ng i n Di scourse
EXPERIMENT
3:
C LOZE E XPERIMENT I I
Although Experiments 1 and 2 provided strong evidence for the decisive role of the ontology of referents in establishing local coherence in discourse,
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
reflects partly the prediction of centering, in that continuing a concrete referent is clearly associated with unmarked pronouns; partly it also exhibits the referent type effect because marked and unmarked pronouns are equally available whenever there is a shift to a concrete referent. The latter, again, indicates that abstract referents are not being considered as competing candidates. The few (and thus insignificant) pronouns shifting to an abstract referent are almost all marked, · as was expected. Pronouns continuing an abstract referent are so extremely rare as to be insignificant. The low number of pronouns referring to abstract referents confirms the referent type hypothesis in that it can be said to signal that abstract referents are of low accessibility and extremely unsuitable to be continued or shifted. The distribution of nominal anaphors is less clear-cut. The numbers of abstract anaphors nicely mirror the centering predictions, although in absolute terms the numbers are so low that the observation is not significant. The number of nominals (and especially marked nominals) used to continue a concrete referent is high. This makes it implausible to look at marked nominals simply as being strong identificational devices. Why would anaphors continuing a concrete referent require (marked) nominals while all other results show a strong association of the concrete continuation condition with unmarked pronominal devices. This result hints at a meaningful difference between the function of pronominal as opposed to nominal markedness. In relation to the results ofExperiment 3, I will account for this result within a non-identificational interpretation of marked nominals, as it is proposed in Maes & Noordman (1995). In sum, the distribution of the concrete anaphors provides us with two remarkable observations: (i) the (expected) asymmetry in the shift condition (nominals > pronominals) is not mirrored by a reverse symmetry in the continuation condition where pronominals and nominals balance each other out, which is not in accordance with the widely acknowledged association of pronominal anaphors and prominent referents; (ii) a standard referential hierarchy (such as the one presented by Ariel 1990) predicts a progressive decrease in the numbers of anaphors in the left column of Table 4 and a corresponding increase in the third column. The actual distribution, however, hardly reflects these predictions.1 1 I will come back to these issues in the next section.
Alfons Maes 223
some of the experimental options may have unintentionally influenced the outcome of the experiments. Therefore, I carried out a final production experiment, in which I revised some of these options. First, the doze task in Experiment I restricted the options to only two types of anaphoric expressions. The completion task in Experiment 2 provided us with a variation of anaphoric and non-anaphoric completions, as a result of which it was difficult validly to compare numbers and proportions of the different types of anaphors. In the third experiment, I combined the best of both methods. I used a doze task which enabled subjects to opt for all the relevant types of anaphoric expressions: unmarked and marked pronouns, unmarked (definite) and marked (demonstrative) nominal · NPs. Subjects were explicitly instructed about the different options, without being given information about the functional differences between them. Secondly, the experimental material in Experiments I and 2 consisted of two-sentence sequences. One introductory sentence may not have been enough for a standard implementation of continuation or shift as this might not be enough for the full .establishment of the referent. In the third experiment, I constructed sequences of three sentences with a referential structure as is indicated in ( I 3 ). ·
b
continue condition
sentence sentence
I 2
sentence 3
Referent x (initial, subject) introduced Referent x (initial, subject)-referent y (non-initial, non-subject) Referent x elaborated
shift condition
sentence sentence
I
2
sentence 3
Referent x (initial, subject) introduced Referent x (initial, subject)-referent y (non-initial, non-subject) Referent y elaborated
Thirdly, the results in the previous experiments might be affected by slight uncontrolled differences between the I2 items used, such as the specificity and the conceptual size of referents, the type of coherence relation be.tween the sentences or differences in the number of anaphoric options for concrete and abstract referents. Therefore, I revised the items and modelled them according to the following characteristics. (i) In the previous experiments, the linguistic and conceptual size of the concrete and abstract referents differed slightly in that concrete referents were introduced as a noun with one preverbal modifier (the/a little pin) whereas abstract referents consisted of a verbal entity and one or two arguments (the increase in pressure (in the cylinders)). As differences in the size of the · antecedent can influence anaphor comprehension (e.g. Murphy 1 98 5), they
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(1 3) a
224 Re ferent Ontology and Ce nte ri ng i n Di scourse
might influence the production result in the previous experiments as well. Therefore, in Experiment 3, I controlled the size of the referents: each referent consisted of a head plus one complex attribute/argument or a head plus two simple attributes/arguments, as is illustrated in { I4).
(I 4) a concrete
(ii) In a number of items in the previous experiments, the referents introduced could have a referential-specific as well as an attributive-generic reading. This might not only result in different semantic and pragmatic interpretations (see e.g. Donnellan I97 I ; Fauconnier I 985) but also in different types of anaphoric processes (see e.g. Muller-Lust & Gibbs I99I). Therefore, I opted for a generic interpretation for concrete and abstract referents: in the revised items, each referent could be paraphrased as this type of object or this type of event/process. This interpretation fitted in with the technical content and context of the experimental items and resulted in natural and comparable interpretations for concrete and abstract referents. Furthermore, as definite and indefinite antecedent NPs in the previous experiments did not differ significantly over concrete and abstract referents, I used only definite NPs to introduce concrete and abstract referents in the first sentence. (iii) In order to exclude effects based on differences in the coherence relation structure of the items, I restricted the relations between sentences to neutral elaborations. In sentence 2, a cause or explanation relation was established, as these relations offer the most natural way of combining concrete and abstract referents within one sentence. (iv) To ensure that the subjects specifically referred to the intended (abstract or concrete) referent, the antecedent NPs in sentence 2 were underlined and subjects were explicitly instructed to refer to these under lined NPs. Especially in the case of abstract completions (as in (I 5 b) and (I 5d) below) this enabled us to rule out a possible propositional inter pretation, in which the abstract anaphor refers to the VP or the whole antecedent sentence, instead of the abstract antecedent NP. This undoubtedly resulted in more valid data. But, of course, it cannot refute
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
the electronic eye on the carburettor head: eye attributes: electronic on the carburettor b abstract the increase in pressure in the cylinders mcrease head: argument: in pressure (in the cylinders)
Alfons Mae s 225
the fact that in completing abstract anaphors the presence of more or less plausible (but not intended) propositional readings might affect the subjects' behavior. All the above-mentioned characteristics are exemplified in (I s).
(I s) a continuation of a concrete referent
The electronic eye in the carburettor is an essential part of the engme. 2 The electronic eye causes the increase in pressure in the cylinders. has the form of a computer chip. 3 b continuation of an abstract referent The increase in pressure in the cylinders is essential for the 1 engine to run properly. The increase in pressure is registered by the electronic eye in 2 the carburettor. is necessary in order to provide the engine with enough 3 pressure. c shift to a concrete referent The increase in pressure m the cylinders is essential for the r engine to run properly. The . increase in pressure is registered by the electronic eye in 2 the carburettor. has the form of a computer chip. 3 d shift to an abstract referent 1 The electronic eye in the carburettor is an essential part of the engme. 2 The electronic eye causes the increase in pressure in the cylinders. . is necessary in order to provide the engine with enough 3 pressure. 1
__
__
·
__
Ex pe rime nt 3: proportion of unmarke d and marke d anaphors transi ti on type (conti nuati on-shift) and re fe re nt type (concre te -abstract)
Table S
conti nuati on of
a functi on of
shift to
concre te re fe re nt 564 n
abstract refe re nt n. 564
concre te refe rent 564 n
abstract refe rent n 564
0.5 9 0.39
0.28 o.69
0.20 0.78
0.09 0.89
=
unmarke d {pron + nom) marke d {pron + nom)
as
=
=
=
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
__
226 Referen t On tology and Centerin g in Discourse Ex perimen t 3: proportion of marked and nominal abstract an aphors as a fun ction of tran si tion type (con tin uati on-shift) and ex pressi on type of the an teceden t (clause-n oun)
Table 6
marked nominal
contin uation of an abstract referen t w here the anteceden t is
shift to an abstract referen t w here the anteceden t i s
cl ause n = r42
n oun 422 n
clause n = r4o
n oun n = 424
0.63 O.J 7
0.72 0.3 r
0.86 0.26
0.90 0.3 8
=
=
=
=
=
=
=
.
=
=
=
=
=
=
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
282 subjects received 8 items, 2 items for each of the conditions given in ( r s ). As in the previous experiments, the morphological gender of abstract . and concrete referents was varied (neuter, masculine/ feminine) as well as the form of the abstract referents (clause, neuter N, m/f N). Tables s and 6 show that the refinement of the experimental material does not alter the basic outcome of the earlier experiments. Table s replicates the results of Table 2. In line with the prediction of centering, for both abstract and concrete referents, the proportion of unmarked and marked expressions differs significantly in the two transition types (abstract referents: X2 ( r ) 66.745. p < .oar; concrete referents: X2 ( r ) 177-722, p < . oor). In line with the results of Experiments I and 2, abstract and concrete referents differ significantly: continuing abstract referents yields proportionally many more marked expressions than continuing concrete referents (X2 ( r ) I09. 1 40, p < .oa r ). Shifting to an abstract referent is significantly more pften expressed by a marked device than shifting to a concrete referent ( X2 ( r ) 27. r s r, p < .oa r ). Similarly, Table 6 replicates and extends the results of Table 3: an abstract anaphor with clausal antecedent is not realized more often with a marked or nominal anaphor than an abstract anaphor with a noun antecedent, on the contrary (marked: continuation: X2 ( I ) 4.65 3 , p < o s ; shift: X2( r ) r .2 r7, n.s., p .269; nominal: continuation: X2 ( r ) 2.076, n.s., p . 149; shift: X2 ( r ) 6. 3 19, p < .os). This result again supports the ontological rather than the transformational hypothesis on the processing of abstract referents. Finally, in two items (n 562) subjects had to refer to a masculine/ feminine abstract antecedent noun, in one item (n 282) a concrete anaphor referred to a m/f concrete antecedent noun. In these items, subjects opted for 3 3 s abstract and 1 74 concrete pronominal anaphors respectively;
Alfons Maes 227
95% of the abstract anaphors were 'illegal' neuter pronouns, as opposed to only 22% of the concrete anaphors. The set-up of Experiment 3 allows us validly to compare the proportions of the four main anaphoric types in the different conditions, as they are given in Table 7· The first thing to note is that marked pronominal and nominal anaphors behave differently, as was observed already in Experi ment 2. As far as pronominal anaphors are concerned, the referent type effect is clearly replicated. Continuing and shifting to abstract referents demands significantly more marked pronouns than continuing and shifting to concrete referents (continuation: X2( I ) 1 74.265, <.ooi; shift: X 2 ( 1 ) 3 3 ·758, <.ooi). Compared to the previous experiments, the proportion of marked pronouns used to shift to a concrete referent has however increased considerably. This is entirely due to the revised experimental material, with three instead of two sentences. Apparently, after two introductory sentences even an abstract referent is fully established and considered a valid compc:;ting candidate (see (I s b)). Hence shifting to a concrete referent requires more referential power. This, however, does not alter the fact that the simple continuation of these established abstract anaphors requires predominantly marked pronouns. The distribution of nominal anaphors in Table 7 shows a picture which is comparable to the results in Experiment 2. In all conditions, the proportion of marked, i.e. demonstrative (as opposed to definite) anaphors is extremely high, which can only mean that the distribution of marked nominals does not reflect accessibility differences of different types of referents in different transition stages. Instead, another functional explanation is called for, which exceeds a simple identificational one. =
=
continuation of concrete referent n 564
abstract referent n 564
concrete referent n = 564
abstract referent n 564
o.64
o.67
0.44
0.6;
=
pronominal anaphors
unmarked marked nominal anaphors
unmarked marked
shift to
0.77 0.23
=
0.28 0.72
0.23 0.77
0.36
0.33
o.;6
0.29 0.71
O.JO
0.19 0.8 1
0.70
=
0.07 0.93
O. J ;
0. 14 0.86
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Experiment 3: proportion of unmarked pronominal, marked pronominal, unmarked nominal and marked nominal anaphors as a function of transition type (continuation-shift) and referent type (concrete-abstract)
Table 7
228 Refe re nt Ontology and Ce nte ring in Discourse
=
·
=
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The high proportion of marked nominals might well be due to the fact that demonstrative nominal anaphors are perfectly suited to ·express a 'classifying' or member-set interpretation of discourse referents. As ( 1 5 ) shows, the NPs used refer to a (generic) specimen of a class of abstract objects or events. Demonstrative nominal NPs fit in perfectly with this classificational generic interpretation. Following Maes & Noordman ( 1995: 262), the use ofclassifying demonstrative NPs is to be regarded, not simply as an identificational tool, but as a way of modifying and adding to the referent by yielding inferences with respect to the class membership of the referent (see, for similar interpretations, Corblin 1983: 123; Kleiber 1 984; Kirsner & van Heuven 1988}. An observation which supports this exceptional status of marked nominals is that abstract referents are associated less with marked nominals than concrete referents (n 297 vs. n 392). A plausible explanation for this might be found in their ontology, in that the classificational interpretation doesn't accommodate abstract · referents quite as well as it does concrete referents. A concrete referent can be conceptualized, classified, and expressed more naturally as a member of its ontological class than an abstract referent: the one-way valve or the executablefile can naturally be expressed as members of the classes ofvalves orfiles, while the natural class ofthe increase in pressure or the control ofsystemfiles is much less readily available. So, the abstract ontology of referents coincides with-or even is partly based on-the absence or at least the lower availability of a natural class, and hence inhibits subjects from opting for a classificational marked nominal. The pronominal-nominal proportions in Table 7 partly cut across the prediction of pronominal anaphors being associated with attentionally prominent referents (continuation) and nominal anaphors with nOn prominent referents (shift). In the continuation conditions, pronominals outnumber nominals. In the shift condition, however, it is remarkable (and against all predictions) that precisely the less accessible type of referent (abstract referents) is realized predominantly by a pronominal anaphor (o.6s). whereas concrete referents slightly prefer nominals (o.s6). These 'irregularities' cannot be explained within an attentional framework with out implicating the ontological explanation discussed in the previous paragraph. Similar irregularities can be observed when we compare the proportions of the four referential types in Table 7 with the standard predictions of referential hierarchies: the four proportions do not gradually decrease in the left-hand columns or increase in the right-hand columns, which means that the standard correlation between referent accessibility and referential types does not hold. Again, the ontology of referents together with the classificational interpretation of marked nominals would seem to offer the only explanation for this.
Alfons Maes
229
C O NCLUSI O N
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The results of the production experiments show an instance of anaphor interpretation in which 'attention-based' determinants are supplemented or even overruled by 'knowledge-based' determinants. On the one hand, they support the assumption that the form of referential expressions reflects attentional fluctuation of discourse referents, since the transition stage of referents appeared to crucially define the form of referential expressions. The fact is that continuing referents resulted more often in the use of unmarked expressions than shifting to a referent. This result is congruent with centering and it confirms a large body of linguistic and empirical evidence on the association of anaphor form and referent prominence (e.g. Garrod & Sanford 1 982: 27; Reichman 1978: 3 1 1; Clark & Haviland I97T 27; Chafe 1987; Marslen-Wilson et al. 1982: 3 5 1 ; Yule 1981: 49; Francik 1985: 59; Anderson et al. 1 98 3: 433; Karmiloff-Smith 1980: 235; Linde 1 979). On the other hand, the type of referential expression . appeared to be dependent on the ontological type of referents in that-all things being equal-abstract referents require more marked (pronominal) forms in the continuation and in the shift condition and that they are much less easily continued or shifted to than concrete referents. This effect, which clearly calls for a knowledge-based explanation, is substantiated by a number of arguments. First, as the experiments focused on two referent types which can be said to be located at the borderline of the concrete/abstract distinction (i.e. inanimate physical objects and event arguments), one can safely generalize the results for the complete classes of concrete and abstract referents. Secondly, the ontological effects are clear and general. Thirdly, the effect of ontology on transition type is supplemented by other observations proving the role of referent ontology, i.e. the high number of 'illegal' abstract anaphors and the irrelevance of the clausal vs. nominal antecedent format. The results fit in with research into other knowledge-based aspects of antecedent-anaphor relations. One such aspect is the interpretation of so-called conceptual anaphors. Gernsbacher ( 199 r ) and Carreiras & Gernsbacher ( 1 992) showed that morphologically singular but semantically plural antecedents are processed more easily when they are referred to by a plural anaphor. That means that the surface characteristics of the antecedent are overruled by the semantic conceptualization of the referent. Similar evidence can be found in experimental research into processing differences between surface and deep abstract anaphors (Hankamer & Sag 1976; Sag & Hankamer 1 984). Experimental (but so far inconclusive) evidence given in Mauner, Tanenhaus, & Carlson ( 1995 ) indicates that, unlike surface abstract anaphors, deep abstract anaphors do not require
2 30 Referent Ontology and Cente ri ng i n Discourse
additional processing effort if the VP-antecedent is not a simple VP structure (e.g. x needs to feed the chickens), but if it is somewhat hidden within a passive antecedent construction (e.g. the chickens need to be fed). This again argues in favor of the irrelevance of antecedent form in processing abstract anaphors. The distribution of demonstrative pronouns in the experiments can be seen as a 'monolingual' confirmation of what Fraurud (in preparation) observes in a crosslinguistic comparison, i.e. the close association of so-called 'hypostatization anaphors' and demonstrative pronouns. In addition, it enables us to refine the definition of hypostatization and to shed light on the explanation of this association. Fraurud defines hypostatization anaphors as non-nominal second or third order entities, which involve the creation of a new discourse referent by hypostatizing the outcome of the interpretation of the antecedent (see, for a similar interpretation, Conte 1996). The results presented above suggest that the abstract nature of the referent, rather than its non-nominal form, is the crucial determinant in the 'proliferation' of demonstrative pronouns, and that the creation of a new discourse referent is not so much a matter of transforming verbal-like antecedent input into a concept-like anaphor . interpretation, but rather a matter of abstract objects obstructing a natural conceptualisation of a referent as a distinct imaginable entity in discourse. So, if we call abstract anaphors more indirect, or less strictly anaphoric or if we assume there is a more oblique interpretation process at work there, the explanation for such a characterization will have to be looked for in this ontological handicap, rather than in structural or formal characteristics. Finally, the production experiments refine the view of different types of anaphoric expressions as being hierarchized according to the degree of referent givenness they express. They question focal position as being a sufficient condition for referents to obtain a prominent givenness status (i.e. statuses like activated, in focus, textually evoked, or highly accessible). If we observe that abstractness has a continued impeding effect on the use of focal markers such as unmarked pronominals, as well as on the tendency to continue that particular referent, one has to conclude that the ontology of referents should be more salient in defining the relationship between expression type and referent status. The 'exclusive association' existing between abstract referents and marked pronominals might support a view in which the ontology of referents is considered to be a decisive factor in withholding a particular giveness status (e.g. in focus) from particular ontological types of referents (e.g. abstract referents), but we will need more empirical backing for a far-reaching conclusion like this. Analogously, the tendency for abstract referents not to be expressed as marked nominals hints ·
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Alfons Maes 2 3 I
at restricting particular types of expressions to particular ontological types of referents, independent of their givenness statuses. Acknowledgements I am ·grateful to Kari fraurud, Jeanette Gundel, francis Cornish, and two anonymous referees of this journal for their valuable comments on an earlier version of this article. I am indebted to Carel van Wijk and Hans Hoeken for their help in analysing the data. ·
ALFONS
Received: 20.06.97 final version received: 25.02.98
MAES
APPENDIX 1 6 conditions (C1 -C 1 6) of the first sentence Cr-C8 Cr-C2 Cr C2 C3-C4
CJ C4
Cs-C6 Cs C6 C7-C8 C7 cs
C9-C 1 6 C9-C r o
concrete subject neuter Noun definite indefinite neuter Noun definite indefinite neuter Noun definite indefinite m/f Noun definite indefinite abstract-subject clause (dat/of)
C9 Cro Cr r-C 1 2
CI J-C I 4
.
Cr r CI2 CIJ CI4
Cr s-C r 6 Crs Cr6
neuter verbal N definite indefinite m/f verbal N definite indefmite m/f verbal N definite indefmite
abstract non-subject clause (dat/of/om te + inf) neuter verbal N definite indefinite m/f verbal N definite indefinite m/f verbal N definite indefinite concrete non-subject neuter N definite indefinite neuter Noun definite indefinite neuter Noun definite indefinite m/f Noun definite indefinite
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Discourse Studies Group Tilburg University PO Box 90 153 5000 LE Tilburg The Netherlands e-mail:
[email protected]
2 32 Referent Ontology and Centering in Discourse
N OTES
8
9
IO
II
not receive all the conditions of the experimental material ( I2 items vs. 32 conditions), the main variables (con tinuation vs. shift, abstract vs. concrete) were nevertheless represented equally often for each subject. Dutch differs from English in that physi cal object nouns, as well as abstract nouns can either be neuter or masculine/ feminine. So, the use of masculine/ feminine antecedent nouns enabled us to inve;tigate the extent to which sub jects opted for either the morphological gender of the antecedent noun (result ing in a masculine/feminine anaphor) or the 'semantic gender' of the referent (resulting in a neuter anaphor). The ontology of referents might well affect this preference. See, for the relevance of the semantic vs. morphological gender and for different positions on the topic, Garnham & Oakhill (I990), Bosch (I987), Gernsbacher (199I), Carreiras & Gernsbacher (1992), and Tasmowski & Verluyten (r98 5). The overall high proportion of illegal anaphors was partly due to the experi mental task, as subjects were asked to use het ('it') or dat ('that'), or if necessary, another pronominal device. In analysing the results, I only took into account anaphoric completions which unambiguously referred to either the concrete or the abstract referent. The majority of the remaining completions were anaphoricalfy related as well, in that they could refer to either the concrete or the abstract referent, to the combination of both, to the whole verb phrase, or to the whole proposition of the antecedent sentence. These remaining completions contain interesting data for further ana lysis, yet I left them aside for present purposes. In Ariel ( I990), the order of unmarked and marked norninals is reversed. But this would not make things sig�ficantly different.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The examples are more o r less literal (and hence not always elegant) trans lations of Dutch sentences used in the experiments. 2 The corpus consists of (i) a compilation of 22 popular scientific texts of soo woo words each (number of words rs,ooo), (ii) a compilation of 3 7 book reviews, TV reviews, autobiographical columns, and personal commentaries from the magazine Vrij Nederland (number of words 28,961 ), and {iii) a compilation of I 8 leaflets (number of words 28,424). 3 I will use the terms marked and unmarked as convenient labels for demonstrative and non-demonstrative expressions respectively: for Dutch this means that the demonstrative pronouns dit 'this', dat 'that' (both neuter, sing.), and deze 'this/these', die 'that/those' (both masc./fe�. sing. and masc./fem./ neuter plural) together with their nomi nal counterparts are labeled as marked, whereas the personal pronouns hij 'he', zij 'she', het 'it' together with the defi nite NPs de N 'the N' (masc./fem. sing. and masc./fem./neuter plural) and het N 'the N' (neuter sing.) are labeled as unmarked. 4 In confining myself to the local level of discourse, I remain aware of the fact that global and structural discourse characteristics can affect the attentional state of referents as well. s This sketchy presentation leaves many details as well as interesting variations and amendments of the basic centering mechanism out of consideration, as e.g. Walker, Lida, & Cote (1994), Brennan, Friedman, & Pollard {1987), Kehler {1997). These, however, do not affect the variables I am focusing on here. 6 The experiments took place during a Dutch language proficiency course in a Translation High School as part of a regular exercise in textual coherence. 7 Although each individual subject did
Alfons Maes 2 3 3
REFERE NCES
'
Conte, M.-E. (I996), 'Anaphoric encapsu lation', Belgian Journal of Linguistics, IO, I - I O. Corblin, F. (I983), 'Defini et demonstratif dans Ia reprise immediate', Le Franfais Moderne, S I , I I 8-J4. Donnellan, K. {I97 I ), 'Reference and defi nite descriptions'. In D. D. Steinberg & L. A Jakobovits (eds), Semantics, Cam bridge University Press, Cambridge, IOO-I 4. Fauconnier, G. (I98s), Mental Spaces:Aspects of Meaning Construction in Natural Lan guage, MIT Press, Cambridge/London. Fletcher, C. R. { I984), 'Markedness and topic continuity in discourse processing'. journal of Verbal Learning and Verbal Behavior, 23, 487-93. Francik, E. P. (I98 s), 'Referential choice and focus of attention in narratives', unpub lished doctoral dissertation, Stanford University. Fraurud, Kari (I992), 'Situation reference: what does 'it' refer to?' GAP Work ing Paper, 24, Fachbereich Informatik, Universitat Hamburg. Fraurud, K. (I996), 'Cognitive ontology and NP form', in Th. Fretheim & J. K. Gundel (eds), Reference and Referent Accessibility, John Benjamins, Amsterdam, 193-212. Fraurud, K. (in preparation), 'Indirect ana phora in a cross-linguistic perspective'. Garcia, E. (I996), 'Measured coherence', Belgian Journal of Linguistics, 10, I 39-60. Garnham, A & OakhiU, J. {I990), 'Mental models as contexts for interpreting texts: implications from studies of anaphora', journal of Semantics, 7, 379-93· Garrod, S. & Sanford, A. J. {I977), 'Inter preting anaphoric relations: the inte gration of semantic information while reading', journal of Verbal Learning and Verbal Behavior, 16, 77-90. Garrod, S. C. & Sanford, A J. (I982), The mental representation of discourse in a
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Anderson, A, Garrod, S. C., & Sanford, A J. { I983), The accessibility of pro nominal antecedents as a function of episode shifts in narrative text', Quarterly journal of Experimental Psychology, 3S. 427-40. Ariel, M. ( I 990), Accessing Noun-Phrase Ante cedents, Routledge, London/New York. Ariel, M. ( I 994); 'Interpreting anaphotic expressions: a cognitive versus a prag matic approach , journal ofLinguistics, 30, 3-42. Asher, N. { I993), Reference to Abstract Objects in Discourse, Kluwer Academic Publishers, Dordrecht. Bosch, P. (I987), 'Pronouns under control? A reply to Liliane Tasmowski and Paul Verluyten', journal ofSemantics, s. 65-78. Bosch, P. (I988), 'Representing and access ing focussed referents', Language and Cognitive Processes, 3, 207- 3 1 . Bosch, P. & Geurts, B . (I990), 'Processing definite NPs', Rivista di Linguistica, 2, I, I77-99· Brennan, S. E. (I99S). 'Centering attention in discourse', Language and Cognitive Processes, IO, 2, I 37-67. Brennan, S. E., Friedman, M. W:, & Pollard, C. J. (I987), 'A centering approach to pronouns', in Proceedings of the zsth Annual Meeiing of the Association for Computational Linguistics, I s s-62. Carreiras, M. & Gernsbacher, M. A (I992), 'Comprehending conceptual anaphors in Spanish', Language and Cognitive Processes, 7. 314. 28 I-99 · Chafe, W. L. {I987), 'Cognitive constraints on · information flow', in R. S. Tomlin (ed.), Coherence and Grounding in Discourse, Benjamins, Amsterdam, 2I-)2. Clark, H. H., & Haviland, S. E. (I977). 'Comprehension and the given-new contract', in R. 0. Freedle (ed.), Dis course Production and Comprehension, Ablex Publishing Co., Norwood, NJ, I-40.
