JOURNAL OF SEMANTICS
AN INTERNATIONAL j OURNAL FOR THE INTERDISCIPLINARY STUDY OF THE SEMANTICS OF NATURAL LANGUAGE
MANAGING EDITOR: PETER BoscH (IBM Germany) REVIEW EDITOR: BART GEURTS (IBM Germany)
EDITORIAL BOARD: PETER BosCH (IBM Germany)
SIMON C. GARROD (Univ. of Glasgow)
BART GEURTS (IBM Germany) PAUL HoPPER (Carnegie Mellon Univ. , Pittsburgh)
LAURENCE R HoRN \'iale University) STEPHEN lsARo (Univ. of Edinburgh) HANs KAMP (Univ. of Stuttgart)
LEo G. M. NoORDMANN (Univ. of Tilburg)
R oB A. VANDER SANDT (Univ. ofNijmegen) PIETER A. M. SEUREN (Univ. ofNijmegen)
CONSULTING EDITORS: R. BARTSCH (Univ. of Amsterdam)
SIR jOHN LYONS (Univ. of Cambridge)
J. VAN BENTHEM (Univ. of Amsterdam)
W. MARstEN-WILSON (MRC, Cambridge)
D. S. BailE (Univ. of Manchester) H. E. BREKLE (Univ. of Regensburg) G. BROWN (Univ. of Cambridge)
]. D. McCAWLEY (Univ. of Chicago)
H. REICHGELT (Univ. of Nottingham)
B. RICHARDS (Imperial College, London)
H. H. CLARK (Stanford University}
A.J. SANFORD (Univ. of Glasgow)
H.-J. EIKMEYER (Univ. of Bielefeld)
A. VON STECHOW (Univ. ofKonstanz)
]. HoBBS {SRI, Menlo Park)
D. IsRAEL (SRI, Menlo Park)
D. VANDERVEKEN (Univ. of Quebec) Z. VENDLER (Univ. of California, San Diego)
E. L. KEENAN (Univ. of California,
Y. WILKS (New Mexico State Univ.,
0. DAHL (Univ. of Stockholm)
H. ScHNELLE (Ruhr Univ., Bochum)
G. FAucoNNIBR (Univ. of California, San Diego)
P. N.JOHNSON-LAIRD (MRC, Cambridge) Los Angeles)
M. STEEDMAN (Univ. ofPennsylvania)
B. L. WEBBER (Univ. ofPennyslvania) Las Cruces)
D. WILSON (Univ. College, London).
E. LANG (Univ. or Wuppertal)
EDITORIAL ADDRESS: Journal ofSemantics, IBM Germany Scientific Centre, IWBS
7000-75, Postfach 8oo88o, D-7000 Stuttgart 8o, Germany. Phone: (49-71 1-) 6695-559. Telefax: (49-71 I ) 6695-500.
BITNET: bosch®ds01ilog.
New Subscribers to the Journal of Semantics should apply to the Journals Subscription Department, Oxford University Press, Pinkhill House, Southfield Road, Eynsham, OX8 1]. For further information see the inside back cover.
Volumes J--(} are available from Foris Publications Holland, PO Box 509, 3300 Am Dordrecht,
The Netherlands.
Published by Oxford University Press
Copyright by NIS Foundation
ISSN 0167-513 3
JOURNAL Of· SEMANTICS Volume 8
Numbers
1
&
2.
SPECIAL ISSUE: FO CUS IN PHONETICS, SYNTAX, SEMANTICS AND PRAGMATICS Guest Editors: Jakob Hoepelman, Rudolf Schnitzer
CO NTENTS JOACHIM jACOBS Focus ambiguities PETR SGALL Focus and the levels of language system
37
JACK HoEKSEMA and FRANs ZwARTS Some remarks on focus adverbs
51
ULRICH F. G. KLEIN Focus: an idea in motion
71
S]AAK DEMBY 'Only' as a determiner and as a generalized quantifier
91
RoB T. P. WieHE External and verb phrase negations in actual dialogues
107
JAY DAVID ATLAS Topic/comment, presupposition, logical form and focus stress implicatures: the case of focal particles only and also PETBR I. BLOK Focus and presupposition
127 149
U\1] @)�(U] [{'@) � l
Associate Editors: Ellen Broselow, Dept. of Linguistics, SUNY, Stony Brook, USA; Carol Georgopolos, Uni versity of Utah, USA; Michael Kenstowicz, Dept. of
Linguistics
and
Philosophy,
MIT,
USA; Frederick J. Newmeyer, Dept. of Linguis
tics, University of Washington, Seattle, USA; David Pesetsky, Dept. of Linguistics and Philo sophy, MIT, USA Natural Language and Linguistic Theory pro vides 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 va riety of point of view. The journal actively seeks to bridge the gap between descriptive work and work of a highly theoretical, less empirically oriented nature. Natural Language & Linguistic Theory i s surveyed by MLA International Bibliography, Linguistic & Lan guage Behavior Abstracts, Current Contents/Arts & Humanities, CC!Social & Behavioral Sciences, ASCA, Arts & Humanities Citation Index, Social Sci
ences Citation Index, Linguistic Abstracts, Linguistic Bibliography, IBZIIBR, Computer & Information Sys
tems Abstracts Journal.
Subscription Information ISSN 0167-806X 1991, Volume 9 (4 issues) Institutional rate: Dfi.302.00/US$171.50 incl. postage and handling Private rate: Dfl.124.00/US$48.00 incl. postage and handling P.O. Box 322, 3300 AH Dordrecht, The Netherlands P.O. Box 358, Accord Station, Hingham, MA 02018-o358, U.S.A.
KLUWIER ACADEMIC PUBLISHERS
OXFOJRD
o
lJNliVIEJR§liTY i' Semantics i'
o
JP�IE§§
-----
A New Approach to lEnglisllt Grammar, on §emamtic JP>rindpftes
R. M. W. DIXON R. M. W. Dixon argues for an approach
to language which starts from the premiss that a speaker 'codes' meaning into grammatical forms in order to communicate them to a hearer, who recovers the 'meaning'. He makes a theoretically original contribution to this school of thought by exploring the semantic basis of grammar.
0 19 824272 7, 400 pages, Clarendon Press £35.00 interpretive Acts In Search of Meaning
WENDELL V. HARRIS 'Wendell V. Harris is rare as a literary critic who seeks guidance from philosophy oj language. Interpretive Acts draws on speech act theory to give a synoptic view of literary interpretation.Many of his individual observations are eminently sensible.' Peter
lLexicall §emantics witlhtout Thematic Roles YAEL RAVIN
This book argues that thematic roles are not valid semantic entities, and that syntax and semantics are indeed autonomous and independent of each other.
0 19 824831 8; 256 pages, Clarendon Press £27.50
A Theory of lLanganage and Hnformation A Mathematical Approach
ZELLIG HARRIS In this, his magnum opus, an eminent
linguist presents a formal theory of language structure. He argues that the combining of words into a sentence constitutes a mathematical object, and that each departure from equiprobability is a contribution both to the structure and to the meaning of the sentence.
0 19 824224 7, 432 pages, Clarendon £35.00 1991
Lamarque, Times Literary Supplement
Press Janu�ry
The §ignifier and the §ignified
Tlb.e lLiar
0 19 812959 9, 208 pages, 4 text-fi�res, l table, Clarendon Press £22.50 Studies in the Operas of Mozart and Verdi
FRITS NOSKE
This provocative and stimulating book, in Dutch, is a first published in collection of studies on the interpretation of opera based on structural analysis.
1977
'a major contribution to tke study ofkow opera
works ...full of brilliant insights and deduaions' Roderick Swanston, Opera Now
0 19 816201 4, 430 pages, numerous musical examples, paper covers, Clarendon Press £14.95
An Essay on Truth and Circularity
1o.�4. ��� '?
JON BARWISE and JOHN ETCHEMENDY
c -r
The authors provide their own solution to semantical paradoxes like the Liar, based on J. L. Austin's idea of truth.
'This is a splendid book.Jon Barwise andJohn Etckemendy have striking new idem and material. These they have thought through deftly and masterfully ...a book to seize the philosophical imagination.' W. D. Hart,
University College London, Mind
0 19 505944 1, 206 pages, figures and tables, paper covers, OUP USA £7.95
For more inform4tion, including a leaflet on new and recent titles in Linpistics, please write to: Acadmlic Publicity, Oxford University Press, Walton Street, Oxjord OX2 6DP.
ILIINCGrlUII§1fiiC§ AIB3§1f�ACC1f§ Edited by David Crystal The focus of Linguistic Abstracts is the theory and
Each issue of ll..inguistics
Abstracts is divided into the
practice of general linguistics,
following sections:
including all the recognized
Linguistics Theory, Semantics,
"core" areas of linguistic enquiry and their major interdisciplinary points of connection. More than 140
journals from over 20 countries
are
currently
covered.
Grammar, Phonology, Phonetics, Pragmatics, Discourse Analysis, Sociolinguistics, Psycholinguistics, Neurolinguistics, Historical Linguistics, Comparative Linguistics. Philosophical Linguistics, Mathematical Linguistics, Computational Linguistics, and Historiography
ORDER FORM
Subscription Rates, Volume 7, 1991
LINGUISTICS ABSTRACTS
Individuals £23.00 (UK), £27.50 (Europe), $55.50 (N America), £30.50 (Rest of World) Institutions £73.50 (UK), £78.00 (Europe), $156.50 (N America), £86.00 (Rest of World) Published quarterly
0 0
D
Please enter my subscriptiontsend me a sample copy
I enclose a cheque/money order made payable to Basil Blackwell Please charge my American Express/Diners Club/ MastercardNisa account number
-----Expiry Date --- For payments via the National Girobank, the Basil Blackwell account number is 236 6053
Name Address
Postcode Signature
------
______
Date
----
Payment must accompany orders
Please return this form to: Journals Marketing Manager, Basil Blackwell,
108 Cowley Road. Oxford OX41JF. England. Registered No. 1802n. Or to: Journals Marketing Manager, Basil Blackwell, Three Cambridge Center, Cambridge, MA 02142,
USA
Basin BHackweiD
Oxford and Camlbrndge MIA
Computatio11al Linguistics James F. Allen, Editor
Published quarterly by The MIT Press for the Association for Computational Linguistics ISSN 0891-2017
Computational Linguistics is the foremost journal devoted exclusively to computational analyses of natural language. It encompasses AI research in language, linguistics, and the psychology of language processing and performance. With each issue providing applied and theoretical papers, book reviews, technical correspondence, and letters to the editor, the journal presents a stimulating
Yearlv Institution,,! R,1te: 570.00. r\ dd SI�.Ofl )WSt,lge and h,mdling outside L'.S.t\. Prep.1yml'nt i5 reqqirL·d.
Send check dr,11n1 ,,g,,inst ,, U.S.
forum for the exchange of ideas and trends in computational language.
b,mk in
U.S. funds.
>'.·l,lsterCard or VIS:\ number
to:
i\IIT Press Journ.,ls
55 H,wwMd Street . C11nbridge. i\IA \121�:! 617-253-211119
Computational Linguistics is �journal serving a rapidly expanding interdisciplinary
FAX 617-25R-6779
!ndi\"idu,JI subscripdons Me
-a\:,lllt:1ble-unly-thntugh --- membership in the Association for
field and should be part of every research library collection. Recommend CL to your librarian today.
Comput,1tion.1l Linguistic;. Plt\15l' cont.1rt: Oon.1ld
E. \\',,lker (:\CL)
Bellcore, i\IRE 2:\:179
.;�:;South Str,•et, llo� 1910 Morristown. U.S.:\.
:'\J
\1/'J(l().j 9 Ill
l@� @j[[i)@j ��o�©�©�lhw An International Journal for Jurisprudence and Legal Philosophy Editor: Alan Mabe, Dept. of Philosophy, Florida State University, Tallahassee, USA
Managing Editor: Roxane Fletcher, Dept. of Philosophy, Florida State University
Book Review Editor: Joan McGregor, Dept. of Philosophy, Arizona State University, Tempe, USA Law and Philosophy is a forum tor the publication of work in law and philosophy which is of common interest to members of the two disciplines of jurisprudence and legal philosophy. It is open to all approaches in both fields and to work in any of the major legal traditions - common law, civil law, or the socialist tradition. The editors of Law and Philosophy encourage papers which exhibit philosophical reflection on the law informed by a knowledge of the law, and legal analysis informed by philosophical methods and principles. Law and Philosophy is surveyed by Current Contents!
Social & Behavioral Sciences, ASCA. Social Sciences Citation Index, Revue Philosophique de Louvain, So ciological Abstracts, Legal Contents, The Philosopher's Index, Index to Legal Periodicals, Current Law Index! Legal Resource Index.
ISSN 0167-5249 Subscription Information 1991, Volume 10 (4 issues) Subscription rate: Dfi.270.00/US$153.50 incl. postage and handling Private rate: Dfl.114.00 I US$65.00 incl. postage and handling P.O. Box 322, 3300 AH Dordrecht, The Netherlands P.O. Box 358, Accord Station, Hingham, MA 02018-()358, U.S.A.
KLUWER ACADEMIC PUBLISHERS
Scope of this Journal The JOURNAL OF SEMANTICS publishes articles, notes, discussions, and book reviews in
the area of narural language semantics. lc is explicidy imerdisciplinary, in that it aims at an integration of philosophical, psychological, and linguistic semantics as well as semantic work
done in artificial imelligence and anthropology. Contributions must be of good quality (to be judged by at least rwo 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 are 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 Typescripts for publication should be sent 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. Footnotes and bibliographical references must
appear at the end of the typescript. Diagrams must be submitted camera-ready. All papers
submitted are subject to anonymous refereeing. Authors receive 20 offprints of their published articles and IO offprints of their published reviews, free of charge. Larger numbers can be supplied at cost price by advance arrangement. Unless special arrangements have been made, copyright rests with the NIS Foundation.
Subscriptions Published quarterly, at an annual subscription of £45 (US$90 in N. America, £40 in the UK); single issues £I3 (US$26 in N. America, £ 1 1.50 in the UK). Personal rates are available on application to the Journals Subscriptions Department, Oxford University Press, Pinkhill House, Southfield Road, Eynsham, Oxford, OX8 I]. Prices include postage by surface mail or, for subscribers in the USA, Canada, Japan, India, Australia and New Zealand, by Air Speeded Post. Airmail rates are available on request.
Order Information Payment is required with all orders and subscriptions are accepted and entered by the volume. Payment may be made by cheque or Eurocheque (made payable to Oxford University Press), National Girobank (account 500 1056), 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, Pinkhill House, Southfield Road, Eynsham, OX8 I], UK. Telex 83I47 OXPRESG.
Advertising Advertisements are welcome and rates will be quoted on request. Enquiries should be addressed to the Journals Advertising Co-ordinator, Journals Productions Department, Oxford University Press, Pinkhill House, Southfield Road, Eynsham, Oxford, OX8 I], UK.
•
joum41 of&mantics
8:
1-36
C> N.l.S. Foundation (1991)
Focus Ambiguities J OACHIM JACOB S University oJWuppertal Abstract
(I) Peter schlagt 'Erich. with main stress on the object NP, this NP may be focus, but also larger units containing it, namely the predicate schltigt Erich or the whole sentence. What I will tty to show is that within the theoretical framework sketched in Jacobs (I 988a) focus projection proves to be a non-general, non-homogeneous and, in a way, a non-essential phenomenon. Focus projection is non-general in the sense that its typical pattern discussed in the literaTUre-'normal' stress, larger foci properly including smaller foci-does not cover all instances of focus ambiguity. Focus projection is non-homogeneous because, even if one confines oneself to the typical cases, it comprises several completely different kinds of ambiguity. Finally, focus projection is non-essential if-as the term suggests-it is interpreted as involving percolation of focus feaTUres. In order to explain any of the ambiguities to be discussed here, the grammar does not have to contain technical mechanisms that move focus feaTUres along certain paths in strucTUral trees.' Rather, all of these ambiguities follow from independently motivated rules relating different levels of grammar or constraining grammatical representations. I will also tty to show that within (a slightly modified version of) the theoretical framework mentioned above some of the notorious empirical problems of focus projection find a solution, and that they do so without introduction of additional theoretical machinery.2
1 FBS IN SEM ANTICS The possibility of differentiating several different kinds of ambiguity related to focus or stress in the theoretical framework of Jacobs (1988a} is a consequence of the fact that in this framework both focus and stress receive a multi-level grammatical representation. FBS appears in at least two levels of gr ammar, namely in semantic structure and in at least one level of syntactic structure. Let us ftrst have a look at semantic structure. In representations of the literal
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
In what follows I will discuss ambiguities related to focus and stress in German sentences. Some of these ambiguities will be typical instances of what is now widely called 'focus projec tion', a term which was introduced by T. Hohle in his seminal paper of 1982. Focus projection arises in phrases with specific 'normal' stress patterns and consists in the possibility of assigning to such phrases several focus-background strucTUres (FBS), differing from each other in the size of the constituent in focus. For example, in example {I),
2.
Focus Ambiguities
conventional meaning of sentences, FBS is a partitioning of formulas within the scope of certain operators, as exemplified in (Mra,b):3 (Mxa) ASSERT (# AXNP[PETER(A.x[X(A.y(SCHLAGT(y)(x)]))}], ERlCH #) (Mrb) ASSERT (# AXVP(PETER(X)], A.x[ERlCH(A.(SCHLAGT (y)(x)])] #)
(2a) Peter kennt nur einen Roman von Goethe. (2b) Peter kennt nur einen Roman von Goethe. In the relational theory of focus, the fact that these two sentences have different truth conditions follows from a variation of FBS within the scope of a focus-sensitive truth-conditional operator, as sketched in (M2a) and (M2b): (M2a) ASSERT(# [NUR(#A.Ypp(PETERKENNT E. ROMAN Y], X#)],VON G. #) (M2b) ASSERT(#AXN(NUR(#A.YN(PETERKENNTEINEN YV.G.).X#)), ROMAN#)
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(Mra) represents the object-focus reading of (r) by splitting the propositional content of this sentence into two parts, one corresponding to the background, with a lambda-bound variable in the object-position, and one corresponding to the focused object. The double crosses flanking the sequence of these two parts indicate that this sequence, in addition to having a truth value in given contexts, is assigned a set of alternatives in every context, depending on the way the proposition is divided into focus and background. (This set of alternatives therefore is the model-theoretical entity most directly corresponding to FBS.) This double-cross formula is the scope of an operator representing the illocu tionary potential conventionally associated with the sentence in question. This operator is sensitive to the set of alternatives assigned to the formula in its scope, so that the illocutionary potential expressed by such formulas will vary with the choice of focus and background. (Mrb), which corresponds to the predicate-focus reading of ( r ) , therefore represents a different illocutionary potential (with the same propositional content, as the propositional content of ASSERT(#B,F#) is B(F) ). This way of semantically representing FBS combines ideas originating in Cresswell's and von Stechow's work on structured propositions and in M. Roath's (1985) semantics of focus with the relational theory of FBS proposed in earlier publications of mine. The relation between these theories is discussed in von Stechow (1989). Exact rules for formulas like ( r ) are given in Jacobs (1988a) and Jacobs (1989b). In Jacobs (1988a), one will also fmd arguments against the exclusion of FBS from conventional meaning. One of the most severe shortcomings of a purely pragmatic theory of FBS will be that it will not be able to account for the fact that differences in FBS may result in different propositional contents, as in the pair of sentences (2a) and (2b):
JoachimJacobs
3
Note in passing that (M2a) and (M2b) also reflect the fact that in the sentences in question the stressed constituents simultaneously serve as focus of an adverbial and as focus of the whole sentence. ( This also holds for another reading of (2a,b) in which the focusing particle together with its focus is sentence focus.) This kind of multiple-focus assignment must not be confused with the one in cases like (3), (3) Peter schlagt 'Erich. interpreted as an answer to the multiple wh-question Wir schliigt win? Here, the focus of a single operator consists of several independent parts, as shown in (M3), while in (2) a single constituent serves as the focus of several operators:
2
FB S IN S Y N TA X
In syntax, FBS looks quite different. But why do we need a representation of FBS in syntax at all? Wouldn't it be more elegant to have just one level of focus background representation? There is a simple answer to this. Quite clearly, stress rules must have access to FBS. But from all we know about the overall organization of grammars, stress rules cannot be applied directly to semantic representations like (MI). Rather, they must be applied to syntactic structures of a suitable kind. Therefore, we need FBS in syntax. Furthermore, in many (if not all) languages of the world, certain purely syntactic processes or restrictions interact with FBS, and this cannot be captured without having some direct indication of FBS in syntactic representations. In the theoretical framework of Jacobs (I988a) this syntactic indication of FBS is the result of an assignment of the feature [fn] (n � I) to syntactic constituents, as in (S1a) (Figure I). (S1a) corresponds to (M1a), i.e. to the object-focus reading of Peter schliigt'Erich. (S2a) (Sla)
Peter
schliigt
e
Erich
e
Figure 1
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(M3) ASSERT (#AXNPA.YNP(X(A.x(Y(A.y(SCHLAGT(y)(x)])])],PETER, ERICH#)
4 Focus
ambiguities
(S2a)
/-J-:rsy ADVP C
TOP
Peter
kennt
e
VP
2
NP,.......--____ y
nw
I
einen
-Nt N"'PP
[[+£2fl)1 +
� Roman von G. e
(S2b)
and (S2b) (Figure 2), on the other hand, correspond to (M2a) and (M2b), respectively. The tocus features in these trees carry an index which identifies the focusing operator. 'fi' means 'focus of the illocutionary operator', 'f2' means 'focus of the operator corresponding to the constituent with index 2', and so on. For example, in (S2a), the PP is marked both as the focus of the illocutionary opera tor and as the focus of the operator corresponding to the adverb. In (S3) (Figure 3), both the subject and the object are marked as focus of the illocu tionary operator.
c2
(53) t
/J-l-,)Z NP NP [+ ) [+II I i I
TOP
C
Peter
schliigt
V
e
Erich
e
Figure 3
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Figure 2
Joachim Jacobs
S
An example where focus features with different indices are assigned to dif ferent constituents is (4), where the subject is focus of sogar and the object noun is focus of nur (see also Figure 4): (4) Sogar Peter kennt nur einen Roman von Goethe. (54)
c2
I 'r'
Sogar
Peter
C
kennt
NP
I
e
ADVP /--- N1 /".. 3 D
I I
nur emen
dfJ
6
Roman von G. e
Figure 4
The general rule behind this assignment of indices to occurrences of the focus feature is that the focus of an operator X will have index n+I in case there is an operator Y which has scope over X and whose focus is assigned index n. There fore, these focus indices reflect the scope relations between the focusing opera tors present in the semantic structure of the sentences. Curiously, these scope relations play a role in some of the rules operating on focus-marked syntactic structures, e.g. in stress rules, as we will see in Section S· But, speaking of rules, the ftrst question to be answered here is: what are the rules connecting these syntactic focus representations with semantic focus representations like (M1)? In order to answer this question precisely and in detail, I would have to discuss many difftcult issues not directly related to FBS. Therefore, all I can do here is to give an informal and somewhat simpli fted condition on the well-formed ness of pairings of semantic and syntactic focus-background representations, roughly equivalent to the rules given in Jacobs (1988a: IOJ). In this condition (C1), I presuppose that not only occurrences of the focus feature but also the double-cross subformulas of semantic representations carry an index indicat ing the scope of the focusing operator: (C1)
A pair(M,S) is well formed only iffor every subformula #B, FI,..,Fm# of M with scope index n and for every Fi (I��) within this subformula I-
3 hold: 1. all and only the constituents of S whose meaning is represented by a constituent of Fi are marked with [fn] or dominated by a constituent marked with [fn];
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
rJ-rr-r-�v ADVP TOP 2 [+£2 1
6
Focus ambiguities
2. only non-empty constituents5 are marked with [fn]. 3· It is not possible to reduce the number of occurrences of [fn] in S with out violating I or 2.
(Ms) ASSERT (# AXNP[PETER(A.x[X()..y [SCHLAGT(y)(x)])])], DEN{HUND) #) C2
(55) 1
/)-,7--NPVP,.__v T I i �·��·r;
Peter
schlagt e
den Hund e Figure 5
(Ms,Ss) violates 3 of {CI) because I and 2 will also be fulfilled when the lower two occurrences of [f I] are dropped. {CI) therefore prohibits the spreading of focus features over proper parts of focused constituents. Such a spreading blurs the distinction between multiple-focus assignment on the one hand and the focusing of complex constituents on the other, a distinction which is important for several of the grammatical processes involving FBS (and which in fact terids to be blurred in the approaches based on focus feature percolation). (CI), in conjunction with other principles of the grammar, also prohibits non-spreading assignments of [fn] in many cases. For example, both (S6a) and (S6b) are blocked (Figure 6). In any semantic representation that can be paired with (S6a) without violation of {CI), the representation of ja would have to be in the background of a wide scope focusing operator.6 This operator could be ASSERT or OBWOHL (or both). But modal parties like ja, due to their special hierarchical position in semantic representations,? cannot be in the background or anywhere else in the scope of illocutionary operators or subordinating conjunctions.8 Therefore, (S6a) is blocked.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
To see how (CI) works, let us Hrst have a look at (M1a) and (S�a). This pair ing of a semantic and a syntactic focus representation is well formed because in (S1a) only the object NP gets [fi], and therefore all and only the non-empty constituents of (S1a) whose meaning is represented by a constituent of the semantic focus ERICH are marked with [fi]. The third condition of (CI) is also obeyed here, as, obviously, the absence of any occurrences of [fi] would violate condition 1. In contrast, the pairing of (Ms) and (Ss), which also obeys conditions I and 2, is ruled out by 3 of (CI) (Figure s).