234 Referent Ontology and Centering in Discourse position in modem Dutch', Lingua, 76, 209-48. Kehler, A (I997), 'Current theories of cen tering for pronoun interpretation: a criti cal evaluation', Computational Linguistics, 2J, 3 . 467-75· Kleiber, G. (1984), 'Sur Ia semantique des descriptions definies', Linguisticae Investi gationes, 8, 63-85. Linde, Ch. (I 979), 'Focus of attention and the choice of pronouns in discourse', in T. Givon (ed.), Discourse and Syntax, Academic Press, New York, 3 37-54. Maes, A. (I996a), Nominal Anaphors, Marked ness and the Coherence ofDiscourse, Peeters, Leuven. Maes, A (I 996b), The markedness of abstract-object anaphors in discourse', Belgian journal of Linguistics, 10, I6I-8J. Maes, A & Noordman, L. G. M. {I99S). 'Demonstrative nominal anaphors: a case of non-identificational markedness', Linguistics, J 3, 2, 25 s-82. Marslen-Wilson, W., Levy, E., & Komisarjevsky Tyler, L. (I 982), 'Pro ducing interpretable discourse: the establishment and maintenance of refer ence', in R. J. Jarvella & W. Klein (eds), Speech, Place and Action: Studies in Deixis and Related Topics, John Wiley & Sons Ltd., Chichester, 3 39-78. Mauner, G., Tanenhaus, M. K., & Carlson, G. N. (I995), 'A note on parallelism effects in processing deep and surface verb-phrase anaphora', Language and Cognitive Processes, 10, I , 1-12. Muller-Lust, R A G. & Gibbs, R. W. {I99 I), 'Inferring the interpretation of attributive and referential definite descriptions', Discourse Processes, 14, 107-J l. Murphy, G. L. (1984). 'Establishing and accessing referents in discourse', Memory & Cognition, 12, s. 489-97. Murphy, G. L. (198 5), 'Processes of under standing anaphora', Journal of Memory and Language, 24, 290-303. Prince, E. F. (I 98 I), Towards a taxonomy of given-new information', in P. Cole (ed.),
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
focussed memory system: implications for the interpretation of anaphoric noun phrases', journal of Semantics, 1 , 21-41. Gernsbacher, M. A (I 989), 'Mechanisms that improve referential access', Cognition, 32, 99- I )6. Gernsbacher, M. A. (199 I), 'Comprehend ing conceptual anaphors', Language and Cognitive Processes, 6, 8 I-IO). Givon, T. (I 992), The grammar ofreferential coherence as mental processing instruc tions', Linguistics, JO, S-S 5 · Gordon, P. C., Grosz, B. J., & Gilliom, L. A (I993). 'Pronouns, names, and centering of attention in discourse', Cognitive Science, 17, 3 I I-47· Gordon, P. C., & Scearce, K. (1995), 'Pro nominalization and discourse coherence, discourse structure and pronoun inter pretation', Memory and Cognition, 23, 3, 3I 3-2J. Grosz, B. J., Joshi, A K., & Weinstein, S. (I995), 'Centering: a framework for modelling the local coherence of dis course', Computational Linguistics, 21, 2, 203-25. Gundel, J. K., Hedberg, N. & Zacharski, R (I99J), 'Cognitive status and the form of referring expressions in discourse', . Language, 69, 2, 274-307. Hankamer, J., & Sag, I. (1976), 'Deep and surface anaphora', Linguistic Inquiry, 7, 391-428. Hellman, C., & Dahl, 0. (I994), 'Situational anaphora and the structure of discourse', in P. Bosch & R. van der Sandt (eds), Focus and Natural Language Processing, IBM, Heidelberg, 3 , 467-73. Karmiloff-Srnith, A { I98o), . 'Psychological processes underlying pronominalization and non-pronominalization in children's connected discourse'. In J. Kreiman & A E. Ojeda (eds), Papers from the Para session on Pronouns and Anaphora, CLS, Chicago, 2 J I -SO. Kirsner, R S. & Heuven, V. J. van (I988), 'The significance of demonstrative
Alfons Maes 2 3 5
••
Tomlin, R. S. ( 1987). 'Linguistic reflec . tions of cognitive events', in R. S. Tomlin (ed.) Coherence and Grounding in Discourse, Benjamins, Amsterdam, 45 s-so. Walker, M. A, Lida, M., & Cote, S. ( 1994), Japanese discourse and the process of centering', Computational Linguistics, 20, 2, 19 J-2J2. Webber, B. L. ( 1991 ), 'Structure and osten sion in the interpretation of discourse deixis', Language and Cognitive Processes, 6, 2, 107-35· Yule, G. ( 198 I ). 'New, current and displaced entity reference', Lingua, 5 5 , 4 1 - 5 2.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Radical Pragmatics, Academic Press, New York, 223- 5 5· . Reichman, R. ( 1978). 'Conversational coherence', Cognitive Science, 2, 2 8 3:-327· Romijn, K. ( 1996), 'Hoe doen we het? Verwijzen naar linguistische en cog nitieve representaties van het voomaam woord "het" doctoral dissertation, Universiteit van Amsterdam. Sag, I. & Hankamer, ]. ( 1984) , 'Toward a theory of anaphoric processing', Linguis tics and Philosophy, 7, 325-345. Tasmowski, L., & Verluyten, S. P. ( 1985 ). 'Control mechanisms of anaphora', journal of Semantics,. 4, 341-70.
Joumnl oJSnnantics
14= 237-309
© Oxford University Press
1 997
A Modular Approach to the Grammar of Additive Particles: the Case of German Auch MARGA REIS A N D INGER ROSENGREN Universitiit Tubingen and Lunds Universitet
Abstract In this paper we give a modular account of the grammar of additive particles. In doing this
possible pattern: the particle preceding the main stressed constituent it relates to (its
Additive particles, however, occur in a second, equally unmarked pattern: the
RC}. . RC
preceding the main stressed particle. Former accounts do not only miss this comple mentary distribution as to position and stress pattern relative to the
RC,
but, as we
demonstrate in detail, they misrepresent the relation between syncax. semantics and focus
structure of these (and similar) particles in general. Using German
just one
Auch
as our prime example, we argue in particular (i) that there is
underlying
Auch
the
± stressed
distribution cannot be explained
by a
variants,
and
that the complementary
analysis; (ii) that the
set
of
operation applies to the material it contains, no matter whether it is the
RC
or
p and some contextually given proposition q induced by Auch belong to, is not supplied by the focus structure of p but by comparing p and q; (iii) that the syntactic scope of Auch is crucial for its semantics in that the adding alternatives the
Auch
proposition
movement
predicative material common to ± stressed
Auch
Auch
p, q; (iv) that the complementary distribution of
follows from the modular interaction of the syntax and semantics of
with focus structure; (v) that
Auch
gives rise to two utterance meanings, 'in
addition/furthermore' and 'likewise', directly correlating with whether or not the scope
of
Auch
contains
RC
material. What we argue, in short, is that so-called 'focus particles'
are in reality 'scope particles'.
1
I NTRODUCTI O N •
The by now accepted term for elements like English only, also, even, or their German equivalents nur, auch, sogar is 'focus particles' (or 'focus adverbs'). The conspicuous way they interact with focus structure is exemplified in ( I ), and has been roughly described like this: the respective focus particle enters a specific relation to the focused constituent (its 'related constitu ent' = RC), in our case Maria, indicating that Maria is the only one (nur), an additional one (auch) or the least expected one (sogar) of a set of candidates to fulftl the proposition in question, in our case: 'Peter wird x besuchen' ('Peter will visit x').
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
we take issue with the standard descriptions of focus particles, which are based on just one
2 3 8 Additive Particles: the Case of Gennan Auch
( I ) a. Peter wird nur MaRla besuchen.
2
b. Peter wird auch MaRia besuchen. c. Peter wird sogar MaRia besuchen. Peter will {only/also/even} Maria visit Peter will visit {only/a�o/even} Maria.
(2) a. *peter wird Maria NUR besuchen. b. Peter wird Maria AUCH besuchen. Whereas restrictive particles like nur typically stick to the pattern in ( 1a), disallowing non-contrastive stressed variants, especially to the right (2a), plain additive particles like Auch3 occur equally unmarked to the left as well as to the right of the RC, exhibiting the following accentuation pattern: if to the left of the RC the RC bears main stress and the additive particle does not (I b); if to its right it is the other way around (2b), main stressed additive particles being no more contrastive or marked than their unstressed counterparts. In other words, unstressed vs. stressed additive particles are in complementary distribution with respect to position and stress of the RC: whichever comes later, the additive particle or the RC, will bear the nuclear accent (NA). As is well known, the regularities just stated and illustrated by German examples hold for restrictive vs. additive particles in a great many languages (see Konig 199Ia). This does not mean, however, that they are always as clearly visible as they are in German; c£ the English examples in ( 3 ): (3) a. b. c. d. e. £
Only PEter will visit Mary. Peter will also visit MAry. *Also PEter will visit Mary Peter, TOO, will visit Mary. *Too PEter will visit Mary. Peter will not visit Mary EITHer.
*Peter will ONly visit Mary. Peter will ALso visit Mary. *peter will visit Mary ALso. Peter will visit Mary, TOO.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Particles of this kind are found in practically all languages (see Konig I 991a). In recent years, an enormous amount of linguistic effort has gone into explicating their status and function, and as the term 'focus particle' indicates, focus structure has been a constitutive part of the explication. Strangely enough, however, there are only scattered remarks on the focus behaviour of these particles themselves, although many of them-as is well known-have stressed as well as unstressed variants. A moment's reflection shows that this cannot easily be integrated into the standard explications, which are tailored to the pattern illustrated in (I), namely 'unstressed particle > stressed RC'. Moreover, not all particles have stressed variants under the same conditions; c£ (I) vs. (2):
Marga Reis and Inger Rosengren
239
- that there is indeed just one Auch underlying the ± stressed occurrences of this form; - that the complementary distribution of ± stressed Auch fragmentarily illustrated in (rb)-(2b) follows from the semantics of Auch interacting
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
While only, just like German nur, visibly behaves as expected (3a), the picture offered by additive particles looks more complicated: unlike German, English has two primary additive particles: also and too, and these are subject to various, not always complementary restrictions: also appears stressed as well as unstressed, following its RC in the former case and preceding it in the latter; but, again unlike German Auch, there are syntactic restrictions as to its position relative to the RC and the status of the RC itself (3 b, c). As for too, this particle is always stressed, but its occurrence is restricted to a few specific positions, the unmarked position being at the end of the clause (3d, e). In this position we also find the particle either, which is always stressed, taking over the additive job in negative clauses (3f). In spite of these restrictions and lexical variations, the main regularity holds in English, too: unstressed (also) and stressed occurrences of particles (ALso, TOO, EITHer} are in complementary distribution with respect to position and stress of the RC. But the fact that more than one lexical item partakes in it obscures this distribution. Nor is it easy to keep the specific semantics of the various lexical items under control. It is the aim of this paper to isolate and describe the properties of additive particles in such a way that their complementary behaviour can be accounted for. In doing this, we shall take issue with the standard approach to focus particles, which is exclusively based on the pattern underlying ( 1 }: it does not only miss the generalization governing the stress behaviour of additive particles relative to their RC, but-as we shall demonstrate in · detail-it also fails properly to account for the relation between syntax, semantics and focus structure of these particles in general. In pursuing our aim, our object of investigation will be German Auch. The crucial advantage of this is the following: whereas in many other languages-for example, in English as we have seen in (3)-the job of additive particles is divided between several lexical items, German Auch has access to all positions open for additive particles, thus exhibiting the central complementary distribution in full and undisturbed from possible lexical differences. Moreover, given the identity of form, Auch has at least a very good chance of being the same word throughout. We hope that our investigation of German Auch will not only lead to a better understanding of the specifics of additive particles but also lay the basis for a more consistent account of focus particle behaviour in general. The hypotheses we shall argue for in particular, are
240 Additive Particles: the Case of German Auch
To put it in a nutshell: we shall argue, using Auch as our prime example, that the so-called 'focus particles' · are in reality 'scope particles'. The . paper is organized as follows: in section 2, we will describe the complementary distribution of the ± stressed variants of Auch, and determine what kind of co-constituents both occur with, arguing that all our findings support the hypothesis that there is only one Auch. We will then show that Auch is a base-generated left-adjunct to XPs and sketch its syntax in GB-oriented terms. In section 3, we will discuss what we consider shortcomings of the previous semantic analyses of Auch · and motivate a different approach allowing us to explain the relation between the syntax and semantics as well as the behaviour of both variants of Auch on a unified basis. In section 4, we will discuss the interaction between the syntax, semantics and focus structure of the Auch clause as well as the interaction of this whole complex with various forms of information structure, and derive the complementary distribution worked out in section 2. Finally, we will show that ± stressed Auch give rise to two utterance meanings, 'in addition/ furthermore' and 'likewise'. Section s contains a brief summary.
2
THE S YNTAX OF A UCH
In section 1 we presented examples showing that unstressed and stressed Auch display equally unmarked behaviour. In this section we shall show (i) that they are just variants of one and the same Auch that are in complementary distribution with respect to the RC and the NA (as part of the focus); (ii) that Auch itself is an XP adjoining to XPs of all kinds; (iii)
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
in a strictly modular way with its syntactic distribution and focus structure; - that the division of the Auch proposition into RC and the rest of the proposition is not determined by its focus structure (as previously assumed), but by comparing it with some contextually given proposition in the same set of alternatives; - that, instead, the syntactic division of the Auch proposition into material inside vs. outside the syntactic scope of Auch plays the crucial role for its semantics: what Auch adds to the corresponding part of a proposition in the same set of alternatives is not the RC (as previously assumed), but the material in its syntactic scope, which may be either the RC or the identical predicative material common to both propositions; - that in principle Auch gives rise to two utterance meanings, 'in addition/ furthermore' and 'likewise', which depend on whether or not the syntactic scope of Auch contains RC material.
Marga Reis and Inger Rosengren 241
2. 1 auch/AUCH
in complementary distribution
In this section we shall look at the distribution of auch/Auch relative to overt material. 2. 1 . 1 The AC is part of the clause
Let us begin with a few typical cases showing that auch/AUCH are in complementary distribution with respect to their ACs and the position of the NA or focus respectively. already mentioned in ( 1 ), unstressed auch typically precedes the whole AC, which then invariably bears the NA (4):-(s); the material to the left of auch is ID material. 2. 1 . 1 . 1 The whole AC is to the left or right of Auch. As
(4)
Auch PEter hat das Buch geleseiL Also Peter has the book read Peter, too, has read the book. b. Peter hat das Buch auch geLEsen. Peter has also read the book. c. Peter hat auch das BUCH gelesen. Peter has also read the book. d. Peter hat auch dem Paul ein BUCH gekaufi. Peter also bought a book for Paul. a.
(nicht nur PA UL)5 (not only Paul) (nicht nur geKAUFT) (not only bought it) (nicht nur die RezenSION) (not only the review) (nicht nur dir das ESsen bezahlt) (not only treated you to the meal)
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
that the comple.mentary distribution of its variants cannot be accounted for by deriving A UCH structures from auch structures, both being equally 'basic'; (iv) that the syntactically relevant notion is 'co-constituent (= CC)' of Auch, i.e. the XP . Auch adjoins to, rather than 'RC'. Finally, we shall describe the various syntactic Auch configurations in GB-oriented terms. Before we start, a brief remark on terminology. In the following the RC of Auch will be called AC (= 'added constituent'). The AC is the 'variable material' in the proposition of the Auch clause compared with some other proposition q in context, the rest the material common to both propositions, the ID (= identical material). The ID is normally contained within the same proposition as the AC, representing what the Auch clause has in common with this other proposition.4 {For details see section 3.) We shall see in this section that AC as well as ID enter interesting relationships with the syntax of Auch.
242 Additive Particles: the Case of German Auch
(s) *PEter (auch) hat (auch} das Buch gelesen. Peter (also) has (also} the book read
Unstressed auch is usually adjacent to its AC (4), but there are also cases of optional non-adjacency (6). They seem to be somewhat degraded, but they certainly occur all the time, at least in spoken language.6 (6) a. Peter hat auch gestern
b.
With A UCH, matters are just the other way around: main stressed AUCH typically follows its AC (7}-(8}, having ID material to its right. Again, as (7) shows, there is adjacency as well as non-adjacency between the AC and AUCH. (7) a. Peter hat AUCH das Buch sofort gelesen. Peter has also the book immediately read Peter also read the book immediately. b. Peter hat das Buch AUCH sofort gelesen. Peter also read the book immediately. {8} a. *Feter hat AUCH das Buch sofort gelesen. b. *Feter hat AUCH das Buch sofort gelesen.
In other words, regarding position relative to the AC as well as stress (focus) behaviour, auch and AUCH are in complementary distribution. Note that this also includes the distribution of secondary stress, for not only the AC preceding AUCH may, optionally, ·have one (9), but also the auch preceding its AC ( w):
(9) {Peter/PEter} hat (AUCH) das {Buch/BUCH} (AUCH) sofort gelesen.
(1o) a. Peter hat {auch/AUCH} das BUCH gelesen. b. Peter hat das Buch {auch/AUCH} sofort geLEsen. c. Ja, { auch/AUCH} mit DIEsem Fall hatten wir Probleme. with this case had we problems Yes, also Yes, also with this case we had problems.
The specific functions of secondary stress on Auch will be discussed in section 4.2. Here, it suffices to notice that it may be assigned. Thus, the complementary distribution of auch/AUCH boils down to a very simple regularity:
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
dem Paul ein BUCH gekauft. (nicht nur eine CD} Peter has also yesterday the Paul a book bought Yesterday, Peter also bought a book for Paul. (not only a CD} Peter hat auch das Buch geLEsen. (nicht nur geKAUFT) (not only bought it} Peter has also read the book.
Marga Reis and Inger Rosengren 243
( I I ) The last element in the Auch/AC pair must carry the NA, the first element may carry a secondary accent.
7
Let us now check this regularity more carefully against cases where an overt AC constituent is moved to the left of Auch and leaves a trace behind in its base position. We see immediately that in constellations resulting from scrambling, topicalization or other instances of wh-movement it is the overt position of the AC constituents that counts. This is especially clear in cases like (r2), where Auch, stressed or not, is in the same syntactic position, the AC being optionally to the right (base position) or to the left (movement position}:
This shows that (I I) pertains only to the overt distribution irrespective of its ongm. There is, however, one apparent exception to the above generalization: the finite verb, moved to the left by verb movement, may figure as the AC of auch (IJa, b), but not of A UCH (IJc}: ( IJ) a. daB Peter das RefeRAT auch TIPPte that Peter the report also typed that Peter also typed the report b. Peter TIPPtei das Referat auch ti . (Er korriGIERte; es nicht nur ti.) Peter also typed the report. (He did not only correct it.) c. *Feter tippte das Referat AUCH ti. But on closer inspection this 'exception' is to be expected, for verb movement is always reconstructed,8 no matter whether or not the verb is stressed, whereas scrambling and wh-movement types are not. Hence, the position of the trace is the relevant position here. Thus, (1 3) also conforms to the basic regularity formulated above, and the complementarity of auch/AUCH is again confirmed.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(12) a. Sie wollte auch Peter HEiraten. She wanted also Peter marry She also wanted to marry Peter. b. [Peter HEiraten]j wolltei sie AUCH tj ti. Marrying Peter is what she wanted, too. c. *[Peter HEiraten]j wolltei sie auch tj ti.
244 Additive Particles: the Case of German Auch 2. 1 . 1 . 2 Split
In the examples given in the previous section the ID material was frequently split up, yielding the following combinations of the AC with ID material: cases.
( 14) a. (ID) auch (ID) AC
( r s) a. Gerd war verargert, weil Dr. No nicht nur einen KolLEgen colleague Gerd was annoyed because Dr. No not only a zu Rate zog, sondern Luise auch IMPFte. to advice drew but Luise also vaccinated Gerd was annoyed, because Dr. No not only asked a colleague for advice but also gave Luise an injection. b. Gerd war verargert, weil Dr. No nicht nur Luise IMPFte, Gerd was annoyed because Dr. No not only Luise vaccinated sondern ihre Krankengeschichte auch seinem KolLEgen but her medical . history also his colleague anvertraute. confided Gerd was annoyed, because Dr. No not only gave Luise an injection, but also told her medical history to a colleague of his. As the nicht
nur continuations show, the object preceding auch always belongs to the AC, thus constituting partial exceptions to the ordering regularity auch > AC as stated. According to Jacobs ( 1 983: 1 16) these exceptions are licensed by the preposed AC-parts being c-construable ('thematic') and
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(ID) b. (ID) AC (ID) AUCH (ID) These combinations reflect the complementary distribution of auch/A UCH in that overt mixture of iD/AC material occurs only to the right of auch and to the left of AUCH. But they also show a potential asymmetry between auch and AUCH in what occurs to the right: auch has always AC material to the right (overtly or reconstructed in the case of the finite verb) but also allows ID material there, whereas the overt material to the right of A UCH is always ID material. (Traces of constituents moved to the left and figuring as AC are ·of course possible; see ( r 2) above, but not relevant here.) In this section we shall look at cases in which the AC is split, one overt part being to the right, the other to the left of Auch. This presupposes an AC consisting of constituents that can be syntactically separated, which more or less boils down to the AC being a V projection. What we will show is not only how splits fit into the complementary picture of auch/AUCH, but also that the asymmetry concerning the right hand side of auch/A UCH is systematic: overt material to the right of AUCH is always ID material. Turning first to auch, split examples like (I s) have been cited and discussed by Jacobs (1983: 1 14-1 17):9
Marga Reis and Inger Rosengren 24 S
consequently non-accented ('de-accented'). But while this is typical, it is not the rule; see ( 1 4), where einem Blinden is clearly rhematic:
(r7) Gerd hat heute ABgewaschen, und einem Blinden (*AUCH) iiber die
Strafie (*AUCH) geholfen. In apparent contrast, cases like (r8) are totally acceptable, although, at first glance, the AC seems to be split: (r8) Paul hat dem Vater nicht geschrieben, und er hat die Mutter Paul has the father not written and he has the mother AUCH vergessen. also forgotten Paul didn't write to his father and he also forgot his mother. Why is this so? The answer is that what is to the right of AUCH may be regarded as ID material, for not to write to someone may be equated with forgetting someone in the given co-text. In other words, in this interpretation (r8) is no split case at all-the AC (Vater vs. Mutter) is totally to the left of AUCH-and only if interpreted in this way, will (r8) be acceptable. Summarizing then, the asymmetry noted above is systematic: overt material to the right of AUCH must be ID material. This finding is also confirmed by looking at semi-acceptable cases like (r9):
(19) Paul brachte eine 5 nach Hause. ??Und zu allem Ungliick hat er Paul brought a 5 to home and to all bad luck has he seine {Uhr/UHR} AUCH verschlampt. his watch also mislaid Paul got a D and on top of this he managed to lose his watch, too.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
( 1 6) Gerd hat heute nicht nur ABgewaschen, sondern einem Blinden auch a blind also Gerd has today not only off-washed but iiber die STRASse geholfen. across the street helped Today Gerd didn't only do the dishes, but he also helped a blind person across the street If so, the only regularity governing these exceptions is· that the preposed AC part cannot bear the NA, in other words: the AC part carrying the NA must always be to the right of auch. Note that the AC part carrying the NA must be the constituent that would normally carry the NA, if auch were not part of the clause. This fact shows conclusively that the regularities in question do not pertain to Auch in relation to the AC as such, but to the AC constituent potentially bearing the NA Turning now to AUCH, we do not find comparable splits; c£ the AUCH counterpart of (r6), which is out no matter where AUCH is placed:
246 Additive Particles: the Case of German Auch
2. I .2 Propositional constituents as ACs
is to be expected, auch and AUCH behave in principle in the same way when the AC is the entire proposition as they do when the AC is only part ofit.