JoachimJacobs (S6a)
ct VP �s / ----
C
� �./"...,. v D�l
NP NP
I I
NP
ADVP
V
ja
klaute
I
I I obwohl Peter Gerda das Buch
C
/
NP
(S6b) VP
....--.---_. _. NP
I
yl
[+fl] NP..,..,....___V o�1 Aovrv
I I obwohl Peter Gerda das Buch
I ja
I klaute
Figure 6 (S6b) is ruled out for basically the same reason. To be part of the focus of ASSERT or OBWOHL, ja would have to be in the scope of one of these operators, which is impossible (see above). In conjunction with the stress rules to be discussed below,(C1) therefore yields an explanation of the fact that(6) (6) ??obwohl Peter Gerda ein Buch ja klaute is bad, regardless of the choice between a narrow or a wide focus inter pretation.9 To see how (C1) works in examples with multiple-focus assignment, have again a look at(M3) and(S3). Here, 1-3 of(C1) have to be applied twice, first to pan one of the semantic focus, PETER, and then to focus part two, ERICH, resulting in the focus-feature assignment found in(S3). Multiple application of 1 and 2 will also be required in examples with more than one semantic background-focus subformula, like (M2a). In this special case, however, a little problem arises, because in the embedded background-focus subformula the focus is a variable, and this variable, strictly speaking, does not represent the meaning of any syntactic constituent Rather, it is only related by possible lambda conversion to a semantic constituent representing the meaning of a syntactic constituent, namely of the PP. Therefore, the assignment of [f2] to the PP in(S2a) is somewhat problematic. But this is not a big problem since, of course, we could defme the relation between syntactic and semantic con stituents in such a way that the variable in question, through its being the target
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
ct �s
7
8 Focus ambiguities
of a possible lambda-conversion of the PP's semantic representation, would indeed count itself as semantically representing the PP. Similar measures will ensure the pairing of e.g. {M7) and (S7) {Figure 7): (M7) ASSERT (# AXyp[PETER{X)], )..x[ERICH(A.y [SCHLAGT(y)(x)])] #) (57)
·schlagt
e
Erich
e Figure 7
To license (M7,S7), which represent the VP-focus reading of (7), (7) Peter schl'agt 'Erich. the VP, although it contains an empty node in the position of the verb, would have to count as being semantically represented by the focus-part of (M7), i.e. it would have to be treated by (CI) as if schlagt were reconstructed in its original position. We therefore have two principles governing the assignment of [fi] to constituents containing empty nodes: (a) if the constituent does not contain any non-empty terminal material, [fi]-assignment is blocked by condition 2 of (CI); (b) if the constituent also contains non-empty terminal nodes, application of (CI) is based on the reconstructed version of the constituent. Note, however, that reconstruction of the V-position in (S7) does not make the actual (i.e. non reconstructed) position of schlagt invisible for (CI). Rather, the C-position of (S7) will get [fi], as it is filled by a word whose meaning is quite clearly represented by a constituent of the focus-part of (M7). In general, (CI) 'counts' both the actual and the reconstructed positions of moved constituents.10 Unfortunately, there is one type of [fi]-assignment to the results of move ment transformations that will not be treated properly by the rules discussed so far. The sentence Bin Buch hater gelesen (with a topicalized object) can be inter preted as focusing on the meaning of the reconstructed VP ein Buch gelesen (e.g. when uttered as an answer to Was hater gemacht? ). However, in order to enable the rules to be discussed in Section s to predict the stress pattern of this sentence, (CI) would have to assign [f I] to the top node of the sentence (as the topicalized element is integrated into the rest, cf. Section 5). But this would violate condition I of (CI), as in the resulting structure at least two constituents that are not part of the semantic focus (namely hat and er ) would be dominated
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Peter
JoachimJacobs
9
3
SEMANTIC F OCU S AMBIGU I TIES
Let us come back to the question of focus ambiguity. The relation between semantic and syntactic focus-background representations as determined by {C1) is not one-to-one. Rather, it is a non-biunique function from semantic to syntactic representations. This becomes transparent in examples like {8): {8) Peter schl'agt Erich. This sentence must be assigned at least two semantic focus-background representations, namely {M8a) and {M8b): {M8a) ASSERT (#AXNP )..Yv[X{A.x[E RICH()..y[ Y(y)(x)])])], PETER, SCHLA GT #) {M8b) ASSERT (#AX.yp[ERICH{X)], A.x[PETER{)..y [ SCHLA GT(x){y)])] #) (M8a) represents the reading activated when the sentence is uttered as an answer to the multiple wh-question W& tutErich wtis?-Peter schl' iigt Erich, Luise striichelt ihn , Gerda strickt ihm n eue Hciusschuhe, while (M8b) represents a single focus interpretacion, namely the one elicited by the question Was geschieht Erich?
(C1) will combine both (M8a) and (M8b) with the same syntactic focus representation (S8) (Figure 8). This is probably clear in the case 'of {M8a), but it may not be so clear f'lr (M8b). Look again at the focus-part of this formula. The only non-empty co1:< ;:'tents of {S8) that are semantically represented by constituents of this lambda-expression are Petei and schlagt. Note especially that there is no larger syntactic constituent whose meaning is represented by this
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
by [fi ]. What this seems to show is that under certain conditions constituents can become 'invisible' for 1 of (C1). A mong these conditions are contextual presence (as in the case of pronominals) and/or phonological weakness (as in the case of auxiliaries and clitics). Note that, for example, in the sentence Bin Buch hat ein Kin d gelesen , with a full, contextually non-present NP in subject position, semantic focus on ein Buch gelesen is impossible, as predicted by {C1) and the stress rules of Section S· Furthermore, Susanne Uhmann has observed that the phenomenon in question is restricted to cases in which the verb denotes one of the prototypical actions involving the object (like reading, in the case of books). For example, Bin Buch hat er zerissen is quite unlikely to receive a VP-focus interpretacion (whereas in weil er ein Buch zerissen hat this inter pretation is normal). This suggests that the phenomenon in question is a highly marked pragmatic option and that any attempt to take care of it directly in the grammatical rules governing [£i)-assignment would be misguided.
10
Focus ambiguities (58)
c2
1
/J-A Iii [+T I fl
TOP
C
NP
Peter
schlligt
e
VP
Erich
e
Figure 8
(9) Ihr Mann hat sie doch mit Petra betrogen. Among the possible interpretations of (9) are (M9a) and (M9b):11 (M9a) DOCH(ASSERT) (#AXNP "AYvi(HAT (X ("Ax (SIE ("Ay (Y(y)(x)])]))], IHR MANN Ax"Ay [MIT P.BETROGEN (x)(y)] #) (M9b) DOCH(ASSERT)(#A.Xvp[HAT (SIE ("Ax [X (x)]))], "Ax [IHR MANN ("Ay [MIT P. BETROGEN (x)(y)] )] #) ,
Again, these two formulas represent a bipartite focus (corresponding to a multiple wh-question) and a simple complex focus (corresponding to the question Was ist mit ihr geschehen ?), respectively, and again they are connected with the same syntactic strUcture by (C1) (Figure 9).12 (M9a,S9) is straight(59)
w � ) �
TOP
[+fl]
(j
Ihr Mann
C
NP ADVP
hat sie doch
NP
/ "V
/' NP v•
I
[fl]
........_ v pp
� I e e mit Petra betrogen e
Figure 9
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
lambda-expression or constituents of it. Therefore, the only way to satisfy (C1) in this case is to have an occurrence of [f1] on both the TOP- and the C-node. Another example of focus ambiguity induced by (C1)-let us call it semantic focus ambiguity-is (9):
Joachim Jacobs
II
(10a) Er hat es [ [f,] mir geschenkt]. (10b) ?Er hat das Buch [ff,] mir geschenkt]. The fact that in (wa), but not in (wb), the complex consisting of the pronominal indirect object and the main verb can be the syntactic focus follows from the interaction of several linear precedence rules governing the surface order of constituents in the 'Mittelfeld', as shown inJacobs (1988a). (Note that ' the contrast between {10a) and (10b) cannot be explained by simply stating that any deviation from the usual 10 > DO > V pattern will block focus projec tion, as both sentences have DO > 10 > V.) What is important for our present concerns is that the blocking of focus projection in cases like ( 10b) can be treated as the result of purely syntactic
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
forward. (M9b,S9) again is a consequence of the fact that there is no single syntactic constituent dominating all and only the constituents that are represented by parts of the semantic focus. Therefore, the focus feature must be distributed over several non-contiguous constituents. Note further that here the syntactic position of the modal particle is not in conflict with (C1) (cf. (S6a,b) ), as the particle in this position is not required to be in the scope of other operators by the given assignment of [fi ].'3 These two examples may suffice to illustrate semantic focus ambiguity. The reader might have noticed that, while the second of these examples, due to its 'normal' stress pattern, might be viewed as a typical case of focus projection, the Hrst of these examples is not of the kind that has usually been discussed in the literature on focus projection. Moreover, in none of these examples do we have the typical semantic pattern of focus projection, where one focus properly includes the other. Of course, there are many further interesting questions concerning the syntactic representation of FBS. For example: on which levels of syntax do focus-features appear? Obviously, the examples given so far as $-structures in the usual sense. But why not focus-features in D-structure? Unfortunately I cannot give a detailed discussion of this difficult question here. But note that the fact that certain types of movement are restricted by the presence or absence of focus-features is not a compelling argument for the assignment of focus in D-structure. As far as I can see, such restrictions can always be formulated as requiring the presence/absence of focus-features in certain positions of the target-structures of movements, i.e. as S-structure restrictions. Better evidence for an assignment of [H] at D-structure might come from the fact that some of our rules, e.g. (C1), require reconstruction of moved con stituents. Another important (and possibly related) topic in the syntax of focusing is the blocking of focus projection by certain 'non-normal' surface arrangements of constituents, as in (10a,b):
12
Focus
ambiguities
restrictions on the placement of focus feamres and does not require any reference to word order in the stress rules operating on such focus-marked S-structures.
4 REP RESENTATION S O F S T RES S
(Sla)
Peter
schlagt
e
Erich
e (Tla)
+ "' + -
Peter
schlagt
e
I I Erich
e Figure 10
Peter • • • •
schlagt • •
Erich • •
(G1a)
• • •
The arguments for choosing such a tree-and-grid notation instead of a tree only or a grid-only representation of stress are rather complex and their
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Before we begin to discuss stress rules, let us have a look at the representations that will be produced by these rules. In the theory lam advocating here, not only FBS but also stress is represented in two different ways, namely (a) in trees describing prominence relations between constituents and (b) in grids representing rhythmical patterns. For example, the prominence tree corres ponding to (S1a) will look like (T1a},14 while the grid-or rather one of the grids-corresponding to (S1a) is (G1a) (Figure 10):
Joachim Jacobs
·
13
discussion would lead me too far away. The reader is therefore referred to Hayes (I984), who has argued convincingly that both levels are necessary (and in fact correspond to quite different aspects of phonological organization). The notation of (Tia) is somewhat different from the usual metrical tree notation, as the latter employs the label's' and'w', while in the former we have '+' and'-'.This is not just a difference in labels, because'+' and 's' do not mean the same thing: 's' means 'more prominent than the w-sister'; '+' means 'more prominent than any minus-sister and equal in prominence to any plus-sister'. This presupposes the possibility of having '+' on more than one sister constituent, and indeed that is what we find in the tree corresponding to (S8) (Figure I I). (T 8) means Peter is equal in prominence to schliigt Erich , schliigt is
(T8)
R
rr(':;-._
n i
Peter
schliigt
e
r1 1 Erich
e
Figure II more prominent than Erich, and so on. The statement that two of the constituents here are equal in strength is certainly fully compatible with the intuitions, and, as we will see, will also give us the correct rhythmical patterns. In general, this plus-minus notation is more flexible (albeit less restrictive) than the strong-weak notation, and it has some rather nice implications. For example, within the strong-weak theory one needs many restructuring operations in order to maintain a binary branching pattern. The plus-minus notation, on the other hand, does not entail binary branching and therefore reduces the amount of restructuring needed at the syntax-phonology interface. Also it provides a perfectly natural way to represent the difference in the prominence relations between e.g. Peter scht iift with a narrow focus on the verb
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(58)
14
Focus ambiguities
and Peter schl' iift with a bipartite focus (as an answer to Wir tut wtis? Peter schl' iift, Hi!ms gucktfirnsehen , un d Gerda liest ein Buch). The difference is simply minus-plus vs. plus-plus, while in the strong-weak notation both would have to be represented as weak-strong and the difference would have to be located-quite artificially, I think-at some higher level of rhythmical organization. -
5
THE RELAT I ON BETWEEN SYNTAX AND P R OMINENCE
{RI) For any syntactic structure S, any set of sister-constituents SC within S and any i (i � 1): If there is an element of SC that is marked with [fi] or dominates a constituent marked with [fi] in S, then'+' is assigned to every element c of SC such that I. c is marked with [fi] or dominates a constituent marked with [fi] in S; 2. there is no j < i such that an element of SC is marked with [fj] or dominates a constituent of S marked with [fj], or a constituent of S marked with [fj] dominates c. To see how (R1) works, have again a look at (S1a) and (T1a) (Figure 12). Starting with the set of sister-constituents {TOP, C1} {RI) will require the marking of C 1, but not of TOP, with'+', as C 1 dominates a constituent marked with [fi] and there is no [fo] in the tree. TOP, on the other hand, gets'-' by 1 default. This is repeated for the other sets of sister-constituents of{Sia) and will ,_,fmally result in {Tia), given that the internal plus-minus-structures of poly syllabic words like Peter is projected from the lexicon and that, in a final step, all non-terminal vocabulary except '+' and '-' is deleted in the tree. Now l�t us have a look at a more complex case like {S4), repeated here for convenience{Figure IJ).{RI) transforms (S4) into (T4){Figure I J). Again, look first at the highest set of sister-constituents of {S4), namely {ADVP,C2}. Here C2 will receive'+', as it dominates a constituent marked (f1] and there is no [fo] in the tree. The next set of sister-constituents to be considered is {TOP,C1 }.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Let us first have a look at the rules that transform syntactic structures into prominence trees. These rules have the form of instructions to assign'+' and'-' to the elements of the sets of sister-constituents that given focus-marked syntactic structures consist of. (In fact, these rules, like (C1), are necessary conditions on the well-formedness of pairings of different kinds of trees. However, in this case it is more convenient-but not necessary-to formulate them as operations on trees.) {R1) accounts for narrow-focus cases:
Joachim Jacobs
15
(Sla)
Peter
scliliigt
e
Erich
e (Tla)
Peter
schliigt
e
I I Erich
e Figure 12
Here, TOP gets '+', as it is marked with [f1]. In {C,S), S gets '+', because it dominates a constituent marked with [f 3] and neither dominates [f2] or [fi] nor is dominated by [f2] or [fi]. This repeats itself for {NP,VP), {ADVP,VP) and so on. The result will be (T4), with the first syllable of Peter as the most prominent syllable and the second syllable of Roman on the second place in prominence. (R1) does not tell us how to assign'+' and '-' to sets of sister-constituents not containing any occurrences of the focus feature, i.e. to sets that will have neutral stess. We therefore need additional rules of neutral stress. The first of these rules is (R2): (R2) For any syntactic structure S, any set of sister-constituents SC within S and any i (i � 1): If no element of SC is marked with [fi] or dominates a constituent marked with [fi] in S, then '+' can be assigned to every element of SC that is not marked with [-ns]. '[-ns]' means 'not neutrally stressable'. This feature is assigned to constituents that cannot be prominent unless they are in a narrow focus. The set of [-ns] constituents of German comprises all empty constituents and constituents consisting of certain 'small words', such as the non-demonstrative deftnite pronouns, the definite and the indefinite article, the complementiser da.ft, the indefinite elements man , jemand, n iemand and the wh-elements wer, wen, was,
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
+ A + -
16 Focus ambiguities
(54) ·
c2
rt-rr-r-�v
ADVP
2
C
TOP
[+£2 1
I r·r
Sogar Peter
NP
I
kennt
+
3
/--- Nl
6
/'..
D
I I
r!fr
nw einen
Roman von G. e (T4)
+ + � n/--++ + I I
- _..,....--___
-
A
-
+
I
I
Sogar
-
� � 1\� A 1\ I I I I I nur einen Roman von G. e +
Peter
-
-
kennt
e
+
Figure 1 3
wo etc. Larger constituents, too, can become [ -nsJ under certain syntactic or pragmatic conditions. For example, a PP will be [ -nsJ if its NP-daughter is [-ns) {just like it will be [+wh) if its NP-daughter is [+wh] ), or anNP will be [-ns) if its denotation is contextually present (in a sense that is different from the contextual presence associated with the background pan of FBS, cf. Jacobs 1988a: 3.4). What is common to all [-ns)-constituents is that they are informationally irrelevant, either intrinsically or contextually. Unfonunately, I cannot discuss the assignment of [ -nsJ in any detail here. So let me just give an example, (Su) {Figure 14). {R2) can be applied to all sets of sister-constituents of {S1 1), and it must be applied, as neither {R1) nor {R3)-which will be discussed below-tell us how to assign prominence marks to the elements of these sets. The result will be (T1 1) {Figure 14). {R2) assigns '+' to S, as there are no occurrences of focus features in C or S, and C is marked with [-ns). In the set (NP,VP}, VP gets the plus, as NP,is [ -ns) . In the set consisting ofPP and V, V
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
R
e
ADVP
JoachimJacobs
(Sll)
ct [+fl]
/��
C
NP ( ]
I-t
r
daB
17
er
p ...-....._ NP
1
neben
[-t ihr
V
1
wohnt (Tll)
�-� · + ,;.._�
I
daB
I
er
i'' i
neben
I
ihr
+
wohnt Figure
14
gets the plus, as PP inherits [-ns] from its NP-daughter. Finally, in the set {P,NP], P gets '+', because NP is [-ns). The result is fully in line with the intuitive pronounciation of (SI I ) under conditions of neutral stress. (This will become more transparent when we specify the grids corresponding to such trees.) For certain cases, we need a second rule of neutral stress, and this rule causes problems. The phenomena that it tries to capture, however, are problematic for any of the existing theories of sentence stress. Consider (S1 2a) (Figure I s).Being an all-new phrase, (S1 2a) must receive the pattern of prominence depicted in (TI 2) (Figure I s). The plus-minus-assignment to {PP,V} and the minus-plus assignment to {C,S) and (P,N"P) is not licensed by (R2). Rather, the application of this rule would counterintuitively yield a plus-plus-pattern in all of these cases. Therefore we need a rule that assigns a plus-minus or minus-plus pattern to the constituent sets in question. But what do these sets have in common? Before we try to answer this question, let us assume that they do indeed have something in common, i.e. that they belong to the same natural class of sets of sister-constituents. Let us say that the deHning criterion of this class is the integration of one of the elements of the sets into another. For example, we want to say that in (S12a) PP is integrated into V and NP into P. On the basis of
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
R
18
Focus
ambiguities (S12a)
(T12)
i
_/ /;:_�
A
i
obwohl
-
I
er
-
I
in
A
i i
-
I
Miinchen wohnt Figure 15
these assumptions and definitions, (R3) will assign the correct patterns of plus . and minus: (R3) For any syntactic structure S, any set of sister-constituents SC within S and any i (i � 1 ) : If element c 1 of SC is integrated into another element c2 of SC, then '+' can be assigned to ct-15 The problem here, of course, is to specify precisely and in a general and non circular way the conditions under which integration of sister-constituents takes place. The following seems to be a (very vague) necessary condition for integration: {C2) Within a set of sister-constituents SC, integration takes place only if the elements of SC do not function as separate informational units in the communication. There are several ways in which elements of sets of sister constituents can function as separate informational units: J.
One, of course, is that they are in focus (or contain a phrase in focus). Therefore, (C2) should be taken to imply that no element of an SC to which (R3) applies is marked with [fi) or dominates a constituent marked with [fi]
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
R
JoachimJacobs 19 (for any i � I). (The fact that (R3) is a rule of neutral stress therefore follows from its central condition.) 2. Informational separation also takes place if at least one of the elements of SC, but not all of them, are informationally irrelevant and therefore marked with [-ns] (see above). Consequently, (R3) cannot be applied to this kind of SCs (which immediately explains why, for example, personal pronoun arguments cannot be integrated into a verb, whereas full NP arguments
)
can .
�·
5· Non-arguments of head constituents in most cases behave like separate informational units and therefore cannot be integrated into the head. Maybe the reason for this is that non-arguments normally locate the event, property or substance described by the head in some dimension (e.g. place, time, manner, quality), thereby adding information to information already present. (Note that in cases of coordination-another way of adding information-(R3) is also blocked.) 6. A very important way of informationally separating sister-constituents in natural languages is to divide them by a topic-comment boundary, i.e. to interpret one of them as denoting the entity about which some information has to be given, the others as providing this information. Such a boundary blocks integration of constituents regardless of whether they are in a predicate-argument relation or not. This has very much to do with the distinction between thetic and categorical statements (cf. Sasse I 987), the latter typically involving topic-comment structure, whereas the former describe events as atomic, non-structured chunks of information. The sets of constituents in (S i2a) to which we want to apply (R3) obey these conditions. PP and the NP Miin chen are non-pronominal, non-topical, adjacent arguments of V and P, respectively, and there are no focus features inside of {PP,VJ and (P,NP). The same can be shown for S in relation to C. In contrast, (C2) excludes application of(R3) to {NP,VP) and to {PP,VP) in (SI 3) (Figure I6). This sentence expresses a categorical statement about Peter, thereby separating the subject and the predicate by a topic-comment boundary. Furthermore, the PP in this case is not an argument of the verb. Therefore, (R2) has to be applied to both sets. The result is (T1 3) (Figure 16), which will yield intuitively correct metrical grids. 1 6 The reader might have noticed that conditions I -6 above do not exclude integration of a head into one of its arguments. This , however, should be
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
3· Another way of separation is separation by non-adjacency. As soon as two constituents are non-adjacent in linear order, they seem to be interpreted as two separate informational units (or as parts of two separate informational units). (R3), therefore, should be interpreted as requiring adjacency of c 1 and
20
Focus ambiguities (513)
(T13)
16
excluded as it never seems to occur. 17 Furthermore, if an argument is integrated into its head, the latter normally (see below) does not contain other arguments. Condition (C3) accounts for these observations: (C3) Within a set of sister-constituents SC, element c1 can be integrated into element c2 only if c2 neither is an argument of c 1 nor contains arguments of the lexical head of c2• (C3), among other things, explains the fact that in a neutrally stressed sequence of objects, it is always the last of these that receives strongest stress (Figure 17). (S1 4) will be transformed into (T14) (Figure 1 7), which corresponds to metrical ; grids (see below) in which no syllable is stronger than Buch. To make Kind rhythmically stronger than Buch , we would have to apply (R3) to (NP ,V 1 }. But this is prohibited by (C3), as V 1 already contains an argument of its lexical head schenkte. One might consider replacing (C3) by the requirement that c 1 should be an internal argument of c2• Actually, this is exactly what von Stechow and Uhmann (1 984) have proposed. However, given the criteria of distinguishing internal from external arguments customary in GB-theory, this will be too restrictive, as there are clear cases of integration of external arguments, e.g. in Das hat mir Gerdageliehen , as an answer to Woher hast Du das schone Armband? (cf. Jacobs 1 988a: 1 32, where I also argue that the verb in this sentence cannot be
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Figure
Joachim Jacobs (514)
ct [+fl)
/:J-r-- v' (R3) rr-r--. (R3) [-ns)
C
I weil
I er
NP NP /'...,. V 1::::::... 1::::::... I einem Kind ein Buch schenkte
R
I weil
-
+
(T14)
+ /"... _
I � � er einem Kind ein Buch
·� schenkte
Figure 17
viewed as informationally irrelevant or pan of the background). The argument integrated here is external according to all criteria: its semantic role is agent, the verb chooses haben (rather than sein ) as auxiliary in the perfect tense, etc. Nevertheless, it is a non-pronominal, non-topical argument adjacent to the lexical verb. There may be another way of replacing (C3) by a more general condition. It starts from the observation that there are many parallels between integration and (non-coordinative) compounding (cf. also Fuchs 1987: 428). In fact, the former might be viewed as a kind of syntactic (i.e. non-lexical) compounding, creating syntactic units which phonologically (maybe also semantically) behave like single words. Now, as is well known (cf. Liberman & Prince 1977), in compounds the accent goes to the immediate constituent in non-head position if the head is not itself a compound. If we apply this rule to lexical as well as to syntactic compounding {i.e. to integration), (C3) follows, . provided that the role of non-head immediate constituents in lexical compounds is played by arguments in syntactic compounds. Unfortunately, in certain cases, it is not clear whether (C2} and (C 3 ) make the right predictions. For example, in (S 1 s ) . application of (R3) to (NP,V) yields (T1 5) (Figure 1 8), which many speakers of German would not accept as a neutral stress pattern of this sentence.18 Probably the reason why the direct object cannot be easily integrated into the verb in this case is that these constituents function as separate informational units. The verbal phrase could be interpreted as a categorical statement about Schonhuber, thereby creating a
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
-
21
22
Focus ambiguities (SI S) ct
[+fl]
/ /-A
C
NP
weil
Tandler
I
I
NP
I
V
I
Schonhuber beneidet
(TIS)
R (R3)
weil
Tandler
Schonhuber
beneidet
Figure 1 8
topic-comment structure. But, of course, a lot.of further research will have to be done here. Another interesting problem is whether and to what extent constituents can inherit the ability to integrate arguments. In many cases, inheritance is already blocked by (CJ), which says that arguments cannot be integrated into constituents already containing other arguments. This condition, however, does not prevent inheritance of possible integration in cases where the head is modified by an adjunct. Indeed, other principles not withstanding (cf. (S6b)), the presence of adjuncts in general does not seem to be a barrier for the possible integration of arguments. For example, the object of a verb negated by nicht can be integrated just in case the object of the verb alone can be integrated, cf. wei/ er das Buch ( nicht) mitbrachte. In other cases, however, the ways of inheritance are not so clear. In a sentenc.e like Er hat ein Uch in die Wandgeschlagen (a perfect neutral stress pattern), the placement of the main stress would only be compatible with (C3) if we could assume that the integration of the direct object somehow 'ignores' the presence of the PP-argument. Inger Rosengren has proposed (in Rosengren 1989) that this can be explained by the fact that combinations of directional arguments and lexical predicates are treated as complex words (i.e. V 0-constituents) in syntax. (A different explanation is proposed in Uhmann 1991.) A last remark on integration: if the head of a constituent is moved, it is nevertheless 'visible' for integration. In other words, integration presupposes reconstruction of moved constituents in the position of their traces. This is the
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
� + (R2) / /--- + (R3) / � � .6. I �
JoachimJacobs
23
oilly way to explain the neutral stress pattern of e.g. Er schlug ein LOch in die Wand. (cf. the verb-final version of this example discussed above. Remember also that in cases like this the focus feature dominating the constituent bearing main stress will also dominate the trace of the moved constituent. cf. our remarks on example (S7).) Finally, s9mething must be said about the order in which {RI}-(R3) have to be applied. The reader may have noticed that there is a certain intrinsic ordering in these rules. This is shown in (IRO) (Figure I9). (IRO) follows from the conditions that are part of these rules and from the metacondition that every constituent must get a '+' or '-'. In one of the terminal nodes of (IRO)
Rl
integration in SC?
(R2) or (R3)
(R2)
Figure 19
there is a disjunction of rules. Nothing we have said so far prohibits application of (R2) to sets of constituents to which also (R3) could be applied. However, it seems that in cases like {PP,V} in (S12a), application of(R3) is obligatory; a plus plus-pattern here is impossible (unless, of course, there is a narrow focus on V). On the other hand, in {C,S) of (S1 2a) a plus-plus-pattern indeed seems to be a possible alternative to the minus-plus-pattern created by (R3). The solution to this problem is a ftlter on metrical grids, which I call 'Uhmann's ftlter', as the central idea behind it is taken from Uhmann (I 99 I ).
(UF) If a set �f sister constituents SC is in the domain of (R3), the �lement c1
of SC that is integrated into element c2 must be rhythmically stronger than c2 •
If (IRO) allows application of both (R2) and (R3) in a set SC, then it will be a consequence of (UF), in conjunction with the grid rules to be discussed below, that (R2) can be applied to SC only if the element c1 that is integrated into element c2 follows c2 in linear order. Thus, (UF) forces us to apply (R3) to e.g. {PP.V) in (S12a), but it leaves the choice between (R3) and (R2) in cases like {C,S} in (SI 2a).