As
2. 1 .2. 1 Unsplit cases. As
for auch, the above regularities make us expect that it occurs to the left of the entire propositional AC, if possible. Syntax provides two ways to realize this option: auch may be adjoined to the CP, 1 0 see the 'initial Auch' case (2oa)' ' as well as (21a, b), or to the VP denoting the whole proposition (2ob, 21c): (2o) a. [Der Fernseher ist kaputt.] Auch hat ein Student mein AUto [The TV set is broken.] Also has a student my car angefahren. [Dies alles hat nicht gerade meine Laune verbessert.] hit. [This all has not exactly my mood improved.] [ The TV is broken.] Moreover, a student hit my car. [All this didn't exactly improve my -mood.] b. (Der Fernseher ist kaputt.] Es hat auch ein Student mein AUto angefahren. It has also a student my car hit. [Dies alles hat nicht gerade meine Laune verbessert.] [ . . . ] Moreover, a student hit my car. [ . . . ] (21) a. [Der Fernseher ist kaputt.] Auch dafi ein Student mein AUto Also that a student my car angefahren hat, [hat nicht gerade meine Laune verbessert.] has hit
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
They are not good, but considerably better than (1 7). The main reason for this is that, while 'eine s nach Hause bringen' and 'seine Uhr verschlampen' cannot possibly be identified, context will at least help us to identify an intrinsic common denominator (things going wrong and/or things usually making people unhappy), making the two propositions reasonably similar (which means that they appear as instances of the same type of event). Note that the part of the AC that would carry the NA, if Aucli did not carry it, stays to the left of A UCH (whereas in split cases involving auch it must stay to the right). To conclude, we have seen once more that (i) auch/AUCH are in complementary distribution with respect to the AC constituent able to carry the NA, (ii) Auch can only bear the NA if in the actual co-text no part to its right can bear it. What we have seen in addition is that AUCH differs from auch in not allowing AC splits.
Marga Reis and Inger Rosengren 247
[ . . . ] The fact that a student hit my car didn't exactly improve my mood. b. [DaB der Fernseher kaputt ist und] auch dafi ein Student mein A Uto angefahren hat, [hat nicht gerade meine Laune verbessert.] c. [DaB der Fernseher kaputt ist und] daB auch ein Student mein AUto angefahren hat, [hat nicht gerade meine Laune verbessert.]
2. 1 .2.2 Split
cases. As expected, auch may also turn up within a propositional AC, provided it is to the left of the constituent that would carry the NA if auch were not part of the clause. C£ (22), the counterpart to ( r6):
(22) Es hat nicht nur Gerd heute ABgewaschen, sondern (auch) Peter It has not only Gerd today off-washed but also Peter (auch) einem Blinden (auch) iiber die STRASse geholfen. blind also across the street helped also a Today Gerd didn't only do the dishes, but Peter also helped a blind person across the street. for AUCH, we also find what we expect: no true splits, but only counterparts to the apparent 'split' case (r 8). In sum, only auch may take propositional ACs in simple clauses. The AC may be entirely to its right or split up, provided the constituent that would carry the NA, if auch were not part of the clause, stays to its right.
As
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Neither (2oa, b) nor (2r a, b, c) are quite equivalent. But in either case the · whole proposition can be the AC only if there is a common denominator the ID-for the propositions connected by auch, in this case that the events are equally annoying. Note that the co-text (dies alles hat meine Laune nicht verbessert) lends a helping hand in identifying the 10, but this is not obligatory. What is obligatory, however, in these cases, · is that the proposition(s) the auch clause is related to immediately precede it (the reason being, presumably, that otherwise the common denominator acting as the ID for propositional ACs would be too hard · to identify). As for AUCH, we do not find any simple AUCH clauses with unsplit propositional ACs, for the same reasons, of course, we did not find split cases with AUCH: overt AC-material to the right of A UCH is disallowed. We shall return to this fact below.'2
248 Additive Particles: the Case of German Auch 2. 1 . 3 Summary
2.2
Can AUCH clauses be derived from
auch
clauses?
Let us now turn to the question, how the complementary distribution described in section 2.1 is to be accounted for. We have already indicated that we take the relative order of Auch and (the NA part oQ the AC at face value and that we will derive the different stress patterns by independent focus principles. The approach in the previous literature, however, has been more or less the other way around: since Auch is commonly taken to be a 'focus particle', i.e. a particle selecting for a focus (the AC) in its scope, the assumption. is almost inevitable that the complementary distribution of auch and A UCH with respect to NA and AC position results from some kind of leftward AC-movement (combined with Auch picking up the focus accent).'3 In this section, we shall show that this assumption is untenable, no matter whether it is implemented by a straightforward AC movement analysis (section 2.2.1 ), or by a more sophisticated AC-bound variable analysis (section 2.2.2). 2.2. I Can
A UCH clauses
be derived from
auch
clauses by
AC movement?
The following arguments show that a straightforward movement analysis would be quite ad hoc, and in some respects impossible:
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
We have seen in section 2.1 . that auch and AUCH are indeed in complementary distribution at surface structure: auch requires that the AC is (totally or partly) to its right and that it (or its rightward part) carries the NA, AUCH requires that the AC is (totally) to its left, ruling out a further accent to its right. Applying structuralistic reasoning to these findings, the complementary distribution suggests that auch and AUCH are indeed the same Auch, and the way in which they are complementary A uch or (a part oQ the AC, whichever comes last, carries the NA-suggests that this distribution is related to very general focus regularities (c£ section 4). There is, however, a remarkable asymmetry in this complementary distribution in that only auch allows split and propositional ACs, whereas AUCH does neither, which is based on the two systematic regularities worked out in this section: (i) auch requires to its right at least the AC constituent potentially bearing the NA; (ii) A UCH disallows overt AC material to its right.
Marga Reis and Inger Rosengren
249
(a) The ACs to the left of A UCH may be in a base-generated position. The most notable case in point is illustrated in (23):
Peter es ihm AUCH zeigen wollte. (23) weil ja because yes[MP) Peter it him also show wanted because Peter wanted to show it to him, too. The standard assumption in this case (well supported by the evidence) is that the subject must be in its base position, since it is below the modal particle {MP) and above weak pronouns that have remained in the VP (see Lenerz 1993, cf also Ormelius-Sandblom 1997).
While (24a) might be possibly explained away by assuming ellipsis of the preverbal constituent to happen after movement, this explanation does not work in (24b) and {24c), for which we will have to assume base-generated pro/PRO elements,14 for why should these elements move at all before Spell out? A second case is provided by wh phrases, which are perfectly good ACs for AUCH (2sa), but impossible as right-adjacent ACs for auch (2sb):1 5 (25) a. [Ich stand vor dem Eingang.] Und wer stand da plotzlich [I stood at the entrance.] And who stood there suddenly AUCH? also [I stood at the entrance] and who was there, too, all of a sudden? b. *Auch WER stand da plotzlich?/ *WoHIN ist [auch WER] Also who stood there suddenly/ Wheretois also who gegangen? gone
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(b) Some types of ACs are possible for AUCH, but not for auch: a first case (cited already by Kowalski 1992: 92ffj are non-overt elements like those in (24): (24) a. preverbal ellipsis: Hab ich AUCH schon erledigt. Have I also already finished I've already finished this, too. Kommt bitte pro AUCH. b. imperative: Come please also Please come, too. c. infinitival subject: Er bat ihn, PRO AUCH zu kommen. 1-:le asked him also to come He asked him to also come. (! # Er bat ihn AUCH, zu kommen.) (He asked him, too, to come.)
250 Additive Particles: the Case of German Auch
(c) Configurations where the AC is excluded from the scope of AUCH: a case where the exclusion is syntactically motivated is (26) (see Kowalski 1992: 9 1), where AUCH is inside a predicate phrase (26a) which cannot possibly host its AC, the overt subject Petra. Hence, (26a) has no grammatical source, which would have to be (26b). (26) a. Petra gilt als AUCH sehr ehrgeizig. very ambitious Petra is supposed as also Petra is supposed to be very ambitious, too. als auch PEtra sehr ehrgeizig. b. *Es gilt It is supposed as also Petra very ambitious.
(27)
[Peter ist 1 .80 groG.] [Peter is 1 .80 metres tall.] a. Von seinen Freunden ist nur Uohannes/JoHANnes} AUCH Of his friends is only Johannes also r.8o groK 1.80 tall Among his friends only Johannes is also 1 .80 metres tall. b. *Von seinen Freunden ist auch nur JoHANnes 1.80 groB.
Whereas (27a) has the sensible meaning Johannes is also 1.80 (with the implicature that there is someone else of the same height, here instantiated by Peter) and nobody but Johannes is also 1.8o', (27b) would have to mean 'nobody but Johannes is 1.80 and nobody but x =/:- Johannes is 1.8o', which is a glaring contradiction. In other words: the movement analysis implies not only reversal of relative scope, but also that auch and nur in (27) apply to the same set of alternatives, which in view of the contrary meanings of these particles is impossible. (d) Crucially different behaviour of post-posed nur/sogar, which come closest to Auch in that they also occur after their AC, and with which 'post-posed' A UCH is often treated on a par (see e.g. Jacobs 1 98 3: 101ff): (28) PEter {nur/sogar/*auch} hat einen Aufsatz geschrieben.'6 Peter {only/even/ also} has an essay written {Only/even} Peter wrote an essay. (29) a. PEter hat {nur/ sogar /*auch} ein Buch verfaik Peter has {only/ even / also} a book written {Only/even} Peter wrote a book.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
A clear case where the exclusion is semantically motivated is (27) (also modelled after an example from Kowalski 1992: 105f).
Marga Reis and Inger Rosengren 25 I
Contrary to auch, (i) adjacent post-position of nur/sogar is possible, c£ (28), (ii) leftward movement of the AC is possible, but only to the preverbal position, c£ (29)-(3o); (iii) even if to the left of nur/sogar, the AC must always bear the NA, c£ (28)-(29); (iv) if in post-position, nur/sogar may never be stressed (although at least nur allows for main stress in principle), c£ (3 r ). Moreover, distant post-position of nur and sogar as in (29) is quite restricted, and apparently incompatible with non-scalar meaning (see also Altmann 1 976:199). Neither is the case for A UCH in (3 2). Now, given (i)-(ii), the nur/sogar cases allow for a straightforward movement explanation: the AC may left-adjoin to the XP it is part of, and from this escape-hatch position it is optionally moved to the preverbal position by the familiar process of topicalization. (That the results are marked is a general phenomenon with extractions from DPs.) If so, the AC must obviously take its main stress along, which may not even optionally be transferred to the particle left behind (iii)-(iv). But then it is also oqvious that the NA on 'post-posed' A UCH cannot be derived in the same way. (e) Problems with extraction and focus theory, showing that a non- ad hoc movement account covering all occurrences of AUCH simply cannot be . given: First and foremost, how is this movement to be motivated? According to recent theory, movement should have a reason, for example the necessity to get features checked or at least some last resort reason. But what could be the reason for moving the AC out of the c-command domain of Auch? The . only idea that comes to mind is Kowalski's (1992) 'topic' proposal (see section 2.2.2), but, as we will show, it is empirically untenable. Second, this movement would have to have properties that are totally without parallel: to begin with, there is no evidence that the AC of auch may move to the escape hatch position of the auch-XP it is part of, see (28),
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
b. GeTIPPT hat Peter das Referat {nur/ sogar/*auch}. Typed has Peter the report {only/ even / also} Peter {only/even} typed the report. . (3o) ?* weil ihn PEter gestern {nur/ sogar} besuchte. because him Peter yesterday {only/ even} visited (3 r ) a. *{Peter/PEter} wird NUR kommen. Peter will only come b. Es wird NUR Peter kommen. It will only Peter come Only Peter will come. (32) Peter kommt AUCH. Peter comes too Peter is coming, too.
252 Additive Particles: the Case of German Auch
2.2.2 Can
AUCH cases
be derived by assuming an AC-bound variable in the scope of auch?
Let us now look at a more sophisticated version of this analysis proposed by Kowalski (1 992). Her idea is that A UCH always 'focuses' a semantic variable (i.e. the variable underlying the AC part} in its scope that may be bound from apparent AC material outside, arid she tries to build a case for it in roughly the following way: (i} A UCH cases with invisible ACs (imperative and infinitival subjects, for example) are taken to be cases where the AC has no syntactic reflex, hence is just a semantic variable (i.e. the variable in the partition, see section 3 below}; (ii} we can assume the variable part to be invisibly in the scope of Auch, if the following restrictions obtain: a) if the variable has no syntactic reflex, A UCH c-commands the base position of the element the variable in question is an argument of, b) if overt ACs to the left of A UCH bind the variable in question, they act as topics, which motivates their appearance to the left; (iii} it is possible to give (iia) and (iib) an interpretation such that the assumption of the variable part being invisibly in the scope of Auch is non ad hoc. While her proposal (and our representation of it} is certainly not fully worked out, the outlines are clear enough to argue against it. First, the basis of the case is not sound: all the invisible elements acting as ACs of A UCH can be shown to be syntactically 'real' elements of syntax (PRO for example)-with no evidence against their being to the left of A UCH-and those that are undoubtedly not projected into syntax (e.g. 'implicit arguments' like the goal argument in Er teilt Uemandem] seinen
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
which would make extraction possible. Furthermore, extraction from DPs, which is presupposed by the movement analysis wherever the AC ofAUCH is a DP, is heavily constrained and marked (see Muller 1995; Pafel 1995a), whereas the putative extraction from auch phrases would have to be totally unconstrained, producing unmarked results only. Last but not least, a non- ad hoc description covering all the focus data cited in (d) seems impossible. Note that the most plausible revision of the movement analysis-taking non-adjacent auch cases as source sentences, with the putative AC of AUCH thus moving out of a VP-will not really help: while the extraction problems could be circumvented this way, the problems cited in (a)-(d) would remain, and the focus problem be aggravated: why should auch pick up the focus accent left behind by a non-adjacent phrase at all? The upshot of this is that a non-ad hoc AC movement analysis is not available, let alone a unified one.
Marga Reis and Inger Rosengren 2 5 3
Riicktritt mit 'He announces [somebody] his resignation') may not act as ACs of AUCH. Hence, all occurrences of AUCH have real ACs to their left, which makes the claim that in reality A UCH always 'focuses' to the right, . implausible. Second, contrary to (iib) the real ACs to the left of A UCH are not always topics: invisible ACs {PRO for example) never are topics, and visible ACs may appear beside a full-fledged topic, c£ {3 3) (where Peter is the AC and a . Gauguin is the topic): ,
.
Peter possesses a Gauguin, too (but he doesn't have other impressionists.) Furthermore, the purported AC restrictions on referentiality and specificity cited in favour of the . topic analysis can be shown to result from straightforward positional restrictions auch and A UCH have in common (see section 2.J.I). Third, the role focus plays in relation to stress and syntactic structure becomes totally obscure: what could it mean for a variable without syntactic reflex to be 'focused', and how would we account then for the stress patterns of A UCH clauses? In particular, why is Auch always stressed, if the AC is to its left? Kowalski seems to assume that its position behind the AC as such is responsible (p. 79), but this means stipulating what should be explained on the basis of general principles of focus and stress assignment. Note finally that 'weak' pronouns like the accusative es and its cliticized form 's may act as ACs of A UCH:'7 (34) [Das erste Proj ekt fordert ihr, hab ich gehort. Und wie habt ihr The first project support you, have I heard. And how have you Euch mit dem zweiten Projekt entschieden?]-Na ja, wir yourselves with the second project decided -Well yes, we {fordern es/fordern's} halt AUCH. support it MP too [You'll support the first project, I've heard. And what did you decide on the second one?]-wen we'll support it, too. Given the fact that these elements may never act as focused constituents, not even in contexts where corresponding nuclear or secondary stress is not assigned (e.g. second instance cases like PETer hat nur das erste Projekt/*nur es/
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(33) [Mensch, Paul besitzt einen GauGUIN!] [Boy, Paul possesses a Gauguin.]Einen GauGUIN besitzt Peter AUCH [, aber ihm fehlen andere Impressionisten.] Gauguin possesses Peter also [ . . ] A
254 Additive Particles: the Case of German Auch
*nur)s gefordert (see also Hoeksema & Zwarts 1 99 1 : 67; Krifka 1995; GeilfuB Wolfgang 1996: r4f), this is a final touch of evidence that the AC has no intrinsic relation to focus assignment. In sum, Kowalski's proposal does not work either. 2.2. 3 Summary
2. 3
The CCs
of Auch
In the beginning of section 2, we already introduced the notion CC, which is of course nothing else but the c-command domain of Auch. In this section we want to clarify the exact syntactic nature of this CC, thus preparing the ground for a consistent description not only of the syntax of Auch but also of its semantics and its focus behaviour. We shall first argue that Auch does not take heads as CCs (section 2.J.I), then determine what kind of XPs it takes as CCs (section 2.3.2). In the course of the discussion, it will become evident that the regularities formulated in the previous sections (see section 2.1.3) in terms of 'to the right' vs. 'to the left' of Auch can all be restated in terms of 'inside the CC' vs. 'outside the CC' of Auch. 2. 3 . 1 Can the CC be just a head?
Let us begin with disposing of the idea that Auch may take heads as CCs.'8 In this regard, the crucial case is (3 s), where the CC o f auch/AUCH seems to be a head constituent, the finite verb: (3 s) daB Peter das Referat auch TIPPte/ AUCH tippte that Peter the report also typed/ also typed that Peter also typed the report/ that Peter typed the report) too However, if the finite verb as such were the CC of auch/AUC� the resulting constituent should also be a V0 category, and it should be possible, in fact mandatory for auch/AUCH to be moved together with the verh But this is strictly ruled out:
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
There is no non -ad hoc analysis that allows us to derive the A UCH cases from auch cases. Hence, the complementary distribution stated in section 2.1 must be accounted for in another way. We will try to show in section 4 that it is an epiphenomenon resulting from the interaction between syntax, semantics, and independent focus principles.
Marga Reis and Inger Rosengren 2 5 5
(36) a. *peter [auch TIPPte]; das Referat t;. b. *peter [AUCH tippte]; da5 Referat ti. Peter also typed the report
(37) a. daB er den Teller {Auch leer iBt /*leer Auch iBt that he the plate {also empty eats /empty also eats /*leer gegessen Auch hat} / empty eaten also has} that he also {eats/ate everything} on the plate b. wei! . er {Auch kommen wiirde/*kommen Auch wiirde} because he {also come would/ ·come also would} because he would come, too sie {Auch klug ist/fahrt II *klug c. weil because she {also intelligent is/drives II intelligent Auch ist/fahrt} also is/drives} because she {is also intelligent/is also an intelligent driver} {Auch nicht kam/*nicht Auch kam20} d. als er Ieider when he unfortunately {also not came/not also came} when, unfortunately, he didn't come either e. wenn Peter {Auch noch kommt/*noch Auch kommt} when Peter {also still comes I still also comes} if Peter will come too The generalization covering these cases is that Auch, no matter in which variant, must be in a position above the VP that comprises not only the final verbal complex, but also its modifying elements including certain types of adverbial particles. While this VP can be very complex, see (3 8), clearly, all its elements except the governing verb are optional. (38) weil Peter sich [VP (Auch) noch (*Auch) nicht (*Auch) because Peter himself (also) still (also) not (also)
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Since Auch + CC has the same distribution in all other comparable cases as the CC by itself (see section 2.4), the only explanation is that the CC in (3 s ) is not the finite verb as such, but something else. We suggest, in keeping with our views on VP structure (see section 2.4.2), that it is the VP immediately dominating the finite verb plus its closest modifiers (inter alia verb particles, directional elements, adjectival parts of resultative constructions, certain types of 'adverbial' particles, among them presum ably nicht 'not', noch 'still', if there are any). There is excellent evidence that this is indeed correct, c£ the positional restrictions of Auch (no matter whether stressed or not) with respect to these elements illustrated in (3 7a-e):'9
256 Additive Particles: the Case of German Auch
gut (*Auch) zurecht (*Auch) gefunden (*Auch) zu haben (*Auch) scheint) . good (also) right (also) found (also) to have (also) seems because Peter does not yet seem to have found his way Hence, the VP in question may just consist of the bare finite verb, which covers cases like (37). If so, we can conclude that Auch does not take heads. Since there is no evidence that Auch takes other nonmaximal projections, we arrive at the generalization that Auch takes only XPs as CCs.
Although we believe that there is only one Auch, which is sometimes unstressed and sometimes carries the NA, it is useful to keep the two variants apart when investigating what kinds ofXPs they may take. The prediction is, ofcourse, that as mere variants they take the same ones. In apparent contrast to this prediction, we will observe an asymmetry in the kinds of XPs figuring as the CCs of auch vs. A UCH, but we will also argue that this asymmetry is the result of interaction with independent regularities (see section 2.5). Thus, the hypothesis that there is just one Auch is not affected. Let us begin with auch. One of the strongest descriptive generalizations of German syntax is the V2 constraint: with a few exceptions well under control, the finite verb in so-called V2 clauses may be preceded by one XP constituent only. If used as a test for constituency, it shows that unstressed auch takes all kinds of XPs-DP, PP, VP, AP, CP-as CCs:2' (38) a. [Auch PEter)i hat ti einen Aufsatz geschrieben. Also Peter has an essay written Peter, too, wrote an essay. b. [Auch den NEUen Aufsatz)i finden nur wenige ti gut. Also the new essay find only few good The new essay too, is liked only by Jew people. c. [Auch an einem A UFsatz)i hat Peter ti gearbeitet. Also at an essay has Peter worked Peter has also been working on an essay. d. [Auch dem Paul ein BUCH gekaufi)i hat Peter ti. Also the Paul a book bought has Peter Peter also bought a book for Paul. e. [Auch fleillig geLEsen)i hat Peter ti. Also diligent read has Peter Peter has also read eagerly.
DP DP PP VP
VP
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
2 . 3 .2 Can the CC of Auch be any XP?
Marga Reis and Inger Rosengren 2 57 AP
AP
CP CP
Since topicalization has applied in all of these cases (except, of course, (3 8h) ), we may assume that auch plus the XPs in question also form a constituent in their base position within the VP. This is confirmed by the fact that auch + XPs of various kinds may also appear to the left of modal particles (39): since according to standard assumptions modal particles top the highest base-generated VP, the auch + XP in question must have been moved out of the VP, and movement applies of course to constituents only. ·
(39) Dann hat auch {PETra/ IHR} halt einer von uns den Weg gezeigt. Then has also {Petra/ her} MP one of us the way showed Then one of us showed the way to {Petra/her}, too. We may conclude then, that auch takes all kinds of lexical XPs as CCs, i.e. all kinds of X:Ps except functional A-projections.22 In the cases just cited auch invariably takes immediate constituents on the verbal projection line of a full-fledged clause. However, it may also appear within clause-like projections (4o)-(41) which are parts of immediate constituents on the verbal projection line: (4o) a. Der [mit sich auch sehr UNzufriedene] Professor ging weg. professor went away The with himself also very un-content The professor who was very displeased with himself, too, went away. b. Der [auch [mit SICH] sehr unzufriedene] Professor ging weg. (41) a. Der Professor, [auch DAmit unzufrieden], ging weg. The professor, also that-with un-content, went away The professor, displeased with this too, went away. b. Der Professor, [nun auch sich SELBST ein Ratsel], ging weg. The professor, now also him (dat.] self a puzzle, went away. The professor, now a puzzle also to himself, went away.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
( [Auch sehr LIEB]; war er plotzlich t;. Als� very sweet was he suddenly All of a sudden he was also very sweet. g. [Auch EHR.geiziger als andere]; ist sie t;. Also ambitious-er than others is she She is also more ambitious than others. h. Auch hat Peter einen A UFsatz geschrieben. Also has Peter an essay written Peter wrote an essay too. 1. [Auch daB PEter einen Aufsatz schreibt,]i wuBten alle ti. knew all Also that Peter an . essay writes That Peter was writing an essay, was also known to everybody.
258 Additive Panicles: the Case of German Auch
c. Peter, nun auch mit PAUL ·im Clinch, wurde noch Peter, now also with Paul in fight, became even miGtrauischer. distrustful-er Peter, now quarrelling with Paul too, became even more distrustful.
(42) a. Der (PROi mit sichi auch sehr UNzufriedene) Professori ging weg. b. Der Professori (PROi nun auch sichi SELBST ein Ratsel) ging weg. If so, we can view the auch occurrences in (40)-(41) in the same way as those in (3 8)-(39): auch takes as its CCs immediate constituents oii a predicate projection line in either case.23 Let us now turn to A UCH. Since we consider AUCH to be just the stressed variant of Auch, we would expect it to take the same XPs as CCs as the unstressed variant auch. But there is an important restriction, as can be seen from (43)-(44): (43) A. Paul hat der Johanna den Aufsatz gezeigt. Paul showed Johanna the essay. B. a. Peter hati der Johanna (AUCH den Aufsatz gezeigt ti]. b. (AUCH der Johanna den Aufsatz gezeigt1 hati PEter ej ti Peter has the Johanna[dat.) also the essay shown Peter showed Johanna the essay too. c. Peter hati der Johanna den Aufsatz [AUCH gezeigt ti)· d. (AUCH gezeigt ti]j hati Peter der Johanna den Aufsatz ej . Peter showed Johanna the essay too. e. Peter zeigtei der Johanna den Aufsatz (AUCH ti]. Peter showed Johanna the essay too. (44) A. Paul hat Johanna auf den Aufsatz gestern aufmerksam gemacht. Paul has Johanna on the essay yesterday attentive made Paul drew Johanna's attention to the essay yesterday. B. a. *(AUCH Johanna)i hat Peter ei auf den Aufsatz aufmerksam gemacht. ·
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The generalization is, clearly, that auch can take XPs in all clause-like projections, i.e. where there is a predicate projection line: in the case of (4oa, b) and (41a) the predication is related to a visible head, the adjective, whereas in (41 b, c) there is an invisible head only. That there is indeed a predicate projection line comparable to those in full-fledged clauses, is shown by reflexive cases like (4oa, b) and (4 1b) above: since they demand a subject antecedent in the governing category, the respective structures obviously contain a PRO co-indexed with Professor (23a, b), which makes these constructions clause-like, in other words, there is an (extended) projection of a predicate head.