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
ORO)
24 Focus
ambiguities 6 SYNTACTI C F O C U S A MB I G U I T I E S
The relation between focus-marked syntactic trees and plus-minus trees created by {R1}-(R3) and constrained by {C2), (C3) and (UF) is a many-to many-relation and therefore yields lots of syntactic focus ambiguities, as I will call them. Among them are the classic cases of focus projection. (T1 2) {Figure 20), for example, will be the prominence pattern not only of {S12a)
(T12) R
i
obwohl
I
-
er
-
I
in
A
i i
-
I
Miinchen wohnt Figure 20
above, but also of (S1 2b}-(S1 2e) (Figure 21). To see how this comes about consider example (S12c). In [C,S}, we have to apply (R1) and assign '+' to S, as this constituent dominates a focus feature. The same holds for the subject NP and VP: VP gets a '+' by (R1 ), because it is marked with [f1 ]. Further down in the tree, {R3) assigns the pluses, just as in (S1 2a). The result is (T1 2) (similarly' for the other variants of (S12) ). Note that this analysis of a classic case of focus projection as a syntactic focus ambiguity does not make use of any feature percolations, or, more generally, of any changes in given feature assignments. The impression that, given a certain pattern of stress, the focus feature can move up or down the tree is explained by the possible assignment of the same pattern of prominence to trees which differ in the hierarchical location of the focus feature, a possibility that follows from the independently motivated rules establishing the relation between syntactic structures and prominence trees. As a consequence, some unattractive assumptions of the percolation theories are avoided, for example the assump tion (see e.g. Rochemont 1986) that focus percolation follows a very special path not found in any other known processes of feature percolation, namely one which first leads from an argument to its head (e.g. from PP to V in (S1 2) ) and afterwards coincides with the usual head feature percolation lines (from V to VP, from VP to S etc.), provided that the first step (from argument to head) has actually been taken. ( Without this Hrst step, the focus feature must be '
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
i
_/ /::_� _
A
JoachimJacobs
25
(S12b)
ct
/£+�] :::VP / /.: � " ns ] [ I I I I
C
NP
I obwohl er
P
in
NP
V
Milnchen wohnt
/'s
/
C
I
(S12c)
t
� _;z [NP p
i1 I
obwohl er
in
c
NP
y
I
(S12d)
t
/£� [
C
I
NP P
-il l
obwohl er
in
c
NP
V
I
I
Milnchen wohnt (S12e)
t
a� [-ns] [ fl]
C
I
I
Milnchen wohnt
obwohl
NP p I er
I
in
+
NP
V
I I Milnchen wohnt Figure ZI
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
c
2.6 Focus
ambiguities
prohibited to follow head feature percolation lines, as a sentence like wei! er in does not have neutral stress.) Note further that our rules also explain focus projection in nominal phrases, e.g. in (S16a,b) (Figure 22). According to our rules, both (S16a) and (S16b) have the possible pattern of prominence depicted in (T16) (Figure 22). This will be quite clear for (S16b), but requires some explanation in the case of (S16a). In (S16a), rule (R2) will Hrst assign a minus-plus-pattern to {D,N1 }. Then (R3) can be applied to (N,PP}, as nach Munchen is a non-topical, non-pronominal, adjacent argument of Reise. Finally, P and NP get '-' and '+', respectively, by (RJ). (In the case of {N,PP}, (UF) would also license application of (R2), which would result in a plus-plus-pattern. Rhythmically, however, the result of this would be practically indistinguishable from (T16), as we will see below.) Munchen wohnt
(S16a)
[+fl]
[-ns] I die
I
Reise
I
nach Miinchen
NP N I /; D N�� p NP
[-ns] I die
I Reise
I nach
(S16b)
[fl] I
Miinchen (T16)
R
�+ j';• -
A
die
+ I I
Reise
_.....-...
+ I
+
.,.....
I
nach Miinchen
Figure 22
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
NP NI /; D N�� p NP
JoachimJacobs
27
7 THE RELAT I O N BETWEE N PRO M I NE N CE A N D RH Y THM
Now let us have a look at the metrical grids associated with prominence trees. The relation between these two kinds of representations is established by condition (C4): (C4)· A pair (T,G) is well formed if every set of sister-constituents SC of T is such that 1 and 2 are true: I. the designated syllables of every '+'-element of SC are stronger in G than all syllables of other elements of SC; 2. the designated syllables of all '+'-element of SC have the same strength in G.2• For example, (C4) allows us to say that the grids (G12) and (G1 2 ) are well formed with respect to (T12), which is repeated here for convenience (Figure 2 3 ). '
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Jorunn Hetland pointed out to me that applying (R3) to neutrally stressed NPs with a prenominal genitive would falsely predict that the genitive will receive strongest stress. However, (R3) can only be applied in such cases if the genitive is a (non-topical, non-pronominal etc.) argument of the head noun. This, however, is dubious, even in examples like Peters Ernennung, where the genitive intuitively specifies one of the semantic roles associated with the noun. Even here, the genitive arguably is not an argument of the noun in the same sense as, say, a direct object is an argument of the verb. Rather, the genitive is somehow a part of the head position of the NP. There is vast evidence for this, e.g. the fact that the genitive is in complementary distribution with the definite or indefinite article (which is the head of NPs in which it occurs). It is quite remarkable that we do not need any extra rules to account for focus projection in NPs, as this is in contradiction to what has been claimed in the literature several times}9 Researchers were puzzled by the fact that NPs have their main neutral stress on the final constituent, whereas verb-final VPs normally do not. Our systems of rules shows that this difference i� a consequence of the fact that in VPs possible integration goes from left to right, while in NPs it goes from right to left.20 By analogous considerations one could explain many of the differences between the neutral stress patterns of German and English sentences, but to demonstrate this would go far beyond the scope of this paper.
28 Focus ambiguities
(Tl2)
R
-
i
/ /---:-· / �
A
i
-
I
I
-
er
obwohl
in
A
i i
Miinchen
-
I
wohnt
Figure 23
•
•
•
•
•
•
•
(GI2)
• •
obwohl er in Miinchen wohnt •
•
•
•
•
•
•
•
•
•
•
•
•
(GI2 ) '
•
This shows, of course, that the relation between trees and grids is not one-to one. But it is not even a one-to-many relation. This is caused by certain optional operations on grids which transform them in order to meet requirements of euphony or to explain certain stylistic options. I cannot go into the details of this part of grid-phonology.22 However, one of these grid transformations must be discussed here, as it is of central importance to the topics of this paper. The name of this transformation will sound familiar. I call it the nuclear stress rule: (NSR) For any grid C corresponding to a full intonational phrase: one further occurrence of ••• may be added to the last of the highest columns of G. Look, for example, at (G17), one of the grids corresponding to (T17a) (Figure 24), which is the pattern of prominence associated with the multiple focus reading one gets in the environment of, say, wei/ er in M' iinchen wohnt, in Stuttgart arbeitet und in Bochum ein F&ienhiiuschen hat. weil er in Miinchen wohnt •
•
•
•
•
•
•
•
•
•
•
•
(G1 7)
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
obwohl er in Miinchen wohnt
•
JoachimJacobs
29
(T17a)
in
weil er
Miinchen wohnt Figure �4
weil er in Miinchen wohnt •
•
•
•
•
•
•
•
•
•
•
•
(GI 7 ' )
•
This explains the fact that-under certain stylistic conditions, e.g. relatively fast speech-the last syllable of (T17a) will be perceived as carrying the strongest beat even though it will be equal in prominence to the syllable Miin- in terms of FBS.
8
AMB I G U I T IE S O F P RO M I NE N C E
Given our condition (C4) above, (G1 7 ' ) is also one of the grids that are well formed with respect to (T1 7b) (Figure 25), which is the pattern of prominence (T17b)
weil
er
in
Miinchen wohnt
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(NSR) allows us to transform (G1 7) into (G1 7 ' ), under the assumption that wei/ er in Miinchen wohnt is a full intonational phrase:
30 Focus
ambiguities
associated with a single narrow focus on the verb. We therefore have here another source of ambiguity. Grids may be ambiguous with respect to prominence trees. To give a second example of this: {GI S) is ambiguous with respect to {TI 8a) and {TI S b) {Figure 26). (T18a)
R
·�
-
I
das
-
I
Haus
+
-
des
·
-
/)__
-
I I I
FUhrers in
+
I
-
Miinchen (T18b)
Figure 26
das Haus des Fuhrers in Miinchen •
•
•
•
•
•
•
•
•
•
•
{GI S)
•
In the case of {TI 8a), {G1 8) is licensed by {C4) and subsequent application of {NSR), while for {TI 8b), only {C4) must be applied to yield (G1 8).
9 NESTING O F F O C US A MB I G U I T IES
What is interesting in this-last example is that the possible readings of {G I 8) on the level of prominence are themselves ambiguous with respect to the next
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
1/--r/---
JoachirnJacobs
31
lower level o f representation, ie. with respect to syntactic focus. (Tx8a), for example, will be assigned both to (Sx 8a) and to (Sx8b) by (Rx)--(RJ) (Figure 27). (Sx 8a) is the all-new reading of this complex NP, while (Sx 8b) is the double focus evoked by environments like weder das Haus des Fuhrers in M' unchen noch das Reichtagsgebaude in Berlin noch die Parteizentrale in Wanne-'Eickel.
NP
(S18a)
" � I /;NP
pp
I
das
[.{" p�
N
N
I
I 1 Haus des
FUhrers
I
I
in
Munchen
NP
/;r� Nl
[-�]
I
das
/ A [-�]
N
I 1 Haus des
N
I
FUhrers
P
I
in
(S18b)
pp
A
NP
I
Miinchen Figure 27
This shows that the ambiguities we are discussing here may be nested. Such a nesting of focus ambiguities may even involve all the levels of representation introduced here, as the next and final example shows. The grid (G19) Peter hat ihm den Ball geklaut 0 0
0
0
0
0
0
0
0
0
0
0
(G19)
0
0
0
is ambiguous with respect to the patterns of prominence (T19a) and (T19b) (Figure 28). Furthermore, (TI9a) is ambiguous with respect to several focus marked $-structures, among them (S19a) and (S19b) (Figure 29).
B
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
[�]
32 Focus
ambiguities (T19a) R
·. . //. A ��
+
-
I I
Peter
-
-
-
hat e ihm
+
-
-
den Ball
+
geklaut
e (T19b)
A// � . � .
·.
+
-
I I
Peter
-
-
-
hat e ihm
-
+
-
den Ball
+
geklaut
e
Figure 2.8
Finally, (S19a) is ambiguous with respect to the semantic PBS-represent ations (M19a) and (M19b): (M19a) ASSERT(# AXw(HAT(IHM(X))) , 1-x (PETER(J.y] DEN BALL (J.z(KLAU(z)(x)(y)] )] )] #) (M19b) ASSERT(# AXNP I.Yv, (HAT(X(J.x(IHM(J.y{Y(y)(x)] )] ))] , PETER, hly(DEN BALL (J.z(KLAU(z)(x)(y)] )] #) (M19a) represents the single focus reading evoked by the question Was ist ihm geschehen?, while (M19b) represents the double focus reading after Wer hat ihm wtisgetan? This complex nesting of ambiguities is summed up in (AN):
(AN)
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
R
JoachimJacobs
33
(S19a)
Tl] I [-t [r] . [-il I
TOP
NP
NP
hat e ihm
D
den
I [-nsI ]
N
v
v
Ball
geklaut
e
(S19b)
TOP Peter hat e ihm
[-nsI ] I D
den
ns ] [ I I
N
v
v
Ball
geklaut
e
Figure 19
10
CONCLUSION
(AN) again shows clearly that the relation between surface strings with certain stress patterns and FBS is very complex. Especially, it shows that the ambiguities involved here are of quite different kinds. On the other hand, as we have seen, it is not necessary to assume mechanisms of focus feature percolation to explain any of the properties of sentences with focus ambiguities. As theories employ ing such mechanisms do not seem to be descriptively or explanatorily superior to the theory sketched here, I conclude that focus ambiguities-those that have been discussed in the literature on focus projection as well as those that have been overlooked-do not result from feature percolations.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Peter
C
34 Focus ambiguities
Acknowledgements I would like to thank Hans Bernhard Drubig, Jorunn Hetland, Tilman Hohle, Marga Reis, Ingr Rosengren, and Susanne Uhmann.
I have benefited
especially with regard to the issues raised in Section 5 ·
a lot from discussions with them,
JOACHIM JACOBS Universitiit Wuppertal FB4 6auBstr.2o D-s6oo Wupperral I Germany
I Such mechanisms are a central ingredient
particles, just like any constiruent in any
mont (I986) and Hetland (I989).
sentence, can be in the scope of an
2 E.g. without special pragmatic mechan isms of focus selection, cf. n. I 3·
3 'XA',
type
'Y,..'
A.
with
the
operator
representing
the
speaker's
intention to correct some utterance in the
etc.are variables of the logical
·
associated
category
syntactic
For example, 'XNP' is a
previous discourse. Consequently, (S6a) will be acceptable if (and only if) inter preted as a correction.
type ((e,t),t) (as NPs are
9 In Jacobs (I98 8b), I tried to explain such
treated as generalized quantifiers). 'x', 'y'
data by requiring that modal particles
variable of
etc. are variables of type e. Unanalysed
c-command the focus of the illocution
semantic
ary
representations
of syntactic
constiruents are written in small capitals. 4 The choice of category symbols is not
5
8 There is one exception, however. Modal
of several theories of FBS, e.g. of Roche
operator (or
a
trace
of it)
in
S-strucrure. This requirement can be deduced from (CI ) in the way just sketched. (6), by the way, will also be
essential here. Non-empty constiruents in the sense
unacceptable when the modal particle is
required here are constiruents containing
replaced by a sentence adverb like Ieider.
at least one non-empty terminal node.
This could be explained in the same way
6 The reason is that the focus NP cannot be
if we could assume that these adverbs,
interpreted as being 'scrambled' out of a
like modal particles, cannot be in the
position within the scope of the modal particle.
(As
is well known, focus con
stiruents cannot
be
leftward scrambled
within the 'Mittelfeld', cf. von Srechow &
Stemefeld I988: 466.) Consequently, the syntactic position of this NP is such that the NP must have scope over the particle. Given the [f1 )-assignment in (S6a),
this is
only possible if the particle is in the back ground of the operator focusing on the
object NP.
7 They
are
modifiers
of illocutionary
operators, cf.Jacobs (1988b).
scope of illocutionary operators. Indeed, there is ·independent evidence for
this
assumption, e.g. the fact that a sentence like Peter kommt Ieider cannot
be properly
paraphrased by Ich teile hiermit mit, dafi es zu bedauern ist, dafi Peter kommt.