Marga Reis and Inger Rosengren 259
b. *[AUCH auf den Aufsatz); hat Peter Johanna e; aufmerksam gemacht. c. *[AUCH gestern); hat Peter Johanna auf den Aufsatz e; aufmerksam gemacht. (43) shows that A UCH takes VPs of all sizes as its CCs,24 in accordance with what we observed so far. But as (44) shows, it does not take DPs, PPs, AdvPs, etc. as its CCs. We can make sure that this holds irrespective of whether only the material in the CC or all the material to the right of A UCH is ID material by looking at cases like (45):
Here the AC is entirely to the left of AUCH, and the position before 'the modal particle allows only the given bracketing (due to the same reasons cited in connection with (39) above). Since constellations like (45) are always out, if Auch is stressed, the only possible cause can be that A UCH combines with a 'wrong' CC. Thus, AUCH is distributionally far more restricted than auch, a reflex of this being that A UCH is always an immediate constituent on the predicate projection line. However, this restriction will turn out to be epiphenomenal, as we shall see later (sections 2.5, 4.2). Since AUCH does take VPs, we expect that it also occurs in other clause-like constructions, and it does; c£ (46): (46) a. Der {damit/DAmit} AUCH sehr unzufriedene Professor The that-with also very un-content professor gmg weg. went away The professor who was quite displeased with this, too, went away. 25 b. Der Professor, [nun sich SELBST AUCH ein Ratsel), ging weg. The professor, now him [dat.) self also a puzzle, went away. The professor, now a puzzle also to himself, went away. c. Peter, nun mit Paul AUCH im Clinch, wurde noch Peter, now with Paul also in fight, became even mifitrauischer. distrustful-er Peter, now quarrelling with Paul too, became even more distrustful. ·
Recall (see above (42)) that these constructions contain a PRO co-indexed with Professor in (46a, b) and Peter in (46c). Given this, the possible ACs of AUCH are all to its left as usual. The possible CCs are again restricted to •.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
wird [AUCH Peter) ja helfen. (45) *Der johanna The Johamia[dat.) will also Peter MP help
260 Additive Particles: the Case of German Auch
projections of the predicate. Hence, A UCH behaves in these constructions just as it does in full-fledged clauses.26 In sum, Auch, unstressed as well as stressed, takes immediate constituents on predicate projection lines as its CCs, with the stressed variant being restricted (by intervening regularities) to the immediate predicate projections on these lines themselves. Now, the following example (47) shows that Auch takes only immediate constituents on predicate projection lines. Thus, inside DPs, PPs, etc. occurrences of Auch like the following are out, no matter whether Auch is stressed or not. Note that the other focus particles, plus nicht, exhibit more or less parallel behaviour.
These occurrence restrictions have been extensively discussed, ever since Jacobs brought them up (1983: 86f£V7 We account for them as the result of a specific syntactic property of Auch ·and focus particles in general, which is to bring about a division of its host clause (corresponding to the proposition it applies to) into the CC arid the rest (the part of the clause not c-commanded by Auch) with the following properties: just one projection line is split, and the resulting parts are categorially homogeneous in that the Auch phrase (i.e. the phrase resulting from adjoining Auch to its CC) must be immediately dominated by projections of the same category as the rest of the clause. It is easy to see that (i) and (ii) can only be fulfilled, if the Auch phrase is an immediate constituent on a predicate projection line, thus dividing up its host proposition. The relevance of this division will show up later (see sections 2.s and 3.2). Here it may suffice to just state the empirical generalization as a division rule, by which cases like (47) are out: (48) Auch divides its host proposition such that the Auch phrase is always an immediate constituent on the projection line of the predicate.
Note that (48) also pertains to other elements having clausal semantic scope like Auch (see section J.I ), e.g. kaum 'hardly', selten 'rarely', immer 'always', in that they do not occur within DPs, PPs, etc. either.28 Moreover, (48) provides a basis for understanding the apparent exceptions to it, inasmuch as they are systematic: Auch within DPs is the more acceptable, the more clause-like the DP is (in particular the more it corresponds to a proposition).29 Hence, in the absence of convincing altematives,30 there is reason enough to consider our approach to this restriction worth pursuing.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(47) Sie vermutete (0pden (*Auch) PaB (*Auch) Pauls] oP She supposed the (also) passport (also) Paul's (ppin (*Auch) dem Koffer]PP· in (also) the suitcase
Marga Reis and Inger Rosengren 261 2. 3 . 3 Summary
We have shown the following: (i} the CC of Auch is always an XP (never a head); (ii) Auch may take all kinds of XPs as its CC; (iii} Auch divides the clause or clause-like constructions it appears in into two parts such that the Auch phrase ('Auch + CC') is always an immediate constituent on ·the predicate projection line. It was also noted that Auch can only be stressed, if . its CC is a predicate projection.
The syntactic status of Auch and its positions in clause structure 2.4
2.4. I The syntactic status of Auch
In section 2.3 we determined the possible CCs of Auch, thereby confirming that there is only one Auch. We will now have to determine the status of Auch itsel£ There are two possibilities: it may be a head or an XP. We will begin with the first alternative. Given the fact that Auch always takes XPs as CCs, there are two options: Auch may be a minor functional head in the sense of Rothstein (199 1 : 7) or it may be the head of a functional projection. The former position has been adopted by Bayer (1996), who assumes that Auch is a small functional head, which is subcategorized for XPs and projects only semantic head features, but not the syntactic ones: (49)
XP
�
y
XP
It is evident that categorial transparency is needed in the case of the particles in question. In (49) this is achieved by stipulating that minor functional heads do. not project category features. But we could get the same result Without stipulation, if Auch were just a maximal (non expanding) projection adjoining to XPs. So there should be more to the minor functional head analysis than just that. All we can see, however, are further disadvantages: if Auch were a functional head: (a) it should take just one type of XP-complement (see Abney 1987; Rothstein 1 99 1). But as we have seen above, it obviously takes all sorts of them;
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Let us now turn to discussing the syntactic status of Auch itself, and then explicate the relation between Auch and its CC in GB-oriented terms.
262 Additive Particles: the Case of German Auch
(b) it would block the possibility of the verb to move to functional projections above Auch (c£ the NegP in French, see inter alia Pollock 1 989). There is no evidence for any blocking of this type; (c) the Auch occupying the initial field could not be accounted for. Instead, we would have to assume that this is a separate Auch, an assumption that is not supported by the evidence (see note 12). All this taken together rules out this version of the head hypothesis. Let us now look at the other version, where Auch is a functional head, also projecting its category features, and takes an XP as its complement: (so)
YP
XP
I
Auch
In this case, we would have to assume that the YP allows the verbal features to extend through it (the extended projection idea proposed by Grimshaw 1 991). This, however, would be no real problem, since we have to assume this for other functional projections, e.g. the NegP in English, too. But the problems (a)/(c) would still remain. Since the various head analyses fail, let us adopt the simple analysis already alluded to above: Auch is a maximal {non-expanding) phrase ( s 1a) adjoining to XPs of all kinds (5 1 b). By a non-expanding maximal projection we mean a head that projects to a maximal phrase in much the same way as sentential adverbs like Ieider 'unfortunately', doch 'however', etc. ( s 1) AuchP b. XP a.
I
Auch0
�
AuchP
XP
2.4.2 The positions of Auch in clause structure
The empirical generalization stated in section 2.3.2 as a division rule (48 ) was that Auch occurs with all lexical XPs as its CCs, i.e. all XPs lexical items can take as long as the (resulting) Auch phrases are immediate constituents on the projection line of a predicate. This rules out at least all functional A-projections on the extended projection line, but, as we have already seen, it does include A-bar projections like CPs. We also tacitly assumed that Auch does not move itsel£ We do not know of any evidence contradicting this assumption. (On the contrary, the modal
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
�
Y
Marga Reis and Inger Rosengren 263
(s z)
CP
� � � � �
SpecC AuchP
C'
CompUfinite V
FP
SpecF
F'
F0
VP
AuchP
VP
-------
DP
�
AuchP
VP.
DP
�
VP
I
yo
AuchP
AP
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
particles with which particles like Auch may be compared, do not move either, see Ormelius-Sandblom 1 997: 43f). The fact that Auch never moves itself, does, of course, not preclude that other constituents to the right of Auch may move to its left, be it by scrambling, topicalization or some other type of wh-movement, whenever such a movement is licit (which, as a rule, it is for constituents within VP, but not within DP/PP). The same movements are possible for Auch phrases. This amounts to the general ization that in every position in which a lexical phrase XPi is licit (by movement or base generation), the phrase [Auch (XPi] ] is also licit, subject to the division rule (48). This has been confirmed over and over again in the previous pages. we shall return to one apparent exception to this generalization below. The following tree (which partially conforms to the clause structure proposed by Haider 1993 for SOV-languages) illustrates four theoretically possible positions of Auch. There are, of course, many more.
264 Additive Particles: the Case of German Auch
The account we have given so far covers all the well-known occurrence restrictions for Auch but one, the positional asymmetry of Auch CPs as exemplified in (5 3): a complement clause with an adjoined Auch cannot be extraposed, but it may well occur in the initial field of the matrix clause. hat Peter behauptet. (5 3) a. Auch [daB der Kanzler zu DICK sei], Also that the chancellor too fat be[subj.] has Peter claimed Peter also claimed that the chancellor is too fat. b. *Feter hat behauptet, auch daB der Kanzler zu DICK sei.
(54) a. Auch hat ein Student dein AUto angefahren, glaube ich. Also has a student your car on-driven believe I A student also hit your car, I think. b. *Ich glaube, auch hat ein Student dein AUto angefahren. I think also has a student your car hit This shows that Auch may not take V2-CPs either, but this is what we would expect, so far. Second, and more importantly, a subclass of adverbial clauses occurs in extraposed position without problems (see also Altmann 1976: 16off): (55)
Ich habe darauf verzichtet, auch weil ich MODE war. I have that-on refrained, also because I tired was I refrained from doing it, also because I was tired. b. Ich habe mich gut unterhalten, auch wahrend Peter WEG war. I have me well entertained also during Peter away was I had a good time also while Peter was away. a.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
This asymmetry, first observed by Altmann (1978: 22ff), has usually been treated on a par with the occurrence restrictions covered by our division rule (48). Accordingly,Jacobs (1983: 46ff) tried to relate the difference to his claim that Auch takes only V projections (bound up with the claim that there is no V2 constraint in German), with Auch taking a V projection in (5 3a) but not in (5 3b), and Bayer (1996) tried to relate it to ± canonical direction of government: in (5 3a) the verb governs the auch phrases to the left, which is the canonical direction, whereas in (5 3 b) it does so to the right. Both accounts have met with strong criticism (see note 30), which we will not repeat here. Rather, what we should like to point out is that this restriction is different from the others and thus should have no impact on the general analysis by which the others are covered. 3 ' First, the following clause shows that a comparable positional restriction also holds for V2 clauses. The clause-initial construction (54a) is good,. the extraposed construction (54b) is out:
Marga Reis and Inger Rosengren 265
This, we would not expect. These data show that Auch cannot be primarily responsible for the distributional difference in ( S J ). . Third, elements like doch 'however', iibrigens 'by the way' exhibit parallel behaviour to Auch: das Referat the report getippt, typed
The doch in the initi�l position has about the same meaning as aber 'but', but cannot appear iri the middle field. Its function is to contrast the meaning of the proposition against another proposition in its co-text. Thus, the doch clause always has to be in a position such that it may be linked directly to the preceding clause. This seems to be the relevant parallel to the Auch clause. Thus, the ungrammaticality of (53 b) may be based on a more general restriction including other particles with similar functions as well. What (ss) tells us is at least that the sharp ungrammaticality of (s 3b) cannot be due to Auch alone: there must be some additional factor ruling out extraposed Auch + dafl/V2 clauses but leaving Auch + weil/wiihrend clauses unaffected. On the other hand, the comparison with doch may give rise to the idea that the acceptability of Auch + CPs in the initial field is due to the fact that in this position Auch directly links its clause to the matrix clause (note that extraposition makes subordinating constructions more like paratactic constructions), very much functioning as a conjunction like und. If this is the primary function Auch has when taking CPs as its CCs, then it cannot occur after the matrix clause, in which position the linking could not take place. In any case, the restriction in question seems peculiar enough to set it apart and give it a separate analysis. 2.4. 3 Summary
In this section, we have argued that Auch itself is not a head, but an XP adjoining to XPs of all kinds, except functional A projections. We also assumed that Auch does not move itsel£ We then described the behaviour of Auch in terms of the clause structure proposed for German in Haider
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(s6) *Ich glaube, doch hat Peter seinem Freund I think however has Peter his friend [dat.) getippt. typed (57) Doch hat Peter seinem Freund das Referat However has Peter his friend [dat.) the report glaube ich. think l Peter typed the report for his friend, I think.
266 Additive Panicles: the Case of German Auch
(1993). Finally, we commented on the absence of extraposed Auch + dafl clauses suggesting that this restriction has no impact on our analysis.
2.5
The CC and the AC/ID division
(s 8) If the CC ofAuch is a nonpredicative projection (DP, PP, AdvP, etc), then it always includes all33 A C material. Ijthe CC ofAuch is a predicative projection, then it may (a) include overt AC material, or (b) it may not; in the latter case it contains only overt ID material or is empty. Obviously, then, the regularity we observed in section 2. 1 . 1.1 -the last element in the Auch/AC pair has to carry the NA-reflects a regularity about AC/ID distribution with respect to the CC ofAuch, yielding a further generalization concerning NA placement: (59) If the CC ofAuch includes (overt) AC material, it gets the NA, if it does not, Auch gets it. Note that the restriction observed in section 2.3.2 that AUCH only occurs with predicate projections as CCs, which also means that the Auch in question is always an immediate constituent of the predicate projection line, follows immediately from (s8)-(s9), for CCs not containing overt AC
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Let us now turn to the relation between the syntactic division of the Auch clause into CC and the rest and its division into AC and ID. In previous accounts this question was reduced to the relation between CC and AC, the communis opinio being that the AC is always included in the CC, i.e. c commanded by Auch (a constraint posited for focus particles in general already by Jackendoff 1 972: 25). Given the evidence accumulated in the previous sections, this is clearly untenable. The main evidence are, of course, AUCH cases, where the AC is always necessarily outside the CC (an AC movement analysis or the like, saving the c-command constraint on the Auch-AC relation, being out of the question (see section 2.2) ). But as has been shown by the split cases discussed in section 2.1 , unstressed auch does not fully conform to this constraint either, for part of the AC material can be outside the CC. All that holds is that the CC of unstressed auch must include the constituent carrying the NA rather than the AC.32 Does the failure of the traditional view mean that there is no interesting relation between the CC of Auch and the AC/ID division at all? By no means. Looking at the regularities worked out in the previous sections from a different angle, we can formulate the following generalization:
Marga Reis and Inger Rosengren 267
2.6
Summary
In section 2.1 we showed that auch/AUCH are in complementary distribution at surface structure-auch requires that the AC is (entirely or partly) to its right and that it (its rightward part respectively) carries the NA; AUCH requires that the AC is (totally) to its left, ruling out a further accent to its right-, from which we inferred that auch/A UCH are variants of the same Auch. We also noted a remarkable asymmetry in the complementary distribution in that only auch, but not A UCH, allows split ACs, which reflects two systematic· regularities: (i) auch requires to its right the AC constituent potentially bearing the NA, not the whole AC; (ii) AUCH disallows overt AC material to its right. In section 2.2 we argued that there is no non -ad hoc analysis allowing us to derive the AUCH cases from auch cases, which suggests that the complementary distribution stated in section 2.1 is an epiphenomenon to be derived from independent regularities. In section 2.3 we discussed the CCs that Auch takes showing that (i) the CC of Auch is always an XP (never a head); (ii) Auch takes XPs of all kinds; (iii) Auch divides the clause or clause-like construction it appears in into two parts such that the Auch phrase is always an immediate constituent on the predicate projection line. By this division rule (48), the well-known restrictions concerning the absence of Auch from inside DPs, PPs, etc. were accounted for. It was also noted that Auch can only be stressed, if its CC is a predicate projection.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
material thus licensing stressed Auch are limited to predicative constituents. Thus no independent explanation is necessary. Note, too, that by (s8) split AC cases can only occur when Auch takes a predicate projection. This is in keeping with what we observed in section 2.1.1.2.34 Thus, the syntax of Auch, in particular the division into CC and the rest of the clause is obviously significant for the possible AC/ID divisions of the proposition, which in turn, together with the syntactic position of Auch (± immediate constituent of the predicate projection line · and position relative to the AC) is significant for the focus structures Auch clauses may have. We have summarized this in the two generalizations just stated, thus also indicating the essential topics that have still to be dealt with. We will return to generalization (s8 ) in section 3, where we will also try to show why the ± predicative nature of the CC plays the important role it does for the AC/ID division, and to generalization (59) in section 4, showing that it can be derived from independent focus principles.
268 Additive Particles: the Case of German Auch
3 SEMANTICS O F A UCH In the foregoing sections we argued that despite conspicuous variation in position and stress, there is just one lexical item Auch, which takes all kinds of lexiCal XPs on predicate projection lines as its CCs. We also argued that the syntactic division of the Auch clause into CC and the rest of the clause plays an important role for the AC/ID division of the corresponding proposition. . In this section, we will try to factor out the semantic properties of Auch in relation to the AC/ID division on the one hand and to the syntactic division into CC and the rest of the clause on the other, and then try to understand why the CC can interact with the AC/ID division in the way it does. We shall first sketch the central features ofprevious descriptions and point out where in our opinion they go wrong (section J . I ) This motivates our own proposal, which we present in section 3.2 and to which the CC is of central importance. In this connection we will also try to derive the generalization (s 8). .
3.1
Previous descriptions
Let us first summarize the aspects of previous descriptions of the semantics of Auch that we find uncontroversial: First, Auch does not affect the status of its host proposition p. This is illustrated by declaratives in (6oa, b) and questions in (6oc, d): whoever
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
In section 2.4 we argued that Auch itself is an XP adjoining to other XPs. We also assumed that Auch does not move itsel£ We then described the behaviour of Auch in terms of the clause structure proposed for German by Haider (1993). In section 2.5, finally, we summarized the evidence showing that the c-command condition on the relation between Auch and the AC is untenable, but that there are nonetheless important correlations between the syntactic division into CC and the rest of the clause and the AC/10 division. These correlations were captured in generalization (s 8), which brings out the importance of the ± predicative nature of the CC for the AC/ID division of Auch propositions. On the basis of (s8) we derived a further generalization (59), which brings out the relation between the nature of the material in the CC and NA assignment in Auch clauses, and we showed that the restriction of stressed AUCH to predicative CCs follows from this generalization.
Marga Reis and Inger Rosengren 269
asserts (6oa) also asserts (6ob), and whoever asks (6oc) also asks (6od). In other words, Auch(p) entails p, preserving the truth value of p, if it has one. (6o)
a. Peter hat Auch den Roman gelesen. b. Peter hat den Roman gelesen. Peter has (also) the novel read Peter read the novel (too). c. Hat Peter Auch den Roman gelesen? d. Hat Peter den Roman gelesen? Has Peter (also) the novel read Did Peter read the novel (too)?
(61) a. johannes hat den Roman gelesen. Johannes read the novel. b. Peter hat die Tagebiicher gelesen. Peter read the diaries. c. Peter hat den Roman gekaufi. Peter bought the novel. d. Peter hat die Kiiche geputzt. Peter cleaned the kitchen. Depending on which of these qs figures in the context of the actual utterance of (6oa), the AC/ID division ofp is different, or, to put it into the by now familiar terms, p and q belong to different sets of alternatives defined by this division. . Naturally, the make-up of the clause denoting p-including stress from which we have abstracted here-may place additional restrictions on what kind of proposition in context could act as q to which p is added. Since this is one, perhaps the major point where we take issue with previous descriptions, we will postpone its discussion to section J . I .J. Third, the propositional domain over which the alternative set induced by Auch is defined, its so-called 'semantic scope', is subject to the normal restrictions we observe with all focus particles (see Konig 1991a: 46ff): it is syntactically restricted to the minimal clause (or, to be exact, the minimal clause-like projection) it appears in, and semantically restricted by other scope-taking elements, if Auch is in their syntactic scope within the same minimal clause.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Second, Auch gives rise to the implicature that there is a true proposition q in context to which p (or parts of it) is added such that q, p are partially alike and partially different, i.e. have the same AC/ID division. In the case of (6oa), q could be, for example, any of the following propositions in context (the AC part in which q differs from p being italicized):
270 Additive Particles: the· Case of German Auch
The syntactic restriction is illustrated by (62)-(63), where the AC/ID division pertains to the proposition of the entire clause in (62a), but just to the proposition of the embedded clause in (62b), which results in a clear meaning difference: in (62), the ID part is Hans wollte, dafi Peter betrachtete in (62a), but just Peter betrachtete in (62b); in (63) the ID part is wollte, dafi Peter den Gauguin betrachtete in (63a), but again just Peter betrachtete in (63b). x
x
x
x
The semantic scope restriction induced by other scope-taking items like selten 'rarely', oft 'often', aile 'all', nicht 'not', etc. is illustrated in (64)-(6s): (64) a. Karl hat sonntags selten auch seine MUTTer besucht. b. Karl hat sonntags selten seine Mutter AUCH besucht. Karl has sunday-s seldom (also) his mother (also) visited On Sundays, Karl very rarely also visited his mother. (6s) a. Karl hat sonntags auch seine MUTTer selten besucht. b. Karl hat sonntags seine Mutter AUCH selten besucht. rarely visited Karl has sunday-s also his mother also On Sundays, Karl visited also his mother very rarely. (64a, b), where Auch is in the syntactic scope of selten, says that it was rarely the case on the (possibly many) Sundays that Karl visited e.g. his friends, that he also visited his mother-in other words, selten is outside p, q-, whereas (6sa, b), where Auch has scope over selten, says that on Sundays Karl rarely visited his friends and also rarely visited his mother-in other words, selten belongs to the semantic scope domain of Auch. Clearly, what is relevant here is the syntactic scope of Auch: only if the other scope-taking item belongs to its CC, is it part of the domain over which the alternative set is defined. This much in the meaning description is common ground. What then is controversial? To see this, let us look at how the meaning of Auch is usually
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(62) a. Hans wollte auch, daB Peter den GauGUIN betrachtete. Hans wanted also that Peter the Gauguin looked at Hans also wanted Peter to look at the Gauguin. b. Hans wollte, daB Peter auch den GauGUIN betrachtete. Hans wanted that Peter also the Gauguin looked at Hans wanted Peter to look also at the Gauguin. (63) a. Hans wollte AUCH, daB Peter den Gauguin betrachtete. Hans wanted also that Peter the Gauguin looked at Hans, too, wanted Peter to look at the Gauguin. . b. Hans wollte, daB Peter den Gauguin AUCH betrachtete. Hans wanted that Peter the Gauguin also looked at Hans wanted Peter to also look at the Gauguin.
Marga Reis and Inger Rosengren
27 I
described, see for example (66) (a simplified version of the analysis by Krifka . 1992: I9, see Kowalski I992: 89):
(66)
AUCH
( ( o:, ,8}) I.
�
o:(/3) & 3X[X � ,8 & X =/- {3 & o:(X)] 2.
3·
/3, X of the same type
3 . I .2 Where former analyses are too weak
What is wrong with Auch contributing no more than a conventional implicature to the meaning proper of Auch clauses? Looking only at
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
What (66) says is this: Part I says that Auch applies to a structured proposition, which means the division ofp into an identical (ID) and a variable part (AC) is given from the start: ,8 corresponds to the AC, a to the ID. The formula does not specify the source of this partition, but the standard assumption is that it is focus background structure, with (3 corresponding to the focus, and a to the background (see inter alia Jacobs I983, 1986; Roath I98s; von Stechow I991; Konig I99Ia, 199I b, I993; Krifka I992; Bayer 1996). In other words, Auch is taken to operate on the set of alternatives supplied by the focus structure of the Auch clause, whence the name 'focus particles', 'focus adverbs' for Auch and its kin. Part 2 says that the Auch clause denoting p entails p. Part 3 specifies the existence of q in that it says {= second and third conjunct) that there is some X, which is different from (3 and also fulfils a. The first conjunct is supposed to express (via �) that this X must belong to the set of alternative focus values that are contextually salient. The status of the conveyed proposition q is that of a conventional implicature. Note that besides contributing this implicature, Auch has no grammatically determined meaning of its owiL We have, of course, no objections to part 2 of (66), or to part 3 inasmuch as it specifies the existence of q such that p. q belong to the same set of alternatives. But we do have problems with part I and certain aspects of 3 : clearly, an adequate description of the meaning of one Auch should b e able to account for both its variants, the unstressed as well as the stressed one. We contend-primarily but not exclusively on the basis of AUCH cases that the standard description is unable to do the job: with respect to part 3. it is too weak in that the proper meaning contribution of Auch is just an implicature. With respect to part 1; it is too strong in that the exact partitioning ofp into an identical (a) and a variable part (fJ) is already given from the start, supplied by the focus structure of the Auch clause. Let us now give substance to this contentioiL
272
Additive Particles: the Case
of German Auch
unstressed Auch, it is difficult to see any problem, for both characteristic properties of conventional implicatures are fulfilled: (i) Unlike p, q is unaffected by negation and illocutionary variation of the respective auch clause. Take, for example, (67) with p being (67a) and q (67b). The negated and question variants of this clause (68) no longer entail that p [=(67a)], whereas they still convey that q [= (67b)]. (67)
b.
=
(ii) Negation of q does not affect the truth-value ofp. C£ (69): although B's answer entails the denial of q, the truth of p remains unaffected; only the use of auch in A's statement is made out to have been inappropriate. (69) A: Auch PEter will kommen ( 67) B: lnteressant, abet bisher will offenbar niemand sonst kommen. Interesting, but so far will apparently nobody else come Interesting, but so for apparently nobody else wants to come. =
But as soon as we turn to stressed Auch, there are problems with this analysis. Obviously, Auch can be focused (7o), and as such the meaning contribution spelled out in (66)/part 3 may even be sensitive to negation: whereas (7r) is always interpreted as the denial ofp, with q remaining intact, (72) (also, perhaps even preferably) allows for an interpretation as a denial of the existence of q rather than of p.35 (7o) Hans will AUCH kommen. Hans will also come Hans wants to come too. (7r) Es ist nicht {wahr/WAHR}, daB auch HANS kommen will. (reading: denial of p) that also Hans come will It is not true It isn't true that Hans wants to come too.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
a. b. (68) a.
Auch PETer will kommen. Also Peter will come Peter wants to come, too. ( p) Peter will kommen Johannes will kommen. (=q) Es ist nicht wahr; daB auch PEter kommen will. It is not true that also Peter come will It's not true that Peter wants to come too. Will auch PEter kommen? Does Peter want to come too?
Marga Reis and Inger Rosengren 273
(72) ?Es ist nicht {wahr/WAHR}, daB Hans AUCH kommen will. (preferred reading: denial of q) It isn't true that Hans also wants to come. In other words, the paraphrase of (72) is (73). (73) Es ist nicht wahr, daB jemand auBer Hans kommen will. It is not true that anybody else except Hans wants to come.
(74) a. *Feter ist intelligent, ABER faul. Peter is intelligent but lazy b. [Das hier ist NEU UND brauchbar,] *das hier ist ALT, this here is old [This here is new and usifu� ABER brauchbar. but useful (75) *SaGAR Peter ist gekommen./*Feter ist soGAR gekommen. Even Peter is come Peter is even come OB rigens gekommen. (76) *Feter ist Peter is by the way come Since conventional implicatures are, by definition (see Grice 1975), disjoint from the truth-conditional meaning of their host proposition, they should not be sensitive to negation either. Hence, we have to assume that the meaning contribution of Auch must be more than a conventional implicature; it must have a reflex in the propositional represe.ntation of the Auch clause that can be focused and negated. All this suggests that Auch, stressed or not, contributes a non-implicated, truth-relevant meaning element, which we shall call ADD (for 'in addition'). Thus, an Auch clause corresponds to ADD (p), meaning roughly, 'in addition' p. Like all other meaning elements, ADD may be focused, and since negation affects focused rather than backgrounded meaning (see Moser 1992), ADD may then also be affected by negation. Note that negating ADD (p) means denying that an adding operation took place, which amounts to denying that there is a q to which p is added. This
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The standard analysis cannot come to grips with these facts, for the meaning element to which focusing in (71) and denial in (72) pertains, would be the conventional implicature as such. But this is quite ad hoc: elements that contribute solely or primarily a conventional implicature usually cannot be focused;36 c£ aber 'but', and sogar 'even', which primarily contribute an adversative and a scalar implicature respectively (74)-(75), or iibrigens 'by the way' (76), which contributes an implicature of 'disconnectedness' of its host proposition.37
274 Additive Particles: the Case
of German Auch
3 . I . 3 Where former analyses are too strong
What is wrong with assuming that the AC/ID division of p is fixed from the start by focus structure (an assumption most clearly expressed by assuming Auch to operate on structured propositions)? Two things: (i) in determining the set of alternatives p, q belong to, the role of the focus background division of p is misrepresented, and, correspondingly, the role of q (and context in general) is totally underrated, (ii) the role of the syntactic division of the clause denoting p into CC and the rest of the clause is misrepresented and its true significance unrecognized. In both regards we think that former analyses have been led astray by their almost exclusive reliance on cases with unstressed Auch.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
suggests, of course, that the implicature concerning the existence of q and the entailment of p be made dependent on ADD.38 Indirect support for describing the meaning of Auch like this comes from related lexical items. Apparently, the properties we ascribe to ADD co-occur whenever the meaning of a lexical item involves the notion of addition. C£ for example adverbs like zusiitzlich 'in addition', iiberdies 'moreover', aufierdem 'besides', plus, extra and verbs like hinzufiigen, addieren 'add': they always involve two elements, the added element A, and the element B to which A is added, whereby the status of A does not depend on its being added (:::::> the 'entailment' property of ADD), but being added would be unthinkable without the existence of some B (:::::> the 'implicature' property of ADD). Clearly, one would not hesitate postulating a meaning element like ADD defined inter alia by these properties for these items, thus also bringing out the meaning component they have in common. If so, one should do likewise in the case of Auch, for Auch overlaps with these items in meamng. Moreover, additive particles have a clear etymological affinity to words denoting 'increase' or 'addition' and to coordinating elements in all languages (Konig 1991a: 164f). Thus, our paradigm case Auch is etymo logically related to the stem of Gothic aukan 'increase' (c£ Latin augere 'increase') and the coordinating particle in Swedish (och) or Danish (og). It is hard to see how these relationships could be captured without appealing to something like ADD as meaning component of (at least some of) these items, including Auch, for ADD by its very meaning combines aspects of addition, increase and connectivity at the same time. In sum, we believe that positing an element ADD as part of the meaning of Auch is necessary, the most salient justification for this being its ability to be focused.
Marga Reis and Inger Rosengren 275
Turning first to (i), there can be no doubt that in cases with unstressed Auch there is systematic covariation between the focus/background and the AC/ID divisiotL Thus, the narrow foci in (77a, b) correspond unambiguously to the AC, which restricts possible qs to those which differ from p just in the focus. value, see the examples right below (77a, b): (77)
Auch PETer hat den Roman gelesen. possible q for (a): johannes hat den Roman gelesetL b. Peter hat den Roman auch geLESen. possible q for (b): Peter hat den Roman gekaufi. a.
(78) a. Peter hat auch den RoMAN gelesetL =p Peter also read the novel. b. Peter hat etwas getan =/= Roman lesen. =q Peter did something =/= reading a novel c. Peter hat etwas =/= Roman gelesen. =q Peter read something =/= a novel Still, the correlation between focus ambiguities of p and possible qs, hence possible alternative sets, is also systematic. Thus it is quite understandable how the unstressed Auch cases could lead to the standard assumption that the AC/ID division is given by the focus structure ofp, in other words that Auch operates on the set of alternatives supplied by the focus structure ofp. But cases with stressed A UCH do not fit the picture at all, for if the NA is on AUCH, indicating a narrow focus, the focus structure cannot possibly tell us anything about the AC/ID divisiotL How then is this division determined in AUCH clauses? In many cases, just looking at p leaves it indeterminate; c£ (79) (the possible ACs being indicated by brackets): (79) daB (Pauij [ [seiner] Mutter] [ (diesen] Roman] AUCH schenkte that Paul his mother this novel also gave that Paul gave this novel to his mother for a present, too In this case the AC (and hence the AC/ID division) can only be determined in co-text or context, i.e. by comparing the AUCH clause with the proposition q it is added to. Trying to resolve the indeterminacy by positing structural ambiguity would be completely ad hoc. Neither can AC movement (c£ section 2.2) account for the indeterminacy.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
In these cases, the AC/ID division shared by p, q is identical with the focus background division ofp, 'given' by it so to speak. In other cases, the focus structure ofp may be ambiguous; c£ (78a), where the focus structure is such that we cannot know which of the two possible qs is the 'right' one (78b) or (78c). Only looking into the co-text or context, i.e. comparing with q, allows us to choose between them.
276 Additive Particles: the Case of German Auch
On the other hand, even if we can determine the partitioning from just looking at p, as in (8o)-(8J), focus structure does not play the role stipulated for it:
Rather, the determining factors, all cooperating with the basic regularity that the AC is to the left of A UCH, come from quite different sources: syntax as in (So), excluding las, which is reconstructed, thus leaving only Peter as possible AC; semantics as in (8 1), excluding aile, which as AC would lead to a contradiction; secondary accent as in (82): those accents must have a reason, which, in the absence of other interpretive options, could be highlighting rhematic material and thus identifying the AC. But it may be a topic, too, and-in the right context-the direct object would then be interpreted as the AC, although it is not accented (see section 4.2). Of course, extra-linguistic knowledge may also play a role, as in (8 3): the same cake cannot be produced twice, by different people, which excludes Peter as AC, etc. The conclusion is obvious: the division of p into AC/ID is not H.xed in advance. This much one might say is not really new, rather just additional strong evidence for the approach advocated by Rooth (1992) (see especially Rooth 1992: 107ff and 1 996: 279f): there, no lexical focus selection is stipulated, which allows the AC/ID division triggered by Auch; Hrst, to differ from the focus-background division of p, second, to be pragmatically H.xed by whatever factors come into play, even without the helping hand of focus. But Rooth and the others following him (see especially von Fintel 1995) go only half-way towards the correct solution, still misrepresenting the role of the CC. This is where our objection (ii) gets iiL What persists even in these approaches is the traditional doctrine about the role the CC plays with respect to the operation particles like Auch perform on the set of alternatives: it is still taken as a matter of course that the operation, in the
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(8o) Peter las AUCH den RomaiL Peter read also the novel Peter read the novel too. (8 1) Alle wollten Peter AUCH einladen. All wanted Peter also invite Everyone wanted to invite Peter too. (82) Und seiner MUTter /Mutter schenkte er den Roman AUCH. he the novel also And his mother[dat.] gave And he also gave the novel to his mother. (8 3) Peter hat diesen Kuchen AUCH gebacken. Peter has this cake also baked Peter baked this cake too.
Marga Reis and Inger Rosengren
277
3 .2
Our proposal
Let us now try to understand why the syntactic division into CC and the rest of the clause can affect the AC/ID division, and how. In doing this, we should be able to account for the generalization (s 8), here repeated for convenience, which shows that the ± predicative nature of the CC plays an important role for the overt distribution of AC and ID material. {84) Ifthe CC of Auch is a nonpredicative projection (DP, PP, AdvP, etc), then it always includes all A C material. Ifthe CC ofAuch is a predicative projection, then it may (a) include overt AC material, or (b) it may not; in the latter case it contains only overt ID material or is empty. The cornerstones of our explanation are the following: (i) the meaning operation ADD, which is characteristic for Auch, does not apply to the AC-elements in its CCs: rather what Auch does by virtue of its meaning contribution to the clause, is adding39 the contents of its CC in its host proposition p to the corresponding part of some contextually given proposition q, both propositions belonging to the same set of alternatives (AS); (ii) the AC/ID division defining the AS results in principle from the comparison between p and q (i.e. from context). Syntax and focus structure of p may place additional restrictions on possible qs; (iii) the added CC may contain either the (relevant part of the) AC or {part of) the ID including the predicate. Adding AC material means that
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
case of Auch the adding operation, crucially affects only the variable part {AC), this being the 'domain of quantification', and that all there is to the CC is that it happens to house the AC. What this amounts to is claiming, in effect, that the specific syntax of Auch does not play a role for its specific meaning contribution at all. Now, we have not only seen that the c-command condition, i.e. that the AC is always in the CC of Auch, is untenable (see section 2.5), but also that there are significant correlations between the CC and both parts of the AC/ID division, which is totally mysterious under the traditional view. The conclusion we draw from this is that the traditional view should be given up: it is not the case that the operation a particle performs is defined over the AC which happens to be housed by the CC, but the other way around: the operation is defined over the objects represented in the CC, which in the case ofAuch may house AC or ID material. This way, the specific syntax of Auch comes into play for its specific semantics, as should have been expected all along.
278 Additive Particles� the Case of German Auch
entities are added to one another that have nothing in common, whereas adding predicative ID material means that events are added to one another that are tokens of the same event type.
Now let us turn to the various constellations mentioned in (84):
(a) The CC of Auch is a nonpredicative projection, say a DP (as in (4a) ). It follows from (8 s ) that it does not represent an event type; it represents a term. Now by (iii) the CC must be added to some suitable part in q, from which it differs, another term. Differing terms, however, must always be ACs. This accounts for the first part of generalization (84). (b) The CC of Auch is a predicative projection, say a VP. If so, it always contains an element giving rise to the predication of the clause in its scope, by which the events are defined that p and q represent. This element occurs either in overt form or reconstructed, represented by a trace. Since the predications pertain to different events, no matter whether they belong to the same or different types, see (iii) and (8 5). the precondition for the adding operation that there be different entities is always fulfilled. Naturally, there is no way in this case to predict what the CC contains in the way of AC/ID except comparing p with q. This accounts for the second part of generalization (84). The comparison with q results in the following subcases: (ba) the overt or reconstructed ·material in the predicative CC is all AC (as in (4b), (4d) ) or mixed, some part being AC, some ID (as in (4c) or (6a, b)). The predicate may belong .to either in this constellation. (bb) the overt or reconstructed material in the CC is all ID material (as in
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The fact that addition, the meaning operation characteristic for Auch, results in adding of both AC and ID material is what distinguishes additive particles from restrictive particles like nur 'only', where the excluding operation necessarily pertains only to the variable part (which corres ponding to (i) is always in the CC of nur), the ID being no more than the defining property of the alternatives. We suggest that all differences between these two classes of particles can be related to this difference, although here we shall exclusively concentrate on what this distinctive property of Auch implies for Auch itsel£ (i}-(iii) provide the basis for deriving the AC/ID distribution with respect to the CC as described in (84). All we need in addition is the rather natural assumption (85), which will allow us to understand why the distinction between ± predicative CCs is relevant. (8s ) Predicative XPs represent event types; nonpredicative XPs (DPs,PPs, etc) represent objects which by themselves cannot define an event type.
Marga Reis and Inger Rosengren 279
What excludes that we have an AC/ID distribution like the one in (86), where the CC of Auch contains only ID material? (86) Auch Hans hat dem PETer das Buch gezeigt. Also Hans has the [dat.) Peter the book shown Hans too showed Peter the book. q: Hans hat Franz das Buch gezeigt. The answer is: the adding operation, which is defined over the CC, requires differing entities, but q shows that this condition is not fulfilled. Hence (86) is out. This answer can be extended to cases where all material is ID material. In this case the propositions just cannot indicate two different events of the same type, they are the same event. What about cases containing AC material not only in the CC, but also beyond it? In (bb), which concerns CCs that are predicate projections, this constellation was not excluded, because there are split cases and also occasional examples of double ACs on the predicate projection line. But cases like (87a, b) with AC material outside the CC are obviously impossible: (87) a. Auch Peter hat seiner Mutter ein BUCH gezeigt. Also Peter has his. mother[dat.) a book shown q: Paul hat seiner Mutter ein Bild gezeigt. b. Auch PETer hat seiner Mutter ein Buch gezeigt. q: Paul hat seiner Mutter ein Bild gezeigt. Obviously, a constraint is at work that AC parts must belong to the same XP projection, even if they are interrupted by ID material (which is not different from what we meet in the case of focus structure interacting with theme-rheme structure). There cannot be such a thing as a proposition with
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(7), (32) ). Then, p, q must be instances of the same event type. Since two events must differ in some part, it follows that the AC will be to the left of the CC. If there is no trace of it in the CC, then the AC has its base position to the left of the Auch in question. (be) the CC contains no overt material and only insignificant recon structed material, the finite verb being an auxiliary or the like (as in ( 12b) ). It follows that the infinite predicate defining the event type (the main verb, etc) is overtly .to the left, and only represented in the CC by its trace. Depending on whether this predicate figures as part of the AC or ID, p and q will be understood as representing events of different types or events of the same type. (a, b) exhaust the acceptable options. Let us now see how the impossible cases are excluded:
280 Additive Particles: the Case of German Auch
two ACs, one lying within the CC and the other outside it. This restricts split ACs to predicate projections as required. Taking all this together, there can be no doubt about the relevance of the syntactic division into CC and the rest of the clause for the interpretation of Auch clauses with respect to the AC/ID division. We shall see later on (section 4·3) how the different constellations in the CC also give rise to different utterance meanings. To sum up, what we think should be represented in the meaning description of Auch is the following: a non-implicated meaning element ADD, with which is associated a) the entailment of its host proposition p b) the implicature that a proposition q exists to which p is added; 2. the implicature that p, q belong to the same AS, i.e. have the same AC/ID division; 3 · access to the division into the CC and the rest of the clause such that the ADDing operation applies to the objects represented by the contents of the CC and the corresponding part in q. 1.
4· that the AC/ID division results in principle from the comparison between p and q (i.e. from context). Syntax and focus structure of p may place additional restrictions on possible qs, but this is strictly optional; 5· that ADDing p to q in the same AS implies that IDs containing the predicate are two different instances of the same event type; hence, not only ACs can be added to each other, but also predicative IDs; 6. that given (S s), it follows that nonpredicative XPs always denote alternative values if added, hence must be ACs, whereas predicative XPs can figure as ACs, but also as IDs. 7· that the correlations between the division into CC and the rest of the clause and the AC/ID division stated in (58)/(84) can be completely derived on the basis of 1-6.
3·3
Summary
In this section, we first sketched previous descriptions of Auch and then discussed what we consider their major flaws arguing (i) that a non implicated meaning element ADD is needed, (ii) that in determining the AC/ID division the focus structure of p is irrelevant, and context (comparing p with q) almost exclusively relevant, (iii) that in the adding
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Furthermore, we argued
Marga Reis and. Inger Rosengren 28 1
operation characteristic for Auch the material contained in the CC plays the crucial role. We then presented our own proposal, in which we try to do justice to (i)-(iii). Finally, we showed how the generalization (sS)/(84) can be derived on this basis.
4 FOCUS STRUCTURE AND ITS RELAT I O N TO THE SYNTAX AND SEMANTICS OF A UCH
4. 1
The focus-background theory
Let us begin with a few remarks on the basics of our focus theory: First, our focus theory is syntactically based.40 By this we mean that the focus-background structure (=FBS) of a clause is determined by the free assignment of a syntactic focus feature [ + F] to a constituent in this clause. The FBS of a clause, therefore, is ftxed when the clause so to speak meets information structure, which is the same as to say that information structure cannot alter it; it can only use it for its own purposes. Second, since the FBS is thus, by definition, ftxed in syntax and only read off at the level of information structure, it is in principle totally independent of topic/comment (TCS) and theme/rheme structure (TRS), the latter being the same as given/new in our approach (see Molnar 1 99 1 , 1 99 3 , and below). With both of them, however, FBS interacts in a systematic way. Thus, rhematic material must in principle be focused and thematic material can never carry the NA, except in contrastive function. Third, the FBS has a very general function: highlighting the information in the focused constituent, thus backgrounding everything else in the clause. Normally, the FBS evokes a set of contextually salient alternatives . thereby highlighting the focused constituent. This, however, is not the only function the FBS has, as we shall see below. Fourth, the overt realization of focusing is accentuation, determined in PF (in interaction with syntactic structure). Since the same accent pattern
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Given the results of the foregoing sections, we will now try to solve the final, modular task: to show that the accent pattern as well as the interpretation of Auch clauses-which comprises the actual division of the host clause of Auch into the variable (AC} and the identical part {ID), plus the observable variations in utterance meaning-derive without further stipulation from the syntax and semantics of Auch interacting with principles of focus structure and context.
282 Additive Particles: the Case of German Auch
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
may sometimes realize different focus structures, clauses may be overtly ambiguous as to their FBS. In this case, and only in this case, the co-text or context may disambiguate them. Now to the specifics of this focus theory: First, [+F) may be assigned to any (lexical) constituent (sometimes to more than one constituent} in the syntactic structure of the clause, but must be assigned to at least one. In order to be identifiable at PF (where the accent pattern is determined), [+F) must be assigned to one or more constituents before Spell-out. We will assume furthermore that there is no covert movement of [+F) after Spell-out, on its way to LF, so to speak (no focus raising, etc.). [+F), finally, has no meaning of its own beyond the informational function of highlighting the whole clause or part of it. Second, the focus domain in this model does not result from [ + F] percolating up the tree (= the bottom-up . approach proposed by Rochemont 1986; Selkirk 1984; Hetland 1 992; Winkler 1 997; and others), but from spreading downwards from a dominating node {top-down} to all constituents dominated by [+F] (see Jacobs 1993; Uhmann 199 1 ; Rosengren 1993b, 1 994). In the case of wide focus (i.e. the whole clause being in focus), the dominating node is the uppermost node of VP. [+F] may also be assigned further down in VP, thus giving rise to focus domains covering only part of the VP, or it may be assigned to DPs, PPs, APs, etc. or parts of them, in both cases giving rise to a more or less narrow focus. Note that all constituents which lie within a focus domain are focused, irrespective of their status. Consequently, if the uppermost VP bears [+F), all adverbials contained in it are focused, too. Hence, the domain dominated by [ + F] constitutes a focus domain. The top-down model restricts the extension of the focus domain in a natural way. {A bottom-up percolating feature e.g. on the direct object would percolate at least to the top of the VP since the percolation could not be stopped without stipulation) A top-down model also avoids another problem bound to a percolating [+F) feature: since percolation by definition is operating bottom-up, it is difficult (if not impossible) to explain how [+F) can jump' from the 'focus exponent' {=FE), as a rule a non-head, to the head and from there percolate up the head line. Restricting focus assignment to lexical projections amounts to saying that functional projections (TP, AgrP, etc} are not part of the focus of a clause, not even if the focus is wide (as for the special status of Verum focus, see note 8). The theoretical consequences of this are too manifold to be discussed here (c£ Rosengren 1993b, 1994) and of no importance for the issue at hand. Let us just note that language specific differences concerning functional A projections such as weak and strong features-which do have
Marga Reis and Inger Rosengren 28 3
(88) [cP Paul hati +F[VP gestern seiner Mutter Grass' neues BUCH yesterday his mother Grass's new book Paul has gekauft ti] ]. bought Yesterday, Paul bought his mother the new book by Grass. Our focus theory tells us that (88) should have a wide focus reading, for (i) [+F] dominates the whole VP and the rightmost argument, the direct object, is the focus exponent, (ii) the clause exhibits the basic relative word order between the arguments, which is not altered by the optional topicalization of the subject, nor by the adjunction of gestern to the VP. However, if the direct object (the focus exponent in (88) ) is scrambled as in (89) or topicalized, passing by the subject or an adverbial, the object leaves
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
consequences for the result of Spell-out (weak features being checked at LF and strong being checked before Spell-out)-do not have any impact on focus structure (at least not in languages like German and English). Third, wide focus enclosing the whole clause presupposes basic word order (being the order in which the arguments of the lexical predicate are projected into syntax). Changes in relative order by which a constituent leaves the focus domain in order to pass by another constituent (e.g. by scrambling, but sometimes also by topicalization; see Rosengren 199 3 b, 1994) will automatically result in a restricted focus domain (which is not the same as a narrow focus domain). In other words, the focus component interacts with a syntactic word order subcomponent (until now left undefined in the minimalist program) which rearranges constituents in such a way that wide focus cannot be obtained any more. Since only these by-pass movements have this effect, topicalization of the subject, for example, which preserves relative order, will be possible without any impact on the focus domain. Finally, there may be more than one [+F] in a clause, resulting in more than one focus domain. This is to be kept strictly apart from [+ F] spreading down within the same focus domain. Fourth, there exists a systematic relation between focus assignment and argument structure, which (at least in part) determines the accent assign ment in PF. We assume that the phonological correlate of [+ F] is an abstract accent marker [+P], which is related to [+F] in a systematic way, and triggers the nuclear accent (NA) in PF. The constituent carrying [+P] in the domain of [ +F] will be called focus exponent (FE). In the case of wide focus, [+P] is normally placed on the verb next argument. Under certain conditions, however, adjuncts and predicates may also be focus exponents. Let us now demonstrate briefly what this model predicts about focus structure in clauses with more than one argument:
284 Additive Particles: the Case of German Auch
the focus domain indicated by [+F] on the uppermost VP, the result being a reduced focus domain. In this case the indirect object, which is now the verb next argument, will take over the function as FE. (89) [CPPaul ha� [ VP[Grass' neues Buch1 +F[gestern seiner MUtter t_j Paul has Grass's new book yesterday his mother gekauft ti] ]]. bought Paul bought the new book by Grass for his mother yesterday.
4.2
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Note, too, that the moved direct object may be unfocused as in (8g), thus being part of the background, but it may also carry a [+F] of its own, this focus then being narrow.
How does focus work in Auch clauses?
In keeping with our modular approach, we expect that in principle the focus behaviour of Auch clauses does not differ from the focus behaviour in clauses without Auch. Likewise, we expect that the interaction of the FBS with TRS and TCS is the same. But since there is one more focusable element-Auch-, we should not be surprised to also find differences between the focus structure of the Auch-less clause and the Auch clause for which this additional element is responsible. In fact, as we actually saw above, there is a systematic relation between focusing and the CC of Auch, as shown by the empirical generalization formulated in section 2.5, here repeated for convenience: {9o) If the CC ofAuch includes overt AC material, it gets the NA, if it does not, Auch gets it. Given the fact that overt AC material in the CC of Auch-whi�h then is always unstressed-always includes the focus exponent, i.e. the constituent that would normally carry the NA in the corresponding Auch-less clause (c£ sections 2.1 and 2.5), the first half of (go) is entirely in keeping with normal focus behaviour. Since Auch says that the proposition of the Auch clause belongs to an alternative set, the AC part being one of the alternative values, the AC is coextensive to a focus domain. Thus, in clauses with unstressed auch the focus does its normal work: it highlights the alternative in question. Note that the focus domain may include auch (in split cases), which means that +F could in principle also directly dominate auch in unsplit cases. (Since in this case the FE will still be within the CC, nothing depends on whether +F is assigned to the entire auch phrase or just to the CC. We will therefore not pursue this question any further).