10 Among other things, this will enable the stress rules
(see
Section
5)
to
correct neutral stress pattern
assign the
in cases like
Er schlug ein LOch in die Wand, where the
de-stressing of the PP in die Wand minors
the neutral stress pattern of ein LOch in die Wand schlagen , cf. Section S· Also we will
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
N O TE S
JoachimJacobs 3 5 get correct predictions in sentences with exttaposed constituents, like Er hat seinn Mutter vmprochm, da.fi er morgen kommt.
=
REFERE NCES Fuchs, A. (1 987), 'Pragmatic as against Hohle, T. (1982), 'Explikationen fiir grammatical factors in the determination ¥normale Betonung" und ¥norrnale Wort of accent placement', inJ. Verschueren & stellung" ', in W. Abraham (ed.), Satzglieder M Benucelli-Papi (eds), The Pragmatic im Deutschen , Narr, Tiibingen, 75-I 54Perspective , Benjamins, Amsterdam, 42 IHogg, R & C. B. McCully (I987), Metrical 34Phanology: A Coursebook , Cambridge U.P., Hayes, B. (I984), 'The phonology of rhythm Cambridge. in English', Linguistic lnquiry, 15: 33-74Jacobs, J. (I988a), 'Fokus-Hintergrund Hetland, J. (I989), 'Satzadverbien im Fokus', Gliederung und Gramrnarik', in H. Alt Diss. , University of Wuppertal (unpub mann (ed.), lnwnationsforschungen , lished). Niemeyer, Tiibingen, 89-I34-
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
I I The modal particle is represented as modifier of the illocutionary operator, cf. n. 7· 12 The pronominal dative object has been moved out of the VP in (S9). The struetural position of the auxiliary is discussed in Jacobs I989a. I 3 In Hetland (I989) and Morsch, Reis & Rosengren (I989), examples like (9), with a 'normal' stress pattern and a spreading of parts of a single complex semantic focus over several non-contiguous con stituents, have been adduced as evidence for the assumption that in addition to the grammatical rules of focus percolation there exists a special pragmatic mechan ism of focus selection which freely picks out as pragmatic 'rheme' one or more of the constituents dominated by an occur rence of the syntactic focus feature. The theory advocated here does not have to make use of such a mechanism to derive the possible foci in cases like (9). (Note that pragmatic focus selection fails to predict some of the possible readings, e.g. the single complex focus reading of (S8), namely (M8b). The reason is the absence of 'normal' stress in this case, which blocks the attachment of the focus feature to the highest node.) 14 'R' 'root'. The rest of the notation is explained below.
1 5 Again, as in {RI) and (IU), constituents that are not assigned '+' get '-' by default. I6 An indication of the rule that has been applied to sister-constituents can be found on their mother node in (TI 3). I7 Fuchs (I987) offers several examples of, as she calls it, concentric integration, where the head, rather than an argument or adjunct, receives main stress. However, in all of her examples (not all of them plausible instances of integration), the de-stressed element is an adjunct of the head. (For integration of heads into adjuncts cf. also Jacobs 1988a: 1 24.) 18 For these speakers, (TI 5) (which corres ponds to a grid with the strongest beat on Schon-) can only convey a narrow focus on the object. I9 For example in von Stechow & Uhmann (I984) and Krifka ( 1984). 20 The explanation of the different patterns of neutral stress in NPs and VPs is one of the major empirical problems of focus percolation theories, e.g. of Rochemont's (1986) theory; cf. Lenerz & Klein (1988). 21 The designated syllables of an element c of SC are those syllables of c that are dominated by the least number of occur rences of '-' in c. A syllable s, is stronger than a syllable s2 in a grid G if s1 has a higher column in G than s2• 22 A good overview can be found in Hogg & McCully (I987).
36
Focus
ambiguities Ph.D. Dissertation, University of Massa.., chusetts, Amherst. Rosengren, I. (1g89), ' W-Interrogativsatz, Skopus und Fokus', to appear in S&P Arbritsberichte. Sasse, H-J. (1987), 'The thetic/categorical . distinction revisited', Linguistics, :35: 5 1 18o. Stechow, A. von (1989), 'Focusing and back grounding operators', Arbeitspapier Nr. 6 der Fachgruppe Sprachwissenschaft, Con stance University. Stechow, A. von & S. Uhmann (1984), 'On the focus: pitch accent relation', GAGL , 25: 22J-6J. Stechow, A. von & W. Sternfeld (1988), Bau steine syntaktischen Wissens. Bin Lehrbuch der Generativen Grammatik , Westdeutseher Verlag, Opladen. Uhmann, S. (1991), Fokusphonologie. Nie meyer, Tiibingen..
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Jacobs, J. (1988b), 'The semantics of modal panicles', to appear in W. Abraham (ed.), Discourse Particles, Benjamins, Amsterdam. Jacobs, J. (1989a), 'Skopus und Kohiirenz', to appear. Jacobs,J. (1989b), 'Probleme der relationalen Fokustheorie', Wuppertal (unpublished). Krifka, M (198�). 'Fokus, Topik, syn taktische Strukrur und semantische Inter pretation', Munich (unpublished). Lenerz, J. & U. Klein (1988), 'Fokus Glasnost', S&P-Arbeitsberichte, 9: 16-3 5· Liberman, M & A. Prince (1977), 'On stress and linguistic rhythm', Linguistic Inquiry, 8: 249-3J6. Moesch, W, M Reis and I. Rosengren (1989), 'Zum Verhiiltnis von Satz und Text', S&P Arbeitsberichte, 1 1 : 1-36. Rochemont, M S. (1986), Focus in Generative Grammar, Benjamins, Amsterdam. Roath, M (1985), 'Association with focus',
Journal ojSmranrics 8: 37-49
© N.l.S. Foundation (1991)
Focus and the Levels of Language System PETR SGALL University Charles, Prague Abstract Recent linguistic trends stress the necessity to investigate not only the system oflanguage, but also discourse patterns. In this context it is important to understand the sentence both semantically and syntactically not just as 'an "assertion", being the linguistic counterpart of a
described in such a way that its properties imposed by the sentence's functioning in
communication are not neglected. This means above all that one must not neglect the topic focus articulation (rFA), which is one of the hierarchies constituting the syntactic pattern of the sentence, and which exists due to the impact of the communicative function on the structure of language. The results of empirical research continuing the tradition of the Prague School and taking into account the methodological requirements of formal linguistics have made it possible to state that TFA (being not only pragmatically, but also semantically relevant; see Section
I
below) can be described in an economical way as one of the aspects of underlying structure (or of sentence meaning, of tectogrammatics), and that this hierarchy has its counterparts (means of expression) on the other levels (see Section 2). The importance of the phenomena now subsumed under TFA has been known since Wei! ( I 8 44). Between the I 86os and I 920S, linguists such as G. von der Gabelentz, H. Paul, P. Wegener, A. Marty, H. Amman and 0. Jespersen introduced 'psychological subject and
predicate' (or 'theme and rheme', later 'topic and comment') into the analysis of general
properties of language, and discussed stress and word order as the means of expression of this dichotomy. In Prague, Mathesius ( I 929, I 939) formulated a linguistic account of the dichotomy from the viewpoint of a structural comparison of Czech (with its 'free' word order) and English, pointing out that the subject typically expresses topic in English (whereas its primary function in Czech is being the Actor). Among his followers, Firbas (I957, I 975)
analysed the interplay of this 'functional sentence perspective' and word order, showing that not only a dichotomy, but a whole scale of 'communicative dynamism' is present. After Halliday ( I 967) brought TFA nearer to the centre of attention of the English-speaking linguistic world, the relevant issues started to be studied also in the context of generative linguistics; see Chomsky ( I 97 I), more recendy Jacobs ( I 98 3) and Rochemont ( I 986). However, not much continuity with continental research has been kept within this trend, so that even in the rare cases in which TFA is given due attention, the approaches used for its analysis do not yield a possibility to describe adequately its position in the system of language Koktova, I 988a, b).
(see
esp.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
fact', but rather as a component part of a discourse, which in the general case is not a
monologue (Hoepelman & van Hoof I 988: 250). The sentence structure should then be
38
Focus and the levels oflanguage system
I.I To gain some insight into the issues ofTFA, let us first observe a few examples:
According to empirical investigations of English, Czech, German and other ·languages, it can be stated that the word-order switch in (I) has certain effects in parallel with the use of an 'inverse' verb form in (2) and with that of the secondary (marked} placement of the sentence stress (denoted by capitals) in (3). Examples parallel to (I}(d)-(f) can be easily added for (2) and (3), and it is not difficult to find other pairs of sentence parts which behave similarly as the from- and to-groups in (I}, or as the object and the prepositional case in (2) or (3); c£ Hajicova and Sgall (I987) on sending colleagues Christmas cards, sleeping quietly in a hotel, etc. It can be stated that the (a) sentences in (1)-(3) are unmarked, since, as the difference inacceptabilityof(c)and(d)intheexamples shows,in(a) boththecom plementations (inner participants and/or adverbials) standing after the verbs belong to the focus on one ofthe readings, which is not the case in (b). As (e) and (f) illustrate, for (a) as well as for (b) there exists a reading on which just one of the complementations belongs to the focus, the other being a part ofthe topic ofthe given meaning ofthe ambiguous sentence.• In other words, a relationship between two elements of the sentences A and B is present, such that if B is a part of the focus of the sentence on a given reading. then on this reading A also is a part of the focus, even though there is
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(I) (a) Jim moved from a village to an industrial centre. (b) Jim moved to an industrial centre from a village. (c) Jim didn't move from a village to an industrial centre; rather, he lived all the time in a quiet small town. . (d) ?Jim didn't mo�e to an industrial centre from a village; rather, he lived all the time in a quiet small town. (e) Jim didn't move from a village to an industrial centre; rather, he moved (from there) to a small town. (f) Jim didn't move to an industrial centre from a village; rather he moved (there) from a small town. . (2) (a) Tom made a canoe out of a log. (b) Tom made a log into a canoe. (c) Tom didn't make a canoe out of a log, but rather he swam across the nver. (3) (a) John dug a ditch with a hoe. (b) John dug a DITCH with a hoe. (c) John didn't dig a ditch with a hoe; rather, he pruned the trees.
Petr Sgall
39
another reading of the same sentence in ·which A and not B is a part of its focus. We say then that A is more dynamic
than B
in such a case. The scale of
(
)
dynamism thus is closely related to Chomsky's I 97 I 'range of permissible focus'. In (4)-even leaving aside certain details concerning the verb-the boundary between topic and focus can be placed in at least five positions (denoted below by slashes): (4) John/sent/a friend/a letter/to a spa/to please her. The position of the topic-focus boundary can be checked in the way proposed by Chomsky I 97 I , or by Posner's I 972 Kommentartest, as well as by
(
)
(
)
( s ),
and they correspond to the boundary positions indicated in (4).
(s ) John didn't send a friend a letter to a spa to please her; rather, he
(a) sent one to her to make her come back.
(b)
sent one to her at the parents' place.
(c) sent her a book of short stories.
(d) sent his sister a parcel. (e) went to his father's place.
As we shall see in Section 1 .3, the notion of dynamism can be generalized to include also those parts that on all the readings belong to the topic of a sentence.
1 .2
()
As we have seen with example 2 , where a direct object and a prepositional group switch their lexical setting, it is rather the level of underlying structure than that of surface syntax that is directly relevant here: if to make into is under stood as a secondary variant of to
make out of, we may say that the relation of
(
)
dynamism concerns the two theta roles (underlying cases, or Tesniere's I 9S9
participants and circumstantials, i.e. valency slots) ofObjective and ofOrigin. In
( I)
the two adverbial (circumstantial) complementations of Direction are
()
concerned, in 3 those of Objective and Instrument. On the basis of these and other examples it is possible to establish a tentative scale of the 'systemic ordering' of the theta roles; see Sgall
et al. ( I 973,
Actor-Addressee-Objective-Origin-Effect-MannerDirectionalftom-Instrument-Directionalto-Locative
)
I 986 . For English, the main
valency slots appear to be ordered as follows:2
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
the often used question test, i.e. by means of sets of acceptable contexts. The responses (a)-(e) are natural after which is the negative counterpart of (4),
40 Focus and the levels of language system
The relationship between systemic ordering (SO) and (communicative) dynamism can now be stated as follows: If A and B are two sentence parts (major constituents) such that the theta role of A precedes
that ofB under SO, then B can be less dynamic than A on a reading of a sentence only ifB is included in the topic on that reading.
(6) (Mary didn't go for a month to Mallorca;) she went for a weekend to her aunt in Belgium. (7) It is JOHN my brother met yesterday (, not Tom). In (6), the Actor together with the verb belongs to the topic, while the focus consists of the temporal adverbial and the Directional. In (7), the Actor and the verb, together with the temporal adverbial, constitute the topic, only the Objective belonging to the focus. Since the latter fact concerns all readings of (7), there clearly are sentences that have no reading on which the focus would cover the whole sentence (contrary to Chomsky's (1971) formulations). A similar case is also example (8):
(8) In Edinburgh a girl was given a HOUSE by her uncle. Here the Locative (standing more to the left than SO would allow) and the Actor (following the intonation centre, i.e. placed to the rightmost position by a shallow rule) belong to the topic in all readings.
1.3
The concept of dynamism, introduced in Section 1 . 1 above, can be generalized in such a way as to distinguish also different degrees of dynamism in those
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Recall that the SO of a language is determined by its grammar (and lexicon),3 whereas the dynamism is specific for individual sentences (more exactly, for their underlying representations). Our SO is manifested more or less directly by the surface order of the parts of the focus of a sentence,4 most of the deviations in surface word order from SO within focus can be accounted for by shallow rules (this concerns the order of head nouns, adjectives and prepositional cases in noun groups in different languages, as well as the grammatically determined positions of the verbs in German, and so on). It was shown elsewhere5 that neither topic nor focus are single constituents ofany level, in the general case; rather, either of them can contain several major constituents or occurrences of autosemantic lexical units. It is possible that the focus or the topic contain the main verb together with any ofits complementa tions (i.e. with a major constituent that has been assigned any theta role), c£ the following examples:
Petr Sgall 41
major constituents (sentence parts) which do not belong to the range of pennissible focus, i.e. which are included in the topic in all meanings of the sentence in question. The following examples, based on those introduced by Chomsky and others, can illustrate our viewpoint
The relationship between the scopes of the two quantifiers in the semantic inter pretations of the clefted versions clearly is identical to that of the original examples, having no cleft construction: e.g. on the preferred reading of ( I I )(a), a single set of girls is present and for each ofits elements there is a set ofproblems, whereas in (I I )(b) we have a single set ofproblems and several sets ofgirls. Since the surface structure of the two cleft sentences is identical on all points except for word order and placement of the intonation centre-these two being the primary means of expressing dynamism-a relatively economic description can . account for this relationship between the scopes by extending the scale of dynamism to the elements which (in all meanings of the sentence) belong to the topic, and by interpreting dynamism as a partial counterpart of the relation ships between the scopes of quantifiers. The latter cannot be identified only on the basis of the dichotomy of topic and focus, since in (9)-{I 1) both quantifiers belong to the topic, so that it can be seen that dynamism itselfis relevant for the semantic interpretation. The semantic relevance of TFA does not consist only in how it affects the scopes of quantifiers (of course, in such cases as with proper nouns the difference of dynamism yields just a redundant opposition of two underlying structures of sentences, since under semantic interpretation such a difference is not relevant, i.e. the two linguistic meanings correspond to the same truth conditions). The scope of negation depends on the boundary between topic and focus in the typical cases, see Hajicova (1973, 1983, 1984); for a brief illustration, c£ the following examples: (12) (a) This time our defeat wasn't caused by Harry. (b) This time Harry didn't cause our defeat
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(9) (a) It is TOM who assumes that everyone in this room speaks at least two languages. (b) It is TOM who assumes that at least two languages are spoken by everyone in this room. {10) (a) It is the PUBLISHER who believes that many people read few books. (b) It is the PUBLISHER who believes that few books are read by many people. ( u ) (a) It was JOHN who talked to many girls about few problems. (b) It was JOHN who talked about few problems to many girls.
42 Focus
and the levels oflanguage system
In {I2)(a)-pronounced with the intonation centre at the end-our defeat, belonging to the topic, is outside the scope of negation, and thus triggers the presupposition that the defeat did occur. On the other hand, in (12){b) this group is under the scope of negation, and thus the defeat is not presupposed: the utterance can be continued, e.g., by On the contrary, he helped us a lot to win. A similar case is that of the often discussed example ( I 3): {I 3) (a) The king of France is bald. (b) Yesterday, Prague was visited by the king of France.
{I4) (a) (b) ( I s ) (a) (b) {I6) (a) (b)
Jim studies linguistics on Sundays. On Sundays Jim studies linguistics. German is spoken in Liechtenstein. GERMAN is spoken in Liechtenstein. One smokes in the corridor. One SMOKES in the corridor. (Or: In the corridor one smokes.)
If it is established that there are cases in which the TFA of a sentence is relevant for the identification of the sentence's truth conditions (as is the case with such complex quantifiers as those in (9)-(I I ), with the scope of negation, and perhaps with some covert kind of quantification of the events in {I 4)-( I 6)), then it can be understood as appropriate to consider TFA to belong to the level of underlying structure {of sentence meaning), i.e. to the disambiguated level that serves as the input language for semantic interpretation.
In the prototypical case, the boundary between topic and focus (in the scale of dynamism understood as underlying word order) can be drawn just before or just after the main verb. Among the peripheral cases, the most important is that of so-called 'focalizers', of such words as even , only (also the sentential adverbs and negation belong here). These words have foci of their own and often occupy the boundary position between topic and focus, so that their focus is identical with that of the whole sentence, as in (17). In other cases they are included in the topic, or they constitute the single element of the focus, as in .
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
A presupposition is triggered, here again, by the noun group included in the topic in (a); however, in (b) the noun group belongs to the focus and thus is under the scope of negation in its negative counterpart, i.e. the presupposition is absent (the entailment has the character of Hajicova's 'allegation' here). Other, less clear examples of the relevance of TFA for truth conditions (and, afortiori, for linguistic meaning), are:
Pett Sgall 43
(I 8), in which case their scope coincides with the rest of the sentence (with the topic); a detailed analysis of the position of these adverbials in TFA was pre sented by Koktova (I 986). (I7) We found onlyJane's SISTER there. (I 8) Tom broke the window INTENTIONALLY. If the boundary position is not occupied by an overt focalizer then it is possible to understand the assertive:negative modality as a virtual focalizer (see Zemb I987; also Jacobs I983). It is possible to work with two more or less equivalent approaches here:
Since the focalizer can belong either to the topic or to the focus, it seems preferable to understand its primary position (on the boundary) as that of the least dynamic element of the focus, rather than as a third part of the sentence (between its topic and focus).
2
If we understand TFA as one of the hierarchies belonging to the underlying structure (together with the valency relations or theta roles, with the relations of coordination and apposition, and with the interplay of the values ofthe category of definiteness of noun groups), then we should also describe the correlates ofTFA on other levels of the language system.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(a) to take a sentence without a focalizer as the prototypical case, as we have just done, and to regard the focalizers in the position of the boundary between topic and focus as the primary occurrences 'of focalizers (c£ Hajicova's treatment of negation, quoted in Section 1.3 above), and to describe their other positions as specific peripheral cases (the syntactic conditions and the semantic interpretations of which have to be specified; an important step towards such a specification was presented in the above quoted book by Koktova); (b) to start the description from the general case (i.e. taking into account the different possible positions of focalizers) and to handle a sentence without any overt focalizer as displaying a 'zero' focalizer, i.e. the affirmative modality of the verb; however, it would then be necessary to admit the existence of such a zero focalizer also in such sentences where an overt focalizer is included in the topic of the sentence, e.g. Who did onlyJohn see? Only john saw MARY yesterday; it would not be quite clear in such a case whether the modality does not function as a focalizer when another focalizer occupies the boundary position.6
44
Focus and the levels oflanguage system
{19) Oohn called Mary a Republican). Then SHE insulted HIM. The fact that such constructions as Concerning her, . . . or Asfor her, . . can be used here corroborates this assumption. In other cases, the presence of more than one intonation centre is connected with the difference between the focus of the sentence as a whole and the focus of a focalizer in a secondary position. Up to now, it is difficult to analyse these cases from the viewpoints of syntax and semantics, since it seems that it has not yet been settled how many different kinds of sentential stress should be distinguished (as phonologically relevant, not only phonetically different). The opinions of linguists such as A LOtscher, C. Keijsper and others vary widely, and more reliable results concerning these issues can only be expected from the inquiries into phonological sentence patterns based on computer analysis of large sets of data. At the level of morphemics, as is well known, some languages in Eastern Asia and in Africa display affixes or particles directly expressing the topic, in a similar way as in other contexts the (morphemic and underlying) cases are expressed. Also the remarks concerning surface word order (and formulated above in connection with phonemics) have a direct bearing for this level (on which the sentence has the shape of a string, rather than that of a tree or of a more complex system). Surfoce syntax, if it is understood as a separate and complete lev:el of its own, includes such specific means of expressing TFA as the English cleft construc tion, left dislocation, topicalization, and also the above mentioned construc tions such as Concerning . . ., Asfor . . ., etc. As is well known, also one of the .
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Starting with the level of phonemics, we can state that in the phonemic patterns of the sentence, the position of the main stress {intonation centre) is the basic means of expressing the focus proper {the most dynamic element of the sentence, which in all its meanings belongs to its focus). This fact (illustrated e.g. by the examples (3), (7) and (14}-(16) above) has been acknowledged by most researchers in the field since P. Wegener. In the prototypical case, the intona tion centre occupies the rightmost position in the sentence; where the word order is not free enough for this-c£ (3)(b) above-the intonation centre occupies a non-final position in the surface shape of the sentence. In other cases such a secondary position of the intonation centre is stylistically marked (as emotionaL e.g. in what Mathesius called the 'subj ective order' in Czech). The situation is not always so simple that a sentence has just a single intona tion centre. In s�ch cases as those discussed by Jackendoff ( 1 972, especially p. 262), the intonation centre of his type B {standing to the left of the other one, type A) can be understood as being characteristic of a contrastive topic, whereas the rightmost intonation centre of type A expresses the focus (proper); c£ LakofPs example (c£ Sgall et al. 1 986: 221):
Petr Sgall 4 5
(2o) (a) The mother picked up the baby. (b) It had been crying nearly all the day. (c) She had been ironing all the afternoon. After the image of the baby has been foregrounded by being mentioned in the focus of(2o)(a), it is natural to refer to the baby in the topic of the next utterance (in the terminology used in anificial intelligence, the image of the baby is now included in the focus list), as is the case in (2o)(b). However, instead of the latter, an utterance such as (2o)(c) can now be used also, with a weak pronoun in the
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
functions of passivization consists in fronting an Objective (or Addressee) included in the copic-c£ examples (9), (I o), ( 1 2), (I s ) above-where many languages, co a certain extent also German, have word order possibilities free enough not co use passive; chis shows char in the quoted English examples the difference in TFA is directly expressed rather by chat in word order, than by the opposition of active and passive, passivization being triggered here by the specific shape of the scale of dynamism in the underlying structure. Also such inversion forms as make into (along with the primary from make out of) have a similar function; c£ example (2) above. As for the layer of discoursepatterns, we would like co scare first that we do not understand it as another level of the language system, since discourse is a kind of human activity (of communication) in which the role oflanguage is that of a means used to achieve the goals of the communication. Therefore we do not speak about text grammars and do not look for units of language 'larger chan the sentence' (a more detailed account of our motivation was presented in Sgall et al. (I986: Section J.7I); see also Kahre (I986). In chis domain, it is indispens able to analyse linguistic phenomena in a close connection with their psycho logical and cognitive background. We would like to present here a shon outline of an inquiry into the psycho logical background of co-reference as connected with the TFA of the sentence (in more derail, see Hajicova & Vrbova I98I, I 982; Sgall etal. I 986: Section 3.73; Hajicova I987). This approach cakes into account the dynamic aspects of the discourse. At the initial point of every discourse, the interlocutors share a certain stock of information (the 'stock of shared knowledge', SSK), a part of which is activated by the situational context.7 During the discourse SSK changes in accordance with what is at the centre of attention at the given time point, what is most salient or foregrounded (activated) in the memory, and what has just been said (or observed). In a smooch discourse, the speaker chooses only chose items to be used in the topic of an utterance which s/he supposes co be among the most salient in the hearer's stock of information; chis enables the hearer relatively easily to identify the objects referred co by pares of the topics of the utterance. Let us consider the following example:
46 Focus and the levels oflanguage system
topic referring to an item that is mentioned in the topic of the preceding utterance. The following observations are rather obvious, as to how the changes of the degrees of salience depend on the TFA of the sentence uttered:
position in the discourse; e.g. utterances metatextually opening a narration or one of its parts, or headings are more prominent than other utterances in the discourse, in that the items introduced by the former retain their activation longer than those introduced by the latter utterances.
It has been pointed out in the writings by Hajicova and others just referred to that these observations can. to a cenain degree, be formulated in a framework that makes it possible to understand how the issues of reference assignment can be captured by finite means, and to describe such (up to now rather vague)
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(a) The items referred to by the focus of the immediately preceding utterance are the most activated ones at every time point of the discourse (in so far as the discourse has not been interrupted by a strong impression coming from outside). (b) If an item b is mentioned in the topic, then at least two issues are to be taken into consideration: (i) a pronominal reference strengthens the activation of the item referred to to a lesser degree than a reference with a full (definite) noun group; (ii) the activation of the item referred to in the topic fades away less quickly than that of the item referred to only in the focus of the preceding utterance(s). (c) Also an item lacking a high degree of activation can be mentioned in the topic, if natural language inferencing makes it easy for the hearer to identify the item mentioned (on the basis of associative links). (d) If an item ofSSK is not referred co in the given utterance, then its activation lowers down (c£ point (b) (ii) above). (e) There are specific expressions which introduce a topic that was not expected, e.g. recalling an object that is being mentioned only after an occurrence of several intervening utterances (Asfor . . . , Concerning . . . ). (fj Under cenain conditions, which still have co be specified, the set of salient items in SSK is reduced so that only a small number of items interesting the interlocutors in the given context is taken into consideration as the relevant partial universe of discourse; c£ e.g. the appropriate use of(I s)(a) after such a restrictive context as In which of the smallest European countries is GERMAN spoken, in Monaco or in Liechtenstein?, as compared with the prototypical contexts, in which (I s)(b), rather than ( I s)(a), may be used as an appropriate and (fully) true utterance.8 (g) Also other scales of hierarchies should be considered. Thus, there is a hierarchy ofprominence with regard to the individual utterances and their
Petr Sgall
47
notions as that of the topic of (a part o� a discourse. Further research based on this viewpoint and on the description of the TFA of a sentence briefly outlined here (and presented in more detail in Sgall eta/. 1986, with a framework suitable for handling the individual lexical occurrences in a sentence with its recursive properties) may contribute not only to a description of the ways in which a specific language system is used in discourse, but also to investigations of relationships between language, the structure of memory and the patterning of cognitive processes.