·
Marga Reis and Inger Rosengren 285
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
But what about the second half of (9o)? In section 3·4· we saw that whenever Auch gets the NA, there is only ID material in its CC or the CC is empty. Thus the obvious explanation would be that in the constellation described in the second half of (9o) Auch is the last focusable material in the clause, which in terms of FBS means that the . rest of the clause is background. In other words, the constituent carrying the NA if there were no Auch, does not carry it; at best, assignment of a secondary accent is possible. Hence, the AC domain does not correspond to a focus domain any more either. Obviously then, focusing Auch overrules the normal FBS, and identifying the AC/ID division is more or less (A UGH signals itself that overt material to its right is ID material and syntax may offer a helping hand) relegated to context, i.e. the comparison between p and q. Clearly, the fact that focusing Auch has these consequences must be related to the function this focusing has. . What then is this function? . A comparison with modal particles (MP) may be instructive. Ormelius Sandblom (1 997: 101ff) argues (i) that MPs themselves may be focused or be included in the focus domain if the focus projects, but may also lie outside it (c£ also Meibauer 1 994: 62ff), (ii) that MPs are (minimally) focused, whenever they are the last focusable material in the clause, and that they may carry secondary stress when to the left of the NA, (iii) that they are not part of the TRS. Obviously, Auch behaves just like MPs regarding (i) and (ii). Here, we are just interested in (iii), which may seem controversial. Ormelius-Sandblom argues in the following way: since MPs always carry new information and, therefore, by definition, should always be rhematic and hence focused themselves, the fact that they may stay unfocused would lead to a contradiction, if we keep to the aSsumption .that they take part in the TRS. Looked upon from the perspective of the TRS, we may also put it in the following way: the distinction between theme and rheme implies that what can . be a rheme should also be able to be a theme, but MPs are obviously never c-construable. Hence, the theme/rheme distinction does not make sense in their case. The parallel between MPs and Auch is obvious: Auch, focused or not, contributes new information to the clause. Hence focusing it cannot be taken as just signalling that it belongs to the rhematic part of the clause, and leaving it unfocused will not tum it into thematic material. We will assume, therefore, that Auch, like MPs, is not part of the TRS either.4' The same argument holds, by and large, for all so-called focus particles. What, then, is the function of focusing Auch, if it is not highlighting that it is rhematic? All that seems to happen is, trivially, that the adding
286 Additive Particles: the Case of German Auch
(91) daB [Peter m] [auch [cc [+F seiner MUtter Ac] that Peter also his mother (das Buch gezeigt hat m] ]] the book shown has that Peter also showed the book to his mother (911) CP
�
SpecC
c·
� daft � C0
VP
DP[ = IDJ
Peter
VP
� �
auch
VP[= CC J
DP+F [ = AC J seiner MUTter
VP[= IDJ
�
DP das Buch
VP gezeigt hat
In (9 I), Auch takes a VP as its CC, which happens to contain a minimally focused constituent. It goes without saying that this focus identifies the AC, which as the last focusable material gets the NA. Everything else must be ID material. Hence, in this case, the AC can be inferred from the position of the NA (seiner perhaps being c-construable). In (92) Auch takes the DP (= 10) as its CC:
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
operation is focused, otherwise the AC is focused. If so, focusing in Auch clauses follows from completely general focus principles. In a way, the only interesting contribution coming from ± focusing Auch is informational: focusing it 'means' that there is only ID material or no overt material to its right and that consequently the AC material is to its left and not focusing it 'means' that the AC material (or part of it, including the FE) must be to its right. In order to illustrate the relation between the CC of Auch, the AC/ID division and the focus structure, let us now look at a few concrete examples. The following three cases have the same AC/ID division but different CCs: 42
Marga Reis and Inger Rosengren 287
(92) daB [Peter 10] [+F auch [cc [seiner MUtter Ac] ]] that Peter also his mother [das Buch gezeigt hat 10] the book shown has that Peter showed the book also to his mother CP (92')
�
C'
SpecC
�
VP
co daft
�
VP
V P[ = I D J
DP+F
�
auch
DP( = CC ) (= ACJ seiner MUTter
�
DP das Buch
VP gezeigt hat
a consequence, the CC must contain all the AC material {see above (5 8)/(84) ), to which the focus must correspond. In every other respect (91) and (92) are alike.
As
(93) daB [Peter 10] [seiner Mutter Ac] [ [+F AUCH] that Peter his mother also [cc [das Buch gezeigt hat 10] ]] the book shown has that Peter has shown the book to his mother, too (93 1)
�
c·
SpecC
� C
0
diJ.f3
VP
�
DP(= ID J Peter
VP
�
DPI = AC ) seiner Mutter
VP
�
AUCH+ F
VP( = CC J ( = I D J
�
DP das Buch
In (93) Auch
is
VP geuigt hat
focused, in which case it must have a
VP
as its CC that
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
�
DP( = I DJ Pete
288 Additive Particles: the Case of German Auch
contains only ID material, thereby indicating that the AC is to the left. The exact AC/ID division can only be determined by comparison with q. Let us now look at a split case, the AC being the entire proposition:
(94) daB [+F [ein Student Ad [auch lee mein AUto angefahren hat Ad ]] that a student also my car hit has that, in addition, a student hit my car CP (94')
�
SpecC
C"
�
V P+ F [ = AC ]
�
DP ein Student
VP
�
auch
V PI = CC I
�
DP+ P mein Auto
VP angefahren hat
In correspondence to this AC, [+F] comprises the highest VP; accordingly the FE must be the DO. Auch is included in the +F-domain, but could also have been adjoined to the highest VP. (Note that in this case a common denominator for p and q functioning as ID has to be inferred.) The overt accent does not tell us that we have a split case, being compatible with a number of other AC/ID divisions and corresponding +F assign ments. This ambiguity can only be resolved in context, i.e. by comparing p and q. Let us now look at a case where the CC is overtly empty:
(95) [seiner Mutter das Buch gezeigt Adi hatj [ [Peter ID] [ [ +F AUCH] lee t; �] ]] In this case, the whole VP, which contains AC material, following Auch is moved to the initial field. This requires at once that Auch carries the NA, since it is now overtly the last focusable material in the clause. A secondary stress on Buch is not obligatory in -this case but may be useful when processing the information structure of the clause. (C£ also below). As for the AC/ID division, the non-normal constituent order tells us that not all material to the right can be AC material. Thus, Peter must be ID material. Of course, the AC/ID distribution could overtly be just the other way around. Between these options only comparison between p and q can decide.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
cO dajJ
Marga Reis and Inger Rosengren 289
Let us finally look at a case where the 'split' of the AC is so to speak upside down in that the AC part within the CC is to the left and higher in the configuration than the AC material outside the CC:
(96) represents a so-called thetic clause, i.e. a clause with wide focus, where the subject is the FE (see Rosengren 1997 for a recent discussion of this kind of clause). This corresponds to the AC being the entire proposition. Still, Auch just takes the DP in the initial field as its CC, thus c-commanding only the AC constituent carrying the NA.43 This case is especially interesting because it is the only case where an Auch+DP may be FE in a wide focus projection, i.e. a projection with a + F on top of the VP (another projection line). This shows even more clearly than normal split cases that not more than the FE need be within the CC ofAuch, which then projects in its usual way. Let us now illustrate a more complicated interaction between the AC/ ID division Auch gives rise to, the FBS and the TRS. We saw in section 2. 1 . 1 .2 that there is no 1 : 1 relation between AC and rhematic material in that the AC may comprise c-construable thematic material, which is in line with how the FBS and TRS usually interact. Much more interesting, however, are examples like the following, where the interaction in Auch clauses differs from what we would expect in the corresponding Auch-less clause: (97) Was tat Gerda, als sie in das Zimmer trat, in dem . Gerd What did Gerda when she in the room entered in which Gerd die Blumen goB? the flowers watered What did Gerda do when entering the room in which Gerd was watering the plants? a. Sie hat [+F Gerd beGRUSST] She has Gerd greeted She greeted Gerd. b. *Sie hat [+F GERD begriiBt]. c. Sie hat auch [+F GERD begriiBt].
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(96) [Heute ist ein furchtbarer Tag. Die KINder weinen dauernd, Today is an awful day. The children cry constantly, der FERNseher geht nicht, und] auch der HUND works not, and also the dog the TV set ist verschwunden. is disappeared [Today is an auful day. The children keep crying, the TV set doesn't work and] the dog ran away, too.
290 Additive Particles: the Case of German Auch
d. *Sie hat auch [+F Gerd beGRUSST]. e. ??Sie hat Gerd [ +F AUCH] begriifk
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The auch-less examples (97a) and (97b) were used by Jacobs (1988) to point out that the NA cannot be on the expected FE (Gerd), since this is thematic material, the NA therefore being carried by the verb. But as soon as Auch is adjoined to the respective VP, the only really acceptable assignment of the NA is to Gerd (in the relevant reading: 'what Gerda did in addition to other things was saying hallo to Gerd'). Why is this so? Obviously, Auch dominates the information structure in that the FBS is taken to be in the primary service of the AC/ID division induced by Auch: the AC corre sponds to the entire focus domain, and must be made 'visible' by the accent pattern regularly signalling this domain; thus the TRS and normal effects of TRS-FBS interaction may be overruled. The deviance of (97d) underlines this point: the FBS is taken at face value by the AC/ID division; hence, the verb carrying the NA is interpreted as minimally focused, and thus should be the only AC material in the clause, but this interpretation is not supported by the relevant context (the question). As for the marginality of (91e), AUCH would require that the material in its CC is ID material, which is also ruled out by the relevant context. Note, however, that the informational dominance of the AC/ID division does not mean that the TRS cannot be derived from context. In the context of a question like in (97), prompting an adequate reply, the TRS may figure, too, but 'behind' the AC/ID division, so to speak. Let us now turn to the function of possible secondary stress in Auch clauses, which as we pointed out in section 2.1.1.1, may occur (i) on a constituent to the left of Auch, if Auch or its AC to the right carries the NA, (ii) on Auch itself, if the AC part carrying the NA is to its right. Concerning the first case, let us again stress the fact that secondary accent is not in the service of identifying the AC. Rather, secondary accents do the same work they always do (we restrict ourselves to focus accents, of course), i.e. (optionally) marking further focusable material, or contrastive constituents, or topics. If the constituent marked this way happens to coincide with the AC, the secondary accent will in a way 'signal' that it is the AC, otherwise it will not. Neither is the assignment of secondary stress dependent on the rhetnaticity of the AC: a rhematic AC need not get secondary stress, as we have seen over and over again, nor need the AC always be rhematic; c£ (98), where Peter is thematic material c-construable via the question (98A), but still the AC of the clause in (98A'-A"). More importantly, this example demonstrates that secondary stress may mark a topic which is not the AC at the same time:
Marga Reis and Inger Rosengren 29 I
(98)
Wer hat denn die schonsten Impressionisten, Who has [MP] the beautiful-est impressionists Johannes oder Peter? Johannes or Peter Who has got the most beautiful impressionists, Johannes or Peter? B. Johannes hat jedenfalls einen GauGUIN. Gauguin Johannes has any case a At any rate, johannes has got a Gauguin. A'. Einen GauGUINi hat Peter/PEter AUCH ti. also A . Gauguin has Peter Peter also has got a Gauguin. A". Einen GauGUINi hat auch PEter ti. A
Aber sie ist auch EHRgeizig. (99) a. Sie ist klug. She is intelligent but she is also ambitious b. Aber EHRgeizig ist sie AUCH. But ambitious is she also But she is also ambitious. (10o) a. Er hat Luise geschwangert, und er hat sie auch geHEiratet. He has Luise made pregnant and he has she also married He got Luise pregnant and he also married her. b. Er hat Luise geschwangert, und geHEiratet hat er sie AUCH. He got Luise pregnant and he also married her. In (99a)-(10oa) we have an AC to the right of Auch. Moving this material out of the scope of Auch (in order to topicalize it) as in (99b)-(1oob), results automatically in Auch now carrying the NA, since nothing is left in its scope
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Two things are of interest here: first, in (98A11) Peter must carry the NA, since it is the last focusable material, in (98A') it may carry a secondary stress, the function of which may be contrastive or to highlight rhematic material, as well as highlighting the AC, or any combination · of them; second in (98A1), and in (98A''), cases of so-called i-topicalization, the preverbal constituent has got a secondary rise-fall contour. This rise-fall contour is typical for i-topics (c£ Biiring 1997; Jacobs 1997; Molnar & Rosengren 1997) and has nothing to do with the division into AC and ID. This underlines that secondary stress may be in the service of other information structures, which can be realized by the Auch clause in addition to its characteristic information structure, the AC/ID division, and the FBS. The following examples are counterparts to (98) in that the material in the scope of Auch is AC material.
292 Additive Particles: the Case of German Auch
that could carry it. (C£ also (104) below). Note that the rise-contour on the moved constituent is optional. Let us now discuss the function of secondary stress on Auch,44 looking Hrst at the normal case in which this accent is optional: . (1o1) Petra ist sehr intelligent, aber sie ist AUCH sehr EHRgeizig. Petra is very intelligent but she is also very ambitious ( 1o2) Ja, AUCH mit DIEsem Fall hatten wir Probleme. Yes, also with this case had we problems Yes, we did have problems with this case too.
·
Geld. (1o3) Es geht nicht NUR urns Geld. Es geht AUCH urns It goes not only about money It goes also about-the money It's not only about money, it's also about money. Let us finally look at examples, where we have obligatory secondary stress on A UCH in the initial field; c£ (1o4a, b) vs. ( 104C, d) and the representation of (104e) in {1041): ( 1o4) a. Peter und Ulrike wolltenj auch zusammen SPIElen tj . Peter and Ulrike wanted also together play Peter and Ulrike also wanted to play together. b. [Auch zusammen SPIElen]; wolltenj Peter ml.d Ulrike t; tj. Peter and Ulrike also wanted to play together. c. Peter und Ulrike wolltenj AUCH zusammen spielen tj. Peter and Ulrike wanted to play together, too. d. [Zusammen spielen]; wolltenj Peter und Ulrike AUCH t; S· Peter and Ulrike wanted to play together, too.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The function of the secondary accent on Auch in (Ioi}-(1o2) is quite different from the function of primary stress on Auch discussed above. What it apparently does is to underline the truth of the added proposition p against contextually salient doubts or even assumptions of the opposite. Evidence supporting this hypothesis is the fact that this type of secondary accent typically occurs in the context of aber 'but'. It implicates the kind of contradictory assumption required, which in the case of (Ioi} would be that being intelligent excludes being ambitious. Also typical is that these accents occur in utterances trying to dispel doubts concerning certain facts, as can be easily imagined as a context for ( 1 02). Note that there are also cases like (IoJ), in which the accent is not only outright contrastive, but plainly corrective: what is conveyed is that nur 'only' is an inappropriate word to be replaced by Auch. In this case, and as it seems, only in this case, Auch carries the NA, although its AC is still to the right.
Marga Reis and Inger Rosengren 293
e. [AUCH zusammen spielenh wolltenj PEter und UlRike ti S· Peter and Ulrike wanted to play together, too. (1o4e') [ [+F AUCH) [cc zusammen spielen IDH wollten [+F Peter und UlRike ti) CP
-------
C"
�
AUCH+F (secondary)
V P( = CC J
zus. spieleni
C0
wollteni
VP
�
DP+ F (=ACl
VP
t i tj
for (104a, b), the AC being in the scope of auch will, of course, carry the NA no matter whether the Auch phrase is in its final position or in the initial field. The interesting cases are { 104C, d). In (104c) AUCH tells us that zusammen spielen is ID material. In (1o4d) only the CC, comprising ID material, is moved to the initial field. Moved material in this position, however, could also be AC material. Hence, {Io4d) is overtly ambiguous: either the moved VP or the subject is the AC. In (1o4e), however, the VP including Auch is moved, the AC gets the NA as the last focusable material in the clause, but Auch nonetheless requires a secondary accent in this position. We cannot really explain this. What is obvious, however, is the parallel to what we have seen when the AUCH VP is in situ: if it contains the AC, it will get the NA, if it contains only ID material, Auch must get the NA, thus signalling that there is only ID material in its scope. What happens apparently, if Auch-constituents are moved, is that they retain their NA, which in the case of (1o4e) can only be realized as secondary stress. This secondary stress may in turn be taken as signalling that the topicalized CC material is just ID material. Summarizing this subsection, we have argued (i) that focusing in Auch clauses does not differ in principle from focusing in Auch-less clauses, (ii) that the function of focus on Auch is to highlight the adding operation, which results in its overruling the FBS, (iii) that as a result of this function focusing Auch has purely informational effects. We also showed that the accent pattern in Auch clauses is just an instance of the independent principle that the last focusable material in a clause must get the NA. Finally we discussed a number of examples displaying interesting aspects of the interaction between syntax of Auch, semantics of Auch, focus structure and context in relation to the AC/ID divisioiL As
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Peter u. U/Rlke
294 Additive Particles: the Case of German Auch 4· 3
The utterance meanings of Auch
(105) a. [Paul war gestem im Kino.) GeARbeitet; hat Paul Paul was yesterday in the cinema. Worked has Paul gestem AUCH t;. yesterday also Yesterday Paul went to the movies. He also worked yesterday.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
So far, we have concentrated on the interaction between the focus structure, the syntactic structure and the grammatically determined meaning of Auch. In section 3.2 we mentioned that this gives rise to two utterance meanings: 'in addition/furthermore' and 'likewise'. We shall now look at these utterance meanings in more detail.45 As shown in section J, Auch says (i) that its host proposition p and some proposition q in context are in the same AS, i.e. p, q are partly different (AC) and partly identical (ID); (ii) that the content of its CC is added to the corresponding part in q. As also shown, the CC must contain the AC, if it is a nonpredicative XP (DP, PP etc), but if it is a predicative XP, all options are possible: only AC material, only ID material, or a mixture of both. In that case only comparison with q can decide which option is realized. How does this give rise to different utterance meanings? Given the central role of th� CC, we may expect that if Auch adds only differing material or corresponding traces, the utterance meaning will be 'furthermore' emphasizing the aspect of difference between p, q, and if it adds only identical material or corresponding traces the utterance meaning is 'likewise' emphasizing the aspect of . sameness. (What will happen if Auch adds both kinds of material is somewhat more difficult to predict. Perhaps this will give rise to a conflict between the two options.) In the previous section, we argued that + F will always be assigned to the last focusable element, i.e. the overt AC part in the scope of Auch, if there is one, otherwise to Auch itself, i.e. there is no overt AC material. Since the CC can be all ID only in the latter case, the utterance meaning 'likewise' will only occur with stressed Auch. But since focusing operates on . overt structures, Auch is also stressed if its CC contains AC traces. Thus, we cannot expect a I : I relation between auch/AUCH and the two utterance meamngs. Let us now look at some telling cases. A minimal pair is provided by (10sa, b), where the covert material in the CC of A UCH is the same, but the trace of the main predicate gearbeitet corresponds to exclusively AC material in (10sa), and to exclusively ID material in (10sb):
Marga Reis and Inger Rosengren 295
b. (Peter hat gestern gearbeitet.] GeARbeiteti hat Paul Peter has yesterday worked. Worked has Paul gestern AUCH ti. yesterday also. Peter worked yesterday. Paul also worked yesterday.
(1o6) a. Paul hat dem VAter nicht geschrieben und er hat auch Paul has the father not written and he has also die MUtter vergessen. the mother forgotten Paul didn't write to his father and he also forgot his mother. b. Paul hat dem VAter nicht geschrieben und er hat Paul has the father not written and he has die MUtter AUCH vergessen. the mother also forgotten Paul didn't write to his father and he also forgot his mother. Again, the contents of the CC determine what is highlighted: in ( 1o6a) it is the variable part (Vater vs. Mutter), giving rise to the 'furthermore' interpretation, whereas in (1o6b) the fact that we have two tokens of the same event types is emphasized, giving rise to the 'likewise' interpretation. Note that in (1o6b) but not in (1o6a), 'not writing' must be identified with 'forgetting', which actually proves the same point. This effect appears in more dramatic form in the examples (1o7)-(1o8) (adapted from Kaplan's 1 984 discussion of differences between English also and too, but originally going back to Green 1 968, 1973): ( 107) Ich sah, daB I saw that a. PEter vom FISCH nahrn und PAUL auch vom Peter of-the fish took and Paul also of-the
·
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
There is a clear intuition that in ( 105a) only the utterance meaning 'furthermore' is possible, in (105b) only 'likewise'. That this is not a chance effect is shown by the cases (98)-(10o), where the material behind Auch has also been topicalized: in (98A') where the trace in the scope of Auch represents ID material, the natural interpretation is 'likewise', i.e. Peter, just like Johannes, has a Gauguin. In (99b)-(1oob), however, where the trace represents AC material, the 'furthermore' interpretation is clearly preferred. Another minimal pair, involving auch vs. AUCH, is (106), which takes up ( 1 8) above: the AC/ID division is in principle the same in both clauses, but the CC of auch contains the AC, the CC of A UCH only ID material.
296 Additive Particles: the Case of German Auch
FLEISCH nahm. meat took
Peter took some of the fish and Paul also took some of the meat.
b. PEter vom FISCH nahm und PAUL auch das FLEISCH Peter of-the fish took and Paul also the meat nicht verschmahte. not scorned
Peter took some of the fish and Paul didn't say no to the meat either. vom Fleisch AUCH das Fleisch AUCH
nahm nicht
Consider first the auch-variants (1 o7a, b), which are more or less understood the same way (i.e. 'von X nehmen' is equated with 'X nicht verschmahen'): Paul took from the meat in addition to taking from the fish. That he took from the fish rather than from something else is an implicature: the second conj unct containing auch implies the existence of a proposition 'Paul nahm von x :j:. Fleisch', with the first conjunct as the closest co-text forcing the further specification of x = Fisch. Note that auch must be interpreted as 'in addition/furthermore'; any 'likewise' reading is impossible. By contrast, the A UCH counterparts in ( 1 08 ) sound almost incongruent: apparently, A UCH induces an identity reading of the predicates suggesting that eating fish --:- eating meat. At first glance, one might suspect that this only happens if Paul is inadvertently taken as the AC of A UCH. However, the incongruence effect remains stable even in the reading where the AC = das Fleisch, which shows that in A UCH-constellations like ( 1 08 ) the aspect of sameness of the tWo propositions in question rather than their difference is highlighted. In other words, A UCH induces the utterance meaning 'likewise', as predicted. The difference showing up in minimal pairs is confirmed by AC/ID constellations in which either auch or A UCH is licit: As pointed out in section 2. 1 . 1 . 1 , if the AC is the finite verb, only unstressed auch is possible, the reason being that the finite verb, visibly or reconstructed, always lies in the scope of Auch. C£ ( 1 09) (corresponding to ( 1 3 ) above):
( 1 09)
a. daB Peter das Referat auch TIPPle I Peter TIPPtei das Referat auch ti. (that) Peter the report also typed
(that) Peter also typed the report tipple das Referat AUCH ti.
b. *Feter
Since the NA unambiguously tells us that there cannot be two events of the
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
( 108) Ich sah, daB a. ? ?PEter vom FISCH nahm und PAUL b. ?*FEter vom FISCH nahm und PAUL verschmahte.
Marga Reis and Inger Rosengren 297
same type, the prerequisite for the 'likewise' reading, we expect the 'in addition/furthermore' reading to be clearly prominent, which it is. In the following case, only A UCH seems possible: (no) A Ich wiinsche Dir einen schonen Sommer. I
wish
you a
nice
I hope you enjoy your summer.
summer
a. B. Ich dir AUCH (??auch ICH DIR). you also ( also I you) I
The same to you.
(Thanks,) the same to you.
A UCH seems
tp be the only variant acceptable in this context, the reason being that the Utterance meaning must be 'likewise' here. That the context (including the conventions for replying to good wishes) do in fact require this utterance meaning, is supported by the possible alternative replies in this situation (n ob), which all turn on the notion of sameness.46 Let us now turn to a case that seems to contradict our predictions: (I I I) a. Wenn DU MICH besuchst, werde auch ICH DICH besuchen. If you me visit will also I you visit
If you visit me I will visit you too.
b. Wer SCHOTten BOURbon verkauft, verkauft auch Who Scots[dat.] Bourbon sells sells also ESkimos EISschranke. Eskimos fridges
Whoever can sell Bourbon to the Scots can also sellfridges to the Eskimos. ·
In these cases the co-text shows that the finite verb is always ID material, but there are two ACs to the right of auch. Still, the utterance meaning 'likewise' is by far the preferred meaning in both ( 1 1 1 a) and ( 1 1 Ib), if not the only acceptable one. Why is this so? The answer may be that in (I I I a) as well as ( u x b), there is a common denominator for the two events: in ( I u a) the compared AC parts are in a way identical (speaker + adressee), and the variation results solely from the different thematic role of the pronouns; in ( 1 1 I b) the two propositions exemplify the same general meaning that there are people succeeding in selling other people things they already have or do not need, in other words, they are two tokens of the same event type. Thus,
co-text or context blocks the utterance meaning 'furthermore', which . would be meaningless here.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
b. B. Danke, gleichfalls. I Danke, dito, dito. I Ich wii�ch Thanks, likewise. / Thanks, dito, dito. / I wish dir das gleiche. you the same
298 Additive Particles: the Case of German Auch
In concluding let us point out that the difference in utterance meanings predicted in our approach is intuitively clear in many cases, but not in all. Thus, to us the auch and A UCH variants in ( 1 1 2)-( 1 1 3) seem practically equivalent in use, or the differences at least very subtle (though going into the direction we predict): (1 1 2) Beck hat den ersten Fall gelost. Beck has the first case solved
Beck solved thefirst case.
a. Wenn er auch den ZWEiten Fall lost . . . he also the second case solves If b. Wenn er den zweiten Fall AUCH lost . . .
(1 13)
he the second case also
If he also solves the second case . . .
solves
A. Ute kommt. Ute comes
Ute is coming.
B. Ich weiB, aber {auch SUsi kommt I know but {also Susi comes
I know, but Susi is coming, too.
I Susi kommt AUCH}. I Susi comes also}
But this should not be really surprising. After all, 'foregrounding' the partial identity between p and the proposition q it is added to does not make the difference between p and q disappear; it just becomes a matter of inference, and the same is true vice versa. Thus we can claim the existence of two utterance meanings related to the character of the material contained in the scope of Auch as further support for the correctness of our approach.
S S U M MARY
From the start it has been our purpose to demonstrate that there is only one and that the differences between its ± stressed variants are derivable from the interaction between syntax, semantics and focus structure. We have isolated and described the specific properties of this central additive particle and thereby shown that the standard approaches do not only miss the complementary pattern characterizing the stress behaviour of additive
Auch
particles relative to their AC, but also the relation between the syntax and semantic ofAuch, which is much more intimate than previously assumed: it
is the syntactic division into CC and the rest of the clause induced by Auch which yields the domain over which the adding operation characteristic for Auch is defined: what Auch adds to the corresponding part of a contextually
·
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
If
Marga Reis and Inger Rosengren 299
INGER ROSENGREN
MARGA REIS
Deutsches Seminar der Universitiit Tubingen Wilhelmstrafle 50 D-72074 Tiibingen Germany e-mail:
[email protected]
Tyska institutionen Lunds Universitet Helgonabacken 14 S-223 62 LUND Sweden e-mail:
[email protected] Received: os.o8.97 Final version received: 26.03.98
NOTES I Preliminary versions o f this paper were presented at the 'Sprache und Prag rnatik' meetings in Rendsburg I 995 and I 996, at the German and Nordic institutes of the University of Lund I 996 and I997, at the lnstitut fiir deutsche Sprache in Mannheim I 995, and at various meetings of the DAAD project 'Vom Satz zur Illokution' I 995
and I 996. We thank the audiences for valuable comments and suggestions. Special thanks go to Jochen Geilfui! Wolfgang and Manfred Krifka for reading and commenting on the semi final version. We also want to thank an anonymous referee for valuable remarks. 2. We shall use italics to mark the RC and
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
given proposition q is the material in its CC. At the same time, the relation betWeen the syntax and semantics of the Auch clause and _its focus structure turned out to be much more superficial than previously assumed, showing in effect that our modular approach to the grammar of Auch is correct. As for particulars, we have argued (i) that the complementary distribu. tion of auch vs. A UCH cannot be explained by a movement analysis or the like, as has been previously assumed; (ii) that Auch and similar particles are XPs adjoining to XPs of all kinds; (iii) that the Auch phrase {Auch + CC) always is an immediate constituent on a predicate projection line; (iv) that the meaning requirement of Auch that its host proposition p belongs to the same set of alternatives as some contextually given proposition q, is not implemented via the focus structure of p but via comparing p with q in context; (v) that the syntactic division of the Auch clause into the CC vs. rest of the clause plays a crucial role for its semantics: what Auch adds to the corresponding part of q is the material in its CC, which may be either the AC or predicative ID material; (vi) that the difference in CC material results in two utterance meanings, 'furthermore/in addition' and 'likewise'; (vii) that the accent pattern of Auch clauses follows from independent focus regularities, the focus on Auch having purely informational functions. In short, we have shown that what has been called 'focus particles' in recent years are in reality 'scope particles'.
300 Additive Particles: the Case of German capitals for stressed syllables whenever
necessary. The expression 'A > B' to be introduced below· is shorthand for 'A precedes B'. 3 Henceforth we shall use Auch as the particle name, also using auch for its unstressed and A UCH for its stressed variant. 4 For the time being we shall use as
(i). (In (ia) non-adjacency is forced by semantics: relative scope of Auch vis-a vis the scope-taking item. zwei Freunde; see section J . I . I ; in (ib) by syntactic reasons, the adjacent position being always ruled out; see section 2.3.2.)
(i) a. Er hat auch zwei Freunden He has also two friends
BUcher gekauft. (:rf Er
hat zwei books bought. ( He has two Freunden auch BUcher gekauft.) friends also books bought.)
b.
He also bought books for two friends. (:rf He also bought two friends books.) Auch ohne GEW war er. Also without money was he. auch GEW war er.) (*Ohne (Without also money was he.)
Neither had he any money.
we do not think that such a sharp
distinction
in
acceptabiliry
between
possible; hence, there is no reason why
there should not exist optional non adjacency as in (6). Thus, we consider Jacobs' principle of 'maximal proximity to the focus' ( I 98 3: 86) as a srylistic rather than syntactic principle. Note, too, that there are occasional examples with two ACs (as already observed by Anderson I 972; c£ (I I} below), as well as systematic discrepancies between pos sible ACs and CCs of Auch (in particular
with respect to head categories; see section 2.J. I}, which also argues against adjacency to the AC as a syntactically relevant notion. 7· Cases like (i)-(ii} (cited as counter examples to this generalization by a reviewer, the intended AC in (i)-(ii) being Napoleon) seem unacceptable to us in the intended interpretation.
(i)
Napoleon kam bekanntlich auch
mit nur drei Stunden SCHLAF pro Nacht aus.
As is well known, Napoleon, too, needed only three hours of sleep per night.
(ii) Auch Napoleon hat sich gelegent
lich wahrend einer Schlicht ein Stilndchen aufs OHR gelegt.
Napoleon, too, used to sometimes take a one hour's nap during a battle.
(i) is grammatical in the reading in which Schlaf is the AC, whereas (ii) is totally out, since Napoleon being the CC of Auch must carry the NA (see section 2.5).
8 The most important cases showing this are coordination (cases like Er sieht einen
'He is watching a movie or listening to CDs' would vio
Film oder hiirt CDs
late the condition that only constituents
can be coordinated, unless it is the
verbal trace that counts:
Er sieht; [einen Film t;] oder hort; [CDs t;]}, focus pro
obligatory and optional cases of non adjacency exists. In fact, it should not even be expected: since non-adjacency
jection (unlike most types of leftward
factors, non-adjacency is syntactically
and above all scope phenomena (with
can
be induced by exclusively semantic
movement, verb movement has no influence on possible focus domains),
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
examples only simple clauses where the proposition of the clause equals the proposition Auch operates on (ulti mately effecting its division into AC vs. ID). We thus avoid the well-known problems of so-called 'semantic scope'; see section J . I . I . 5 The nicht nur continuation indicates a possible co-text in accordance with the AC marked in italics. 6 Jacobs (I983: 8 3f� considers them as outright deviant, as opposed to cases of so-called obligatory non-adjacency like
Auch
Marga Reis and Inger Rosengren 30I co-textual restrictions (a clause begin ning with initial Auch must imme diately follow the proposition(s) to which the Auch proposition is added, initial Auch may never bear the NA) can be derived as consequences of its syntactic posmon interacting with independent principles (concerning the AC/ID division of Auch clauses and focus assignment). Otherwise it fits into all Auch generalizations formulated in this paper. This is reason enough to assign it to the same category as the other occurrence of Auch's considered so far. 1 2 Of course, there are complex ones like (i), which follow the usual regularities.
(i) [Er WOLLte es nicht nur tippen,) er TIPPte es auch/HAT es auch getippt, He didn't only intend to type it, he did in fact type it. Reconstruction comprises, ofcourse, all kinds of finite verbs (main verbs as well as auxiliary verbs). However, as will become evident in section 3, only the reconstruction of the main finite verb counts as semantically relevant. 9 In these examples the material to the right of auch is totally AC-material (which is not necessary; c£ er hat einen Anzug gekauft und dann seiner Mutter auch ein KLEID gekauft 'he bought a suit and then he bought his mother a dress, too'). We shall keep to this option here, since what is important is only that there must be AC material to the right of auch. I o In the following, we will refer to all clausal projections as CPs. The question whether CP is to be split up or whether all clauses correspond to the same functional projection is immaterial to our discussion. I I In former analyses, initial Auch (i.e. Auch occurring all by itself in preverbal position) has often been set apart as belonging to a different category. But its obvious peculiarities regarding
·
(i) DaB jemand mein AUto That someone my car angefahren hat, hat mich AUCH hit has, has me also geargert. annoyed That someone hit my car, annoyed me, too. I 3 This assumption is common (see also Hoeksema & Zwarts I 99 I ), but rarely as explicitly stated as in Jacobs ( 1983: I 04). (Although his analysis is formulated in a different framework, it is undeniably in the spirit of the 'AC movement' approach to post-posed particles, and subject to the same criticism.) On the whole, the tendency to take the priority of the 'unstressed particle > stressed AC' pattern for granted is coupled with either not dealing with the stressed variants of additive particles at all, or reconciling them ad hoc with the pre vailing pattern of analysis by claiming that in the case of AUCH, TOO, etc. focus (lying on the AC) and stress Qying on the particle) are divorced (see inter alia Konig I 99Ia, b, I993), which clearly presupposes a kind of 'AC movement' analysis. The only thorough study of the problems raised by this analysis of post- · posed additive particles is Kowalski
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
respect to scope-taking elements the fronted finite verb always behaves as if it were in their scope, i.e. in final position, see Hohle I988: 7). Since Auch (and .so-called focus particles in general) are also scope-taking elements, data like ( 1 2) are certainly no surprise. Note that this also holds for the so called 'Verum' element (Hohle I 988, 1 992), although this is associated with the left-peripheral verb position and not with the finite verb itsel£ Thus, cases like (i) where Verum functions as the AC of auch and necessarily pre cedes it, are not true exceptions to the regularities in question either.
302 Additive Panicles: the Case of German Auch I 9 This positional restriction takes care of a number of much-discussed dif ferences between auch and A UCH (see Schwarz I 99I; Kowalski I992: 66ff), and also of the distinctive specificity effects discussed in this connection, see (i): {i) a. Theo hat auch {einige/keine} Theo has also {a few/ no } LIMericks gedichtet. Limericks composed Theo also wrote a Jew limericks/Theo didn't write any limericks either. b. *Theo hat {einige/keine) Limericks AUCH gedichtet. If we take into account in (a) that the position of indefinite DPs including their existential reading must be after Nicht/ neg (possibly within the smallest VP, see Pafel I99sb), whereas the position before Nicht/neg induces the specific readings, (b) that the lowest possible position of Auch is before Nicht/neg-note that we are not talking about the higher negation position active in some clause types, see note 2o-then it follows that Auch must precede an AC having the existential reading in the middle field, which rules out the stressed variant. (The fact that (ib) is out is a natural result of the object being specific, and dichten being an 'effec tive' verb: specific objects cannot be created more than once). 20 In some clause types, notably condi tional clauses, the ordering nicht > Auch is acceptable, c£ wenn nicht auch PETer zustimmt, /wenn Peter nicht A UCH zustimmt, ('if Peter does not give his ok, too,'). Since there is probably ·no way around assuming a second negation position higher up in these clauses, this does not impinge on the above generalization. Cases like Dafi er nicht auch am Donnerstag gekommen ist, ist erstaunlich ('that he didn't come on Thursday, too, is amazing') cited to us by a reviewer can most likely also be subsumed under this type of negation. 2 I This is also the traditional view (most
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(1992), which we will comment on below. 14 Regarding PRO, this is common knowledge. Concerning pro in impera tive clauses, see Platzack & Rosengren (1998), where it is also maintained that this element does not move. 1 5 Note that (24)-(25) do not testify to an essential difference between auch and AUCH after all: as to (24), the impos sibility of *[auch PROJ etc. follows from the fact that unstressed auch to the left of its AC is only licensed by an NA on its AC, which empty elements cannot realize. Concerning the fact that auch does not combine with + wh operator phrases (nor other SpecC operator phrases), no explanation has come for ward yet (see Taglicht 1993: 1003). but the likely options for such an explana tion seem all to be independent from the Auch variation. 16 In the eighteenth century immediate postposition of auch was still possible, c£ Der uble Weg auch hatte ihn verdriefilich gemacht 'the bad road, too, had made him moody' (Goethe, Leiden des jungen Werther, p. 145). This suggests that its degraded status in present-day German (see also Jacobs I983: 99) as such has no systematic linguistic causes. 17 Including this example was motivated by an argument by M. Krifka (p.c.) in favour ofa Kowalski-type analysis, which turned on examples involving weak pronouns. I 8 That Auch, and focus particles in gene ral, may take heads as its CCs is assumed in more surface-oriented analyses (see especially the treatment of so-called 'ad-adnominal' and 'ad article' cases in Jacobs I983: 64ff), but sometimes also assumed elsewhere with regard to verbs (see Konig I993: 98 3). We concentrate on the verbal head because it is the only head that syntac tically occurs all by itself, hence there could be no additional syntactic restric tion preventing Auch from adjoining to it, if Auch were able to do so.
Marga Reis and Inger Rosengren 303
'
(i)
(v "' auch (v "' der NEUe Professor also
the new
professor
(v " hat (v ' schon ein Haus has
already a house
gekauft.])]] bought
The new professor, too, has already bought a house. (ii) *Auch der neue Professor hat schon em HAUS gekauft / ein Haus geKAUfT. (iii) Auch nach PaRIS, glaube ich, Also to Paris think I daB er fahren will. wants that he go
To Paris, too, I think that he wants to go.
None the less, we think that there is a valid intuition behind Jacobs' position, which we propose to capture in terms of an independent division rule for Auch domains (see below), also taking care of the syntactic restrictions cited by Jacobs (1983: 45 ff) in support of his view. 22 This means that Auch also takes com plementizer-introduced CPs as well as V2-CPs. Note that the initial Spec position may also be regarded as an adjunct position (c£ Kayne 1 994). 23 Note that DP nominalizations may also contain auch; c£ die TeiltJ4hme/Verleum dung auch des Chefs ('the participation/ denouncing of also the boss'), although these are somewhat marked. More marked, though still at least half-way acceptable in many cases, are DPs like Fotos auch von Peter ('pictures also of Peter') or das Haus auch des Chefs ('the house of also the boss'), whereas auch within DPs like der Inbegri./f(*auch) eines Genies ('the epitome of a genius'} are out. Since nominalizations correspond to propositions, and possessive phrases to predications, whereas (Inbegri./f+ XP] does neither, this is in keeping with what we observed above. Note, too, that DPs like Fotos (?auch) von Peter are much better than Fotos (??auch) Peters ('pictures also Peter's'). This corre sponds to PP attributes being much less cohesive than genitive attributes, which also shows up in their ability to undergo DP extraction; c£ von Peter/*Peters habe ich (auch)Fotosgesehen ('ofPeter/Peter's have I pictures seen'). Since movement, as a rule, is a privilege of immediate con stituents on predicate projection lines, this is again in keeping with the above. 24 Note that stressed A UCH does not cooccur with CPs, which are extended V projections; c£ (i): (i) *AUCH (hat Paul Johanna auf den Aufsatz aufmerksam gernacht]. This, however, has a simple reason: as already pointed out above, A UCH can
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
explicitly formulated in Altmann 1976, 1978: 84ff). The alternative position maintained by Jacobs (1983: 4off) Auch takes only V projections (with the V2 clause also counting as such), which leads to V2 clause structures roughly like (i)-cannot be upheld for mainly two reasons (see Konig 199 1 a: 23ff; Bayer 1 996: 1 9ff): (a) (i) violates the V2 constraint, for the CC of auch is not the preverbal constituent (moreover, this auch can also combine with excep tional 'second' preverbal constituents as in Auch PEter aber/indessen wufite . . . ('also Peter however knew . . . ') showing that preverbal auch is not on a par with them), (b) the structure assigned to (i) by Jacobs wrongly implies that auch-AC patterns like (ii) should be good, for according to the structure in (i) the AC in (ii) is c-commanded by the auch in question. Note also that preverbal auch constituents may be the result of extraction (iii), which is also inexplicable under Jacobs' analysis, for extraction applies to just one constituent (whereas for Jacobs auch nach Paris would be a string of constituents). Obviously, the traditional syntactic analysis has no problem in either case. (Concerning a further argument from so-called thetic cases, see note 4 3 ).
3 �4 Additive Particles: the Case of German Auch
Verleumdung} (*AUCH) seines Chefs (*AUCH). We think that this is related to the fact that Auch may only be
stressed, if its CC is a true predicative projection, which implies that the predicative head must always be in the CC. In the DPs in question this require ment simply cannot be fulfilled without violating others (in particular that there be an AC to the left in the sa,me projection). 27 To these restrictions may be added the observation by Bayer (1996: 1 5 7fT) ·that PP objects of adjectives, which have their unmarked position to the left but may also be placed to the right of the adjective, are considerably worse in this position, if combined with auch. (i)
a.
daB Peter unseres ErachtenS that Peter our opinion (gen.] {auch auf seinen SOHN stolz/ /?stolz {also on his son proud/ proud auch auf seinen SOHN} war also �m his son} was
that in our opinion Peter his son, too
was
proud of
Since movement in the middle field is always to the left, and stolz (auch) auf
seinen Sohn forms a constituent in the initial field, it is quite likely that this constituent is still a P projection. But then it is also quite likely that the same constraint outlawing Auch within PPs and DPs is at work here. As for the remaining asymmetry concerning Auch + CP mentioned by Jacobs (I983: 47ff), see section 2.4.2. 28 The similarity between 'adsententials' and focus particles has already been clearly worked out by Jacobs (I983: 59); in fact it determines his analysis. What is missing in his analysis is that it does not pay attention to the syntactic difference between focus particles and those 'adsententials', which is, first and foremost, that focus particles take all kinds of XPs as CCs, whereas adsenten tials obey the additional condition that their CC must be a predicate projection. 29 See notes 2 3 and 26. Perhaps not all ofthe examples of focus particles within DPs, PPs, etc. that have come to light over the years (for an overview c£ Hoeksema & Zwarts I 99 I : 59ff and Bayer I 996) can be accounted for this way. But the remajning exceptions are often marginal (especially concerning Auch within PPs), and do not seem to follow a systematic pattern within a given language, nor has it been convincingly shown so far that language-specific differences in viola tions of (48) are a matter of linguistic principle rather than just convention. 30 for a critical review of the previous attempts by Jacobs (I98 3), Rooth (198 5) and Bayer ( I 996) (which is a comprehen sive version of his earlier work), see Hoeksema & Zwarts (199 1 : 59ff) .and Kowalski (I992). Jacobs' and Rooth's approach is extensively reviewed in Bayer (I 996: I 9ff); see also note 2 1. 3 I Note that our division rule will also exclude (5 3b) provided we subscribe to traditional theories of extraposition. Given Haider's (1993) theory of extra position, however, which is not implau sible, extraposed constituents are also
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
only take ID material as its overt CC. In the case of (i) the whole clause would have to be ID material, which is communicatively unacceptable in itself and opens up no possibility to infer alternating ACs. Hence (i) is out. 25 Examples like these were cited by a reviewer as counterexamples to our claim that A UCH may not take straightforward DPs/PPs, etc. as CCs, but only projections that are predica tively used. As is evident &om above, they are not counterexamples to our claim, but rather confirm it. 26 Note that stressed AUCH is licit in strictly clause-like projections only; in contrast to auch (see note 23), it never appears within DPs, not even in norninalizations. c£ the counterpart to the example cited there: die {Teilnahme/
Marga Reis and Inger Rosengren 305 counterexamples
immediate constituents of the predi ·
too-see
thus
36 Stressed modal particles and Verum
focus might at first glance look like
accessible
to
focusing.
(Being
cannot carry an accent. Thus,
aber
in
clause. 3 8 It has been suggested to us (M. Krifka, p.c.) that ADD is not necessary since it is
possible to derive the corresponding
effects from the interaction between two components of the meaning of
Auch:
the assertion of p and the impli
cature of an existing alternative q to
p.
This may be so, but our argument was that independent evidence-focusing and negation-forces us to assume ADD
anyway. Once available, it must of course be taken to be the carrier of the two
meaning components in question. 39 Note that the meaning part of
Auch
which we assume to be ADD, is also a
meaning part of other elements, in ·
particular conjunctions like
und
(see
section 3.1 .2). But these conjunctions lack the rest of the meaning of
·
Auch,
which is that the addition must be to a
proposition q in the same AS. 40 For further details we refer to the relevant
literature as it is found in Jackendorff (1 972), Gussenhoven (1984). Culicover
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
No, Hans won't come probably, he will most certainly come.
of modal
(70) may carry secondary stress, if the NA lies on foul. Auch, however, can carry the NA as the only accent of the
er kommt ganz sicher.
quite certainly.
descriptions
these elements cannot carry the NA of the clause. This does not mean that they
B: Nein, Hans kommt nicht vielleicht, No, Hans comes not probably, he comes
the
accessible to focusing does not imply being accessible to negation, as shown by the MPs, which due to their different function always take scope over the negation.) 3 7 Note that what we maintain is that
metalinguistic negation, i.e. as denial that using the word Auch is appropriate
(i) A:. Hans kommt vielleicht. Hans comes probably. Hans will probably come.
that
the declarative sentence mood operator (see inter alia Rosengren 1993a)-and
ich bin erstaunt, dafi Hans nicht A UCH kommen will ('I am amazed that .Hans
will not really do: metalinguistic nega tion is always followed by an appro priate rewording of what was negated, c£ (i), which is not the case in (7 1 )).
claim
particles in Ormelius-Sandblom (1 997) and the Verum element introduced in Hohle (1988, 1992) and identified with
wants to come.-No, that's not true') are preferably read as denials of p. As for cases with clause internal negations like
doesn't want to come either') (see note 20), they are always read as denials of p. However, the fact that ± stress on Auch leads to different interpretations in one and the same negative environment remains, calling for an explanation. (Note that accounting for the 'denial ' of q effect of the negation in (7 1) as
the
that what is focused in these cases are non-implicated meaning elements,
acceptable. 32 Additional evidence is offered by the so-called thetic cases; see section 4.2. 3 3 Recall that the finite verb is always
reconstructed (see section 2. 1. 1 . 1 ), thus counting as overt material. 34 The one apparent exceprion-thetic cases, where Auch takes a DP but the AC comprises the entire constituent will be dealt with section 74.2. 3 s This effect . is apparently confined to negations expressed by matrix expres sions embedding the Auch clause. As has been pointed out by a reviewer, dis. course denials like in Haus will AUCH kommen.-Nein, stimmt nicht ('Hans, too,
to
implicated meanings cannot be focused. But on closer inspection, it turns out
cate projection line; thus by (48) extra posed Auch CPs should be syntactically
306 Additive Particles: the Case of German Auch
41
42
44
meanings going along with auch vs. Auch predicted by our theory (see sec tion 4-3) is by and large home out. 45 As is well known, Auch also allows for scalar utterance meanings aacobs 1 983: 1 49; Ki:inig 1991a), but only in the case of unstressed auch; c£ (i) vs. (ii): (i) Auch der Klassenprimus schrieb Also the class-primus wrote eine 5· 5 a
Even the best student in class got a D.
(ii) Der The eine a
Klassenprirnus schrieb AUCH class-primus wrote also 5· 5
The best student in class got a D.
The scalar meaning in (i)-the best student of the class being the least likely to also get a D in the exam-is triggered by the lexical meaning of Klassenprimus, which suggests ordering the alternative values to which the AC belongs, on a scale. Since this meaning may only arise if the variable part is lying in the scope of Auch, this might constitute indirect evidence for positing rwo 'basic' utterance meanings, the scalar interpretation being naturally compatible with 'in addition' and incompatible with 'likewise', but we will not pursue the question here. 46 Much the same point concerning the semantics of too is made by Goddard (1 986: 63 8f).
6 . REFERE N CE S Abney, S. (I 987), 'The English noun phrase in its sentential aspect', Ph.D. dissertation, MIT, Cambridge, MA. Altmann, H (1976), Die Gradpartikeln im
Deutschen. Untersuchungen zu ihrer Syntax, Semantik und Pragmatik, Niemeyer, Tiibingen (= Linguistische Arbeiten 33). Altmann, H ( I 978), Gradpartikel-Probleme. Zur Beschreibung von gerade, genau, eben,
ausgerechnet, vor allem, insbesondere, zumindest, wenigstens, G. Narr, Tiibingen
( Studien zur deutschen Grammatik 8). =
Anderson, St. (1 972), 'How to get even', . Language, 4 8 , 893-906. Bayer, J. (1 996), Directionality and Logical
Form: On the Scope of Focusing Particles . and Wh-in-situ. Kluwer, Dordrecht. Biiring, D. (1997), The Meaning of Topic and
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
43
& Rochemont (1983), Selkirk (1984), Rochemont ( 1 986), Uhmann ( 1 991), Hetland (1 992), Rosengren (1993b, 1994), Winkler (1997), and many others. A further case where the function of focus is similar to that of MPs and Auch may be so-called Verum focus (see Hohle 1 992, where a number of addi tional focus functions differing form the typical function of evoking alternatives are discuSsed). As for auch being part of the focus domain, see the comment to (9o) in the text. Jacobs (198 3 : 89f, I986: 1 2 5, note 2 I ) has used examples like these as an addi tional argument that, given the c-com mand relation berween Auch and its AC, clauses like these are V projections, which implies giving up the V2 con straint for German (see note 2 1). As we maintain above, it is just the other way around: we should not give up the V2 constraint, but the c-command condi tion on the Auch-AC relation. Regarding accents on Auch, we had the opportunity to check through prosodi cally transcribed material (a transcript of a reading of 'Die Leiden des jungen Werther' containing I 16 occurrences of Auch), for which we are indebted to Gregorz Dogil and his crew (University of Stuttgart). Our fmdings are fully in line with the views stated above: there are secondary accents on Auch having the specific functions in question; more over, the distribution of utterance
Marga Reis and Inger Rosengren 307 Jacobs, J. (I98 3), Fokus und Skalen. Zur Syntax und Semantik der Gradpartikeln im Deutschen, Niemeyer, Tiibingen J). (= Linguistische Arbeiten 138). Culicover, P. & Rochemont, M (198 3), Jacobs, ]. (I984). 'The syntax ofbound focus 'Stress and focus in English', Language, in German', Groninger Arbeiten zur S9. 1 23-65. Germanistischen Linguistik, 25, 1 72-200. Fintel, I(; von (1995), 'A minimal theory of Jacobs, J. (I986), 'The syntax of focus and adverbial quantification', MS, MIT. adverbials in German', in W. Abraham & Geilfui.\-Wolfgang, J. (1996), Uber gewisse S. de Meij (eds), Topic, Focus, and Con Fiille von Assoziation mit Fokus, Niemeyer, figurationality, Benjamins, Amsterdam, Tiibingen (= Linguistische Arbeiten 358). I03-27. Goddard, C. (1986), 'The natural semantics Jacobs, J. (I988). 'Fokus-Hintergrund of too', Journal of Pragmatics, IO, 63 5-44· Gliederung und Grammatik', in H. Green, G. M. (1968), 'On too and either, and Altmann (ed.), Intonationsforschungen, not just on too and either, either', CLS, 4, Niemeyer, Tiibingen, 89- 1 3 4. (= Lin 22-39· guistische Arbeiten zoo). Green, G. M. (1973), 'The lexical expression Jacobs, J. {I993), 'Integration', in M. Reis of emphatic conjunction: theoretical (ed.) { I 993), 63-1 I 6. implications', Foundations of Language, Jacobs, J. {I 997), '1-Topikalisierung', Lin IO, I97-248. guistische Berichte, r68, 9 I - I 3 3· Grice, P. (197 5), 'Logic and conversation', in Kaplan, J. (I 984), 'Obligatory too in English', P. Cole & J. Morgan (eds), Speech Acts, Language, 6o, 5 I o- I 8. Syntax, and Semantics, Vol. J, Academic Kayne, R (1 994), The Antisymmetry of Press, New York, 41-58. Syntax, MIT Press, Cambridge, MA and Grimshaw, J. (199 I), 'Extended projection', London. MS, Brandeis University. Konig, E. (199 1 a), The Meaning of Focus Gussenhoven, C. (1 984), On the Grammar Particles: A Comparative Perspective, and Semantics of Sentence Accents, Foris, Routledge, London and New York. Dordrecht Konig, E. (199 1 b), 'Gradpartikeln', in A. von Haider, H. (I993), Deutsche Syntax-generativ. Stechow & D. Wunderlich (eds), Seman Vorstudien zur 'Theorie einer projektiven tics: An International Handbook of Con Grammatik, Narr, Tiibingen. temporary Research, de Gruyter, Berlin, Hetland, J. (1 992), Satzadverbien im Fokus, 786-803. Narr, Tiibingen (= Studien zur deutschen Konig, E. (I993), 'Focus particles', in J. Grammatik 43). Jacobs et al. (eds), Syntax: An International Hohle, T. N. (I988), 'VERUM-Fokus', Handbook of Contemporary Research, Vol. Sprache & Pragmatik, 5, 2-7. I , de Gruyter, Berlin, 978-87. Hohle, T. N. ( I992), 'Dber Verum-Fokus Kowalski, A. (I992), 'Zur Syntax und im Deutschen', in J. Jacobs (ed.), Informa Sernantik von Gradpartikeln im tionsstruktur und Grammatik, West Deutschen. Syntaktische Bedingungen deutscher Verlag, Opladen, I I 2- I 4 I . fiir die Ausdehnung von Partikelskopus (= Linguistische Berichte, Sonderheft 4 ). und -fokus', Magisterarbeit, Universitit Hoeksema, J. & Zwarts, F. (I 99 I), 'Some Wuppertal. remarks on focus adverbs', Journal of K.ri£ka, M (I992), 'A compositional Semantics, 8, S I -70. semantics for multiple focus construc Jackendoff, R S. (i972), Semantic Interpreta tions', in J. Jacobs (ed.), Informations tion in Generative Grammar, MIT Press, Cambridge, MA. struktur und Grammatik, Westdeutscher Focus: The 59th Street Bridge Accent, Routledge, London and New York (= Routledge Studies in German Linguistics
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
308 Additive Particles: the Case of German Auch Verlag, Opladen, 1 7- 5 3 (= Linguistische
Berichte, Sonderheft 4).
Krifka, M (1995), 'Focus and/or context: a second look at second occurrence expressions', MS, University of Texas at Austin. Lambrecht, K. (1 994), Information Structure
and Sentence Form: Topic, Focus, and the Mental Representations of Discourse Referents, CUP, Cambridge (= CSL 7 1).
konzeptuelle Verschiebung. Studien zur Syntax und Semantik deutscher Modalpar tikeln, Niemeyer, Tiibingen: (= Linguis tische Arbeiten 314). Moln:ir, V. (1991), Das TOPIK im Deutschen und im Ungarischen, Almqvist &
Wiksell International, Stockholm
(= Lunder germanistische Forschungen 58).
Moln:ir, V. (1993), 'Zur Pragmatik und Grammatik des TOPIK-Begriffes', in M Reis, (ed.), (1993), 1 5 5-202. Moln:ir, V. & I. Rosengren (1997), 'Zu Jacobs' Explikation der 1-Topikalisier ung', Linguistische Berichte, 169, 2 1 1-47· Moser, M. G. (1992), 'The negation rela tion: semantic and pragmatic aspects of a relational analysis of sentential nega tion', Ph.D. dissertation, University of Pennsylvania. Muller, G. (1995), A-bar Syntax: A Study in Movement Types, de Gruyter, Berlin & New York (== Studies in Generative
Grammar 42).
Orrnelius-Sandblom, E. (1997), Die Modal partikeln ja, doch und schon. Zu ihrer Syntax, Semantik und Pragmatik, Almqvist & Wiksell International, Stockholm
(= Lunder germanistische Forschungen 6 I ).
Pafel, J. (1995a), 'Kinds of extraction from noun phrases', in U. Lutz & J. Pafel (eds),
On Extraction and Extraposition in German,
Benjamins, Amsterdam, 145-77Pafei, J. (1995b), 'Die "kleine VP" und ihre Bedeutung fur Extraktion und Fokus projektion', MS, Tiibingen.
Comparative journal of Germanic Linguis tics, I, 1 77-224. Pollock, J. Y. (1989), 'Verb movement, universal grammar, and the structure of IP', Linguistic Inquiry, 20, 365-424. Reis, M. (ed.) (1 993), Wortstellung und Infor mationsstruktur, Niemeyer, Tiibingen. (= Linguistische Arbeiten 306) Rochemont, M (1 986), Focus in Generative Grammar, Benjamins, Amsterdam. Rooth, M. (198 5), 'Association with focus', Ph.D. dissertation, Amherst, University of Massachusetts. Rooth, M. (1992), 'A theory of focus inter pretation', Natural Language Semantics, 1, 7 5- 1 1 6. Rooth, M. (1996), 'Focus', in S. Lappin, (ed.),
The Handbook of Contemporary Semantics,
Blackwell, Cambridge and Oxford, 271-97· Rosengren, I. (1993a). 'Riickblick und Aus blick', in I. Rosengren, (ed.), Satz und Illokution, Vol. II, Niemeyer, Tiibingen, 2 5 3-60 (= Linguistische Arbeiten 279). Rosengren, I. (1 993b), 'Wahlfreiheit mit Konsequenzen. Scrambling, Topikali sierung und FHG im Dienste der Infor mationsstrukturierung', in M. Reis (ed.) (1993). 2 5 1 - 3 1 2. Rosengren, I. (1994), 'Scrambling-was ist das?' in B. Haftka (ed.), Was determiniert Westdeutscher Wortstellungsvariation? Verlag, Opladen, 175-96. Rosengren, I. ( 1 997), 'The thetic/categorical distinction revisited once more', Linguis tics, 3 5, 439-79· Rothstein, S. ( 1991 ), 'Heads, projections, and category determination', in K. Leffel & D. Bouchard (eds), Views on Phrase Structure, Kluwer, Dordrecht, 97- 1 1 2. Schwarz, B. ( 1 99 1 ), 'Some observations on German auch', MS, Saarbriicken. Selkirk, E. (1984), Phonology and Syntax: The Relation between Sound'and Structure, MIT Press, Cambridge, MA. Stechow, A. von (1991), 'Current issues in
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Lenerz, J. (1993), 'Zur Syntax und Semantik deutscher Personalpronomina', in M. Reis, (ed.), (1993), 1 1 7-53. Meibauer, J. (1 994), Modaler Kontrast und
Platzack, C. & Rosengren, I. (1 998), 'On the subject of imperatives: a minimalist account of the imperative clause',
Marga Reis and Inger Rosengren 309 the theory of focus', in A von Stechow & D. Wunderlich (eds), Semantics: An
nationales Handbuch zeitgenossischer For schung. Vol. 1 , de Gruyter, Berlin and
International Handbook of Contemporary Research, De Gruyter, Berlin, 804-25.
New York, 998-1006. Uhmann, S. (199 1), Fokmphonologie, Nie meyer, Ti.ibingen (= Linguistische Arbeiten
Groninger Arbeiten zur germanistischen Linguistik, 25, 223-63.
Winkler, S. (1 997), Focus and Secondary Predication, Mouton de Gruyter, Berlin and New York (= Studies in Generative
Stechow, A von & Uhmann, S. (1984), 'On the focus-pitch accent relation',
Taglicht, J. ( 1993), 'Focus and background', in J. Jacobs et al. (eds), Syntax. Bin inter-
252).
Grammar 43)
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
© Oxford University Press
Journal oJ&rtlllntia 14= ) 1 1-) 1 8
1997
Book Review Diego Marconi Lexical Competence. MIT Press, Cambridge, MA, 1 997, xiii + 2o6 pp. ISBN o-262 - 1 3 3 3 3 -4. $25 (cloth).
ALBERTO VoLTOLINI
rv
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
In this clear, lucid, and solidly argued book, Marconi presents a dual theory of lexical competence, according to which knowing the meaning of a word is being able both to draw inferences from sentences containing that word to other sentences and to make both the world fit the word and vice versa: in other terms, to be both inferentially and referentially competent. The inferential aspect of lexical competence shows itself in the fact that we would not regard someone as competent e.g. on 'bachelor' if they did not recognize the validity of inferences such as John is a bachelor' . · . John is not married'. Traditionally, one is considered to be inferentially competent if one has command of Carnapian meaning postulates of the kind '( Vx) (bachelor (x ) ::) married (x))', which validate such inferences. Carnap, however, chose meaning postulates to represent inferential competence because he wanted them to account for the notion of analiticity and therefore to support the analytic/synthetic distinction (a/s). Once this distinction is given up, as Marconi does with Quine (p. 30), there is no longer any need to represent inferential competence via meaning postulates. What replaces meaning ·postulates in Marconi's picture of inferential competence is a weak form of molecularism of comprehension which does not presuppose a/s. According to this form, it is necessarily the case that understanding a sentence p entails that for every term t of p there is some t-sentence or other q different from p which is understood as well (for instance, in order for me to understand 'gold is a metal', I must understand another 'gold'- as well as another 'metal'-sentence (not two specific ones) (pp. 54-5)). For him, this form of molecularism is an amendment of a related version of this doctrine, which is discarded because it implausibly accounts for inferential competence by linking the understanding of a sentence p to the understanding of an utterly different sentence q (p. 5 3). One might wonder, however, whether Marconi's own molecularism does not suffer from the same drawback as the version he criticizes. Suppose someone understood 'this is the bank close' as long as this person understood both 'small boats veer at this bank' and 'the day is drawing to a dose', which share with the former sentence nothing but the more
3 1 2 Book Review .
undoubtedly perceived to be semantic, to encyclopaedic sentences, whose truth is regarded as merely factual. As often happens with taxonomies, even this one seems questionable. For instance, Marconi takes sentences which are particular, contingent, and perceived as constitutive of normal competence as more encylopaedic than dictionary-like (p. 43). This may be the case with his own example ('France
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
morphosyntactic shapes 'bank' and 'close'. Nevertheless, according to Marconi's own molecularism, such a person would eo ipso have to be regarded as competent on 'bank' (and 'close'). Marconi would probably dismiss this alleged counterexample by saying that, in so far as an ambiguous word counts as different words ('bank/ and 'bank2', say), no sentential connection such as the above may legitimate an attribution of competence on either disambiguated word. In order for him to give such a reply, however, he would have to assume that the criterion of identity for a word included a morphosyntactic shape plus its semantic interpretation. But this assumption is precluded by the fact that he also claims that inferential competence regards only knowing relations among uninterpreted words: 'it is incorrect to require or presuppose that such relations involve not the words themselves but their meaning' (p. s s ). Another (unnoticed?) consequence of this weak molecularism is that inferential competence is intransitive. According to Marconi's alternative formulation of his molecularism given in terms of belief possession rather than of sentence understanding, necessarily two individuals share the same belief p about a 1 • • • an only if for any number i such that 1 :S i :S n, there is some ai -belief (i.e. a belief about ai) or other q which those individuals share (p. 5 3). Now, suppose that A has the following five beliefs: (i) that gold is a metal, (ii) that gold is expensive, (iii) that gold is mainly mined in Mrica, (iv) that a metal is a chemical element, and (v) that a metal can be found in nature as a mineral in its native pate. .Moreover, suppose that B has only beliefs (i), (ii), and (iv), whereas C has only beliefs (i), (iii), and (v). Thus, if we stick to Marconi's aforementioned formulation it turns out that A and B share the belief that gold is a metal as well as A and C, but B and C do not. This consequence may sound unhappy if one moves from the interpersonal to the intrapersonal level, as Marconi allows us to do (p. 54). For by parity of reasoning it may be the case that me1 (=me at time') and me2 (=me at time t") share the same beliefp as well as me1 and me3 (=me at time t"'), but not me2 and me3 • As we said before, Marconi rejects a/s. To be sure, for him one must not overlook the indubitable fact that there are sentences recognized as analytic. This may be taken into account by constructing 'being recognized as analytic' as a scalar predicate (pp. 41-4). In this vein, he puts forward a taxonomy which leads from dictionary-like sentences, whose truth is
Journal of Semantics 3 1 3
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
is a republic'). But take any token whatsoever of ' I am the present speaker'. Undoubtedly, such a token is not only both particular and contingent, but also perceived as constitutive of normal competence regarding 'I'. The latt(!r feature depends on the fact that the concept being the present speaker is . constitutive of the Kaplanian character of 'I', i.e. the linguistic meaning of this indexical which ftxes its reference in any context of its utterance. This constitution, however, makes it the case that this token is also a priori.1 Thus, unlike 'France is a republic', this token is more dictionary-like than encyclopaedic, as is witnessed by the fact that dictionaries contain the entry 'I: the present speaker'2 (in this respect, tokens of the above sentence are unlike other analytic tokens such as 'if this patch is red, then this patch is not green' which, as Marconi rightly suggests, 'are out of place in a dictionary' {p. 42, fn. 20)). It is precisely because, as we saw above, inferential competence links only words with words that Marconi feels that lexical competence as a whole is something more than inferential competence. In this vein, he remarks that not only do meaning postulates not exhaust lexical meaning {for they are unable to supply symbols with a semantic interpretation), but they do not exhaust lexical competence either. Nor will translation of a sentence into an interpreted formula of a formal language be sufficient (i.e. a formula plus its truth-conditions), which gives that sentence its semantic interpretation. For when we understand a sentence we know more than semantic interpretation cum-meaning postulates; 'we know . . . how to apply [the sentence's] words in the real world' (p. 19). This knowledge is precisely one of the abilities constituting the other factor of lexical competence, namely referential competence. Marconi acknowledges that sometimes referential competence is mediated by inferential competence; one may be able to apply a word w ' derivatively, that is, on the basis of the fact that one is able to apply a word w" which is inferentially linked to w ' (p. 62). Still, he stresses that inferential and referential competence are separate skills, for one may have one without the other {pp. 5 8-9, 67-8). To be sure, Marconi's final picture of the relationship between referential and inferential competence is more complex. First he flanks the afore mentioned word-world activity of ftnding real items corresponding to a certain verbal label, application, with the converse world-word activity of ftnding a term for an object, naming. Marconi assigns naming to referential competence as well (p. 69) . Moreover, naming is distinguished by recognizing, i.e. the ability of knowing one's way about an object even while ignoring its name (p. 70); still, the former presupposes the latter (p. 71 ). Since recogniz ing may also take place without word application, it turns out that one may access the conceptual, or semantic, lexicon even while failing to access the output lexicon. One can therefore access an inferential network and thus
3 r 4 Book Review
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
mobilize inferential competence even without starting the relative inferential chain from an output word (ibid.). Marconi is explicit as to what makes one apply a word, or even an image associated to that word, to a worldly item: a general rule of application (something like a Kantian schema) (pp. 1 45-9). In so far as this rule operates unconsciously (and presumably also in a 'mechanical' way), he is able to address the related question as to how an artificial system may be credited with referential competence over and above inferential competence. In systems of artificial vision, he writes, the role played by the above procedural rule is taken over by an algorithm matching a relational structure associated with the word (i.e. a labeled graph) with the relevant portion of a scanned scene (p. I so). To be sure, Marconi remains neutral as to whether those artificial systems are to be credited with the same kind of referential competence as humans (p. I 6o). A reason for this neutrality may be that it is not altogether clear whether algorithms matching scanned scenes really implement the system's inferential competence in the way required in order to ascribe to the system a genuine lexical competence. What if a system scanned two scenes but matched algorithmically the same relational structure with both scenes, although what those scenes contained were utterly different, albeit indistinguishable, portions of reality? Marconi notices this problem, but tends to limit its scope. To his mind, it regards only cases which at most involve concepts whose instances cannot be differentiated on the basis of their external appearance (e.g. men and uncles: p. I 5 s) or where the indistinguishable realities are Fs and false Fs (e.g. pistols and toy-pistols (ibid.)). The latter case is however more complex than the former one. If something is individuated as a false F, it is singled out as a representational entity: namely, as something whose nature consists in representing something else, in being the image of the latter (pistols, in the above example). This representational nature -not only is not, but also cannot be grasped by an artificial system which, when faced with a false F, erroneously sees that there is an F over there. This inability looks more troublesome when the system is faced with an ambiguous representational entity, i.e. something which is both a false F and a false G and is accordingly perceived either as an F or as a G (e.g. the duck-rabbit picture). For the system is not only unable to tell this entity both from Fs and from Gs, but also moves back and forth from the erroneous perception that there is an F over there to the erroneous perception that there is a G over there even though nothing in the external appearance of what it is perceiving changes. As a consequence of the distinction between inferential and referential competence, as regards one and the same word nothing prevents anyone from a mobilizing a usual inferential chain but applying it unusually, or vice versa (pp. 78-9). In a certain sense, this abnormality does not affect
Journal of Semantics 3 1 5
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
competence. Indeed, inasmuch as both inferential and referential com petence are to be conceived ir:zternistically (p. 6I), i.e. as 'the complex of knowledge and abilities' 'in the head' 'that make it possible for [a] person to use the language the way she uses it' (pp. I 30- I ), we are left with nothing but a (possible) variation in individual competences (p. s6). In another sense, however, abnormality makes us withdraw an attribution of competence. For, in so far as competence is conceived communitarianly, i.e. as regarding the correct use of a term, where correctness is taken 'as convergence with the linguistic use of speakers who are generally regarded as competent' (p. I 30), then people mobilizing unusual inferential chains or unusual applications of a word are judged as acting not competently (p. 78). According to Marconi, however, there is no tension between the individualistic and his communitarian picture of competence. Rather, the two pictures are compatible (we might take Marconi's position as a whole as outlining a weak form of social-externalism regarding competence), for the latter enters the stage only in so far as the issue of correctness enters too. There really are wrong, or incorrect, uses of terms; these are the uses which do not conform to, or better converge with, standard uses of such terms. That there must be standard uses of terms is according to Marconi undeniable, for actual use is supported by an inclination to converge with a standard use. What really is a standard use of a term may be subject to question. For, although a standard use of a· term is an authoritative use of it, there may be disagreements within the community as to what has to be regarded as the authoritative use of that term (and this disagreement may never come to an end) (pp. 1 26-9). What he denies, on behalf of Wittgenstein's ( I 9 S 3 ) private language argument, is that an individual's ·deviant uses may turn out to be correct with respect to this person's own standards; rather, this person's uses will be judged as incorrect with respect to his or her community standards (p. I 3 I ). As I just outlined, Marconi considers his conception of normativity to be Wittgensteinian. In spirit, certainly; but to the letter? Wittgenstein's picture of normativity seems to allow what Marconi explicitly denies, namely that a deviant use of a term is correct with respect not to our, but to its own paradigmatic application (actually, Wittgenstein makes this point with respect to a (factually) solitary use of a term3). Thus, for Wittgenstein a use of a term is not mistaken in that it is deviant, as Marconi instead wishes to defend (p. I I9). Be it the majority's use or not, a mistaken use of a term is simply that which does not conform to its own paradigmatic application, i.e. to what ftxes the meaning of that term, hence the concept expressed by it. The point is that in Wittgenstein's frame community enters one step back, as a mere presupposition of correctness:4 any use whatsoever of a term, deviant or not, could not turn out to be correct if the term's paradigmatic
·
3 16 Book Review
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
application were only a purported one, i.e. it did not tell correct from incorrect uses of the term itself; and in order for that application to have a genuinely paradigmatic character any later use of the term must be such that it must be in principle possible for everybody to distinguish its real from its merely alleged correctness. Dealing with Burge's (1979) famous case of Bert's abnormal use of 'arthritis', Marconi maintains that allowing for linguistic deviance is-contra Burge and like Bilgrami (1992)-allowing for formation of a concept different from that expressed by the linguistic term involved. Unlike Bilgrami, however, for him being linguistically deviant means being linguistically mistaken, in the aforementioned sense of diverging from uses recognized as standard (pp. 1 16-20). But what generates linguistic deviance? Enormous falsity of beliefs, Marconi answers: Bert 'is deviant because he is badly mistaken' (p. 121). This dependence of linguistic incorrectness on factual incorrectness of beliefs seems highly problematic, however. Firstly, why should Bert's belief be false, if it concerns his notion of arthritis? Surely it would be false if it concerned our notion of arthritis, as Burge sustains; but this is precisely what Marconi is prepared to deny (p. 98). In holding that Bert both has a grossly false belief and uses 'arthritis' deviantly, Marconi seems reminiscent of Quine's famous thesis according to which 'assertions startingly false on the face of them are likely to turn on hidden differences of language' (Quine 1960: § 1 3). However, Quine's point is not that we have to ascribe someone both a different concept and a grossly false belief about that concept, but rather that we have to choose between ascribing someone a different concept and attributing him a grossly false belief about an ordinary concept. Secondly (and more sub stantially), enormous falsity in belief is neither a necessary nor a sufficient condition for the formation of a new concept. It is not a necessary condition (as the previous remark on Bert's 'arthritis'-belief may have made us already suspect). By believing that I am feeling pain on the top of my nails and expressing this belief accordingly, I use 'pain' deviantly; thus, I coin a new concept of pain. But, in so far as such a belief concerns a sensation of mine, it is incorrigible, i.e. it cannot turn out to be false. It is not a sufficient condition either. By badly misperceiving a chair as a unicorn, I mobilize the . grossly false belief about unicorns that some unicorn or other stands in front of me. But, in so far as in expressing that belief I do not make any linguistic mistake in using 'unicorn', i.e.-in Marconi's own understanding of the matter-! do not use 'unicorn' deviantly, I do not coin any new notion of unicorn. Although he repeatedly stresses that his is a theory of competence and not of meaning, he is aware that his internistically based picture of competence is incompatible with some theories of meaning and reference,
Journal of Semantics 3 1 7
·
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
notably the causal-externistic ones. If the famous Putnamian beech/elm and Twin Earth cases supported these latter theories, Marconi's own theory would be eo ipso ruled out, for it might then implausibly occur that two subjects had the same referential competence while actually referring to utterly different things. But Marco�i argues that the Putnam cases (at most) show that not only competence, but also reference is social-externistically oriented: before I 750, 'water' refers on Twin Earth (as well as on Earth, one may presume) to what Twin-Earthling experts at that time singled out as water, namely both to XYZ and to H2 0 (p. 10 I). This raises the following question. Marconi must admit that after I750 'water' has changed reference (on Twin Earth as well as on Earth; on Twin Earth, it merely refers to XYZ, whereas on Earth it designates H2 0). For experts have in the meanwhile mobilized a referential competence so different, that it cannot make them pick out the same substance by their respective 'water' -utterances. But what reason is there in principle which justifies such a change? To be sure, experts have changed their mind as to what constitutes what they refer to by 'water'. But why should this change of opinion also make for a reference change? There are plenty of cases in which no such change occurs. A final word on the book's methodology. Marconi squares philosophy with science, as a research which has to find in empirical data the same kind of evidence as science does (p. 3). He intends his main theoretical distinction between inferential and referential competence to be confirmed by neurological data (pp. 67-73). Inasmuch as this is the case, it is vain to look for arguments in the book that take e.g. physical realizations in the mind of the notions Marconi claims to be theoretically distinct as either sufficient or both necessary and sufficient conditions for those notions themselves. For this would still be a performance in the 'conceptual analysis' style which Marconi no longer advocates as the philosopher's task. To my mind, this attitude of his explains one of the main merits of this book, namely its intellectual honesty. At first sight, a metaphysically oriented philosopher might be dissatisfied by the fact that throughout the book Marconi always opts for the weaker version of a philosophical theory. But the point is precisely that Marconi never tries to defend a thesis whose empirical support may be poor. Nevertheless, scattered about in the book one finds reflections that one is forced to classify as instances of conceptual analysis, that. is, as something which experience cannot be called upon to confirm, for it is rather an assessment of all possible experience. Take for example Marconi's remarks on how a spoon does not depend on its being used for its being a spoon (p. I 5 5). But also the main thesis of the distinction between inferential and referential competence is apparently originally argued for in a 'conceptual analysis' style, by appealing to the possible difference of the extensions of the notions of being inferentially and being
3 I 8 Book Review
referentially competent: even if accidentally the two extensions actually coincide, we might think of someone who had one but not the other (e.g. the bookish scientist who knows practically everything about Aulularia Clemensi and still fails to pick out a single one of these butterflies (p. s8); when Marconi makes us think of such a person, we are not led yet to assume that this scientist suffers from brain damage.) The impression is therefore that Marconi himself cannot avoid feeling nostalgia of that philosophical style which kept philosophers in touch with the Platonic heaven of concepts; as he himself writes in the Introduction, 'heaven is, of course, a much better place' (p. s).
NOTES I C£ Recanati (I993: I 57). 2 C£ Kaplan (I 989: 520).
3 C£ Wittgenstein ( I 9 5 3 : §204, I9782: VI, §4I). 4 C£ Wittgenstein (I95J: §240, I 9 8 I 2: §428) 0
REFERENCES Bilgrami, A ( I 992), Belief and Meaning, Blackwell, Oxford. Burge, T. (I979), 'Individualism and the mental', Midwest Studies in Philosophy, 4, pp. 7J- I 22. Kaplan, D. (I989), 'Demonstratives', in J. Almog et al. (eds), Themes from Kaplan, Oxford University Press, Oxford, pp. 48 1-563.
Quine, W. V. 0. ( I96o), Word and Object, MIT Press, Cambridge, MA Recanati, F. ( I993), Direct Reference, Black well, Oxford. Wittgenstein, L. (I95 3), Philosophische Unter suchungen, Blackwell, Oxford. Wittgenstein, L. ( I 9782), Bemerkungct'iiber die Grundlagen der Mathematik, Blackwell, Oxford.
Wittgenstein, L. (I98 r}, Zettel, Blackwell, Oxford.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Alberto Voltolini Institute of Philosophy and Human Sciences Faculty ofScienze della Jormazione University of Palermo, Italy Piazza 1. Florio 24 - I-90139 Palermo {Italy) email:
[email protected]