NOTE S I Our terms 'meaning' and 'reading' perhaps can be replaced by 'logical form'. However, we prefer to use a description including just a single underlying structure (dis ambiguated, although language-specific in some values or even categories); see a detailed discussion motivating this choice in Sgall et a/. (Ig86), where a definition of the notions 'focus' and 'topic' within the given framework was also presented. 2 Even if some of our examples deserve further discussion, the principle itself, i.e. the existence of SO, seems not to be affected by empirical questions about details. SO is in part language-specific (e.g. in English the Objective or Patient pre cedes several free adverbials, whereas in German or some other continental lan guages the corresponding adverbials pre cede the Objective). We assume (with Chomsky and Fillmore) that every occur rence of an autosemantic lexical unit different from the main verb carries a single theta role (or deep case), i.e. that it occupies a single valency slot. Prepositions, conjunctions, auxiliary verbs and empty subjects emerge as lexical units only on the level of surface syntax or of morphemics; on the underlying level their counterparts
occur as indices or parts of the complex labels of the nodes. 3 The prototypical choice of SO does not depend on the individual verbs: However, there are peripheral cases in which one verb may differ from another (perhaps with some variation which may be con nected with a process of change of SO, such as those processes which gave birth to the difference between English and Ger man mentioned in n.2 above); e.g. the relationship between Addressee and Objective may vary between groups of verbs, so that SO should be specified separately, in marked cases, for individual verbs in their respective theta grids or case (valency) frames. 4 In this point SO differs from various discussed orderings of parts of the sent ence, the authors of which either do not pay attention to the difference between the topic and the focus, or do not draw a clear boundary line between underlying struc ture and shallow rules (and mostly do not classify the valency slots in detail to the necessary degree). 5 See Hajicova and Sgall (I975), where the difference between topic and presupposi tion is also discussed.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
PETR SGALL University Charles Faculty of Philosophy Celetna I 3 I 1000 Prague
48 Focus and the levels of language system
6
If the affirmative modality plays the role of
activation of elements of SSK is the object
a focalizer, e.g. in Jack helped hisfather, then it
is
not easy to understand why it should
lack this role in Jack readily helped hisfather;
.7
of a similar assumption made by the speaker.
8
This reduction of the salient part of SSK
in any case, in Jack helped even hisfather the
belongs to a domain of the impact of the
modality (together with the lexical mean ing of the verb) is included in the topic.
context similar to that of the pre-check
More precisely, SSK can be characterized
ence on 'small number'; see Hoepelman van Hoof(I982: 25 2).
as the stock of information the speaker
needed to prevent the paradoxical infer
&
assumes to share with the hearer(s); also the
Chomsky, N.
( I 97 I), 'Deep strucrure, surface
salience of the elements of the stock of
strucrure and semantic representation', in
shared knowledge',
D. D. Steinberg
29I-303.
& L. A. Jakobovits (eds), Semantics: An Interdisciplinary Reader, CUP, Cambridge, I93-216. Firbas, J. (I957), 'Some thoughts on the
Folia Linguistica , XV:
Haj icova, E. & J. Vrbova
(1982), 'On the role
of the hierarchy of activation in the process of natural language understand
COUNG
1982, J. Horeck}' (ed.),
function of word-order in Old English
ing',
and Modern English', Sbomik praci filo
North-Holland, Amsterdam, 107- I 3.
zoficke fakulty brnenske univerzity As, 72-I OO.
Halliday, M A. K. (1967), 'Notes on transitiv ity and theme in English',
Firbas, J. (I97 S), 'On the thematic and the non-thematic section of the sentence', in
H Ringbom et al. (eds), Style and Text: Studies Presented to N. E. Enkvist, Skriptor, Stockholm, 3 I 7-3 4·
Hajicova, E. ( I 973), 'Negation and topic vs.
Linguistics, 3: 37-8 1 , 199-244; 1 79-21 S ·
journal of 4 (1968):
Hoepelman,J. Ph. & A.J . M . van Hoo£(1988), 'The success of failure',
COUNG
1988,
Vol. I, John von Neumann Society for Computing Sciences, Budapest, 25o-4.
Philologica Pragensia , I6: 8 I -93· Hajicova, E. ( 1 983), 'Topic and focus', Theor etical Linguistics, 10: 268-76.
Jackendoff, R S. (1 972), Semantic Interpretation
Hajicova, E. (I 984).
Jacobs, J. ( 1 98 3 ),
comment',
'Presupposition and
allegation revisited',jou mal ofPragmatics, 8:
in Generative Grammar,
MIT Press, Cam
bridge, Mass.
Fokus und Skalen: Zur Syntax und Semantik der Gradpartikeln im Deutschen,
Niemeyer, Tiibingen.
1 5 5-67. Hajicova, E. (I987), 'Focusing: a meeting
Kohrt,
M.
(I 986), 'Dialoggrammatik und/
point of linguistics and Artificial Intellig
oder Konversationsanalyse?', in F. Hunds
ence', in Ph. Jorcand & V. Sgurev (eds), Artificial Intelligence II: Methods, Systems, Applications, North-Holland, Amsterdam,
& E. Weigand (eds), Dialog analyse, Niemeyer, Tiibingen, 69-82. Koktov:i, E. ( 1 986), Sentence Adverbials in a Functional Description, Benjamins, Amster
3 1 1 -2 1 . Hajicova, E .
&
P . Sgall (I97S). 'Topic and
focus in transformational grammar', Papers
in Linguistics, 8: 3-5 8. & P. Sgall (1 987), 'The ordering principle',journal ofPragmatics, I I: 43 S-54· Hajicova, E. & J. Vrbova (198I), 'On the Hajicov:i, E.
nurscher
dam. Koktov:i, E. (r988a), Review ofJacobs (r98 3),
journal oJPragmatics, n: 253-70. Koktova, E. ( I 988b), Review of Rochemont
(1986),journal ofPragmatics,
12: 24 1 - S J .
Mathesius, V. (1 929), 'Zur Satzperspektive im
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
RE F E RE N C E S
Petr Sgall 49 Sgall, P., E. Hajicova &J. Panevova ( 1 986), The Meaning of the Sentence in its Semantic and Pragmatic Aspects, Reidel, Dordrecht and Academia, Prague. Tesniere, L (1959), Eliments de syntaxe struc turale, Klincksieck, Paris. Weil, H. (1 844), De l'ordre des mots dans les langues anciennes comparies aux langues modernes, Joubert, Paris; translated as The Order of Words in the Ancient LAnguages Compared with that of the Modern LAnguages, Boston ( 1 887), re-edited, Benjarnins, Amsterdam (1978). Zemb,J.-M. (1 987), 'Zum Begriff des "Pradi kats" ', in W. Wilss & K.-D. Schmitz (eds), Maschinelle Obersetzung: Methoden und Werkzeuge, Niemeyer, Tiibingen, 101- 1 1 .
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
modemen Englisch', Archivfur das Studium der nn�eren Sprachen und Literaturen , ISS= 202-10. Mathesius, v. (1939). ·o tak zvanem aktual nim Cleneni very', Slovo a slovesnost, s: 1714; translated as 'On information-bearing structure of the sentence', in S. Kuno (ed.), Harvard Studies in Syntax and Semantics, Harvard University Press, Cambridge, Mass. (1975), 467-80. Posner, R (1972), Theorie des Kommentierens, Athenaum, Frankfurt-on-Main. Rochemont, M S. (1986), Focus in Generative Grammar, Benjarnins, Amsterdam. Sgall, P., E. Hajicova & E. Bene5ova (1 973), Topic, Focus and Generative Semantics, Scriptor, Kronbergffaunus.
Jourru�l of&mtlntics 8: S 1-70
C
N.I.S. Foundation (1991)
Some Remarks on Focus Adverbs JACK H OEKSEMA a n d FRANS ZWARTS University oJGroningen Abstract Much of what we have to say in the present paper should be thought of as an attempt to get a better understanding of the intricate complexities surrounding focus adverbs. To this end, we discuss a number of restrictions which govern the occurrence of different classes of focus primarily to the descriptive level-a decision which some may well disapprove o£ Indeed, when Pullum (1 988) speaks of the 'slow agony of empirical endeavor', he characterizes this position in the following way: 'Simply listing facts will do us no good . . . It it easy. It is boring. Anyone can do it But ultimately it is useless to serious linguistic research.' It goes without saying that we, too, completely disagree with this way of presenting the matter. When one tries to characterize focus phonemena, it is immediately clear that there are numerous lexical differences which interact in subtle ways with regular syntactic and semantic patterns. In our opinion, a linguistically interesting theory of focus should be able to account for these
(1976, 1978) and Jacobs (198 3), and implicitly by many others, which is to concentrate a few common focus particles (often the words for only and even ), and to hope that the rest will somehow conform to the
differential patterns. The sttategy followed explicitly by Altmann
patterns exhibited by the selected items, may lead to some initial progress, but must eventually be replaced by a more comprehensive efforr. It is dangerous to rely too much on the assumed homogeneity of linguistic classes, especially in the closed-class systems. It is also important to gain a comparative perspective on focus adverbs by comparing items from different languages if a theory is to be constructed with the explanatory depth and broad empirical coverage of current theories of WH-movement or anaphoric dependencies. We give a number of examples where it is useful to compare items taken from English, German and Dutch. To summarize: this paper reviews some of the major problems which a comprehensive theory of focus adverbs needs to address, describes some of the variations to be found among focus adverbs, and places this against the background of some of the available accounts of focus adverbs.
1.
W H A T I S A F O C U S AD VERB ?
1.1
Focus and alternatives
To set the stage for our remarks on focus adverbs, it will be useful to briefly consider our terminology. The term 'focus', while perhaps not so badly a�used as its peer 'topic', has been used with quite a few intended referents. It is there fore necessary to indicate what we mean by a focus constituent. For us, a
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
adverbs in Dutch, German and English. We are aware that in doing so we limit ourselves
52 Some remarks on focus adverbs
linguistic expression is a focus expression when it has an accentual peak or stress which is used to contrast or compare this item either explicitly or implicitly with a set of alternatives. So, for instance, both Amos and Hilary are focus expressions in (I). (I) Jedediah chastised AMOS, and not HILARY. In (I), there is an explicit comparison and contrasting of Amos and Hilary. Sentences such as (2) show that the contrast in question does not have to be explicit (2) Jedediah only chastised AMOS.
(3) Did Jedediah chastise Amos? The adverb only, used in (2), is called a focus adverb, because it requires a focus constituent in its environment 1 The reader can check this quickly by noticing what happens when the focus expression Amos is replaced by an unstressed pronoun. In that case the sentence becomes ungrammatical, unless heavy accent is placed on the verb chastised. If the latter happens, the verb will serve as the focus associated with only, and the interpretation of the sentence changes accordingly. Now, the relation 'chastise' is compared with other relevant relations that might obtain between Jedediah and Amos, and it is asserted that only 'chastise' actually holds. For an explicit theory of focus along the lines sketched informally here, we refer to Rooth (I 985). Our understanding of the term 'focus' is that ofRooth, and has nothing whatsoever to do with discourse notions such as 'new information' or the very different use of the term found in AI approaches to automated understanding of discourse such as Grosz (I977) and Sidner (I979). 1 .2
Focus and scales
A distinction is sometimes made between focus particles and scalar particles. We think that scalar particles form a subclass of the focus particles. In particular, it seems to us that scalar particles are distinguished by the ordering which they impose on the set of alternatives. Consider, for instance, sentence (4).
(4) We are only linguists.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
In (2), Amos is implicitly contrasted with other individuals under consideration who were not chastised by Jedediah. If no other individuals are in fact under consideration, (2) cannot be uttered felicitously. So, for example, (2) is not a proper answer to question (3), because (3) does not concern itself with what happened to those other than Amos.
Jack Hoeksema and Frans Zwarts
53
2.
SYNTA C T I C PROPERTIES O F F O C U S ADVERBS 2.1
Two types ofuses: phrasal qnd sentential
In this section, we review some evidence for making a two-way distinction between focus adverbs with sentential scope and focus adverbs with phrasal scope. The evidence consists of a number of differences in behaviour, such as -
order constraints; c-command constraints; stress; scope ambiguity.
2. 1 . 1 Order
The syntactic relation between a focus adverb and its associated focus con stituent or constituents is subject to considerable variation. It is clear, for example, that some focus adverbs must precede their targets, that some must follow them and that yet others may occur in either position. Examples illustrating this claim are given in (sa-f). Similar data could be given for Dutch or German. (s) Obligatory Precedence. a. Even then, Nixon lied. b. 4PJ'hen, even, Nixon lied.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Under the scalar interpretation, one thinks of an ordered set of alternatives for the interpretation of linguists, say a set of predicates indicating professional status, such that the property of being a linguist is towards the bottom end of the list and the claim is made that no higher predicate applies to the speaker. Under the non-scalar interpretation, no such ranking is understood, and it is asserted that none of the alternatives applies to the speaker. In Dutch, the two readings are expressed by different adverbs, namely maar, 'but only', for the scalar reading and aileen , 'only, alone', for the non-scalar interpretation. Similarly, the difference between even and also appears to be simply that even posits its associated focus toward the low end of some probability ranking, whereas also assumes no such ordering. Further properties of even , such as its quasi-universal character, can be derived from its scalar behaviour by straight forward pragmatic inferences (c£ Fauconnier 1975).2
54
Some remarks on focus adverbs Illegal Precedence. c. O>'foo Nixon lied. d. Nixon too lied. Optional Precedence. e.
£
She lied only for Nixon. She lied for Nixon only.
In these sentences, the focus adverb is directly adjacent to the focus expression. The data seem to suggest a fairly unsurprising binary distinction between
Notice, for example, the difference between (6a) and (6b). (6) a.
The POPE has even permitted this practice
b. O>J'he POPE has only permitted this practice. The intended readings of these sentences are those of(7a) and (7b), respectively. (7) a. Even the Pope has permitted this practice. b. Only the Pope has permitted this practice. Whereas the examples in
( s)' purported to
show that
even
must precede its
target, while only may precede it, the examples in (6) show the converse pattern. Here, even may follow the focused expression, while only may not. It seems to us that these examples make it necessary to distinguish between cases where the
( s ).
focus adverb forms a syntactic constituent with the focused expression, as in and cases where it is not, as in (6). We further note that the difference between even and only in (6) does not, perhaps, reflect accidental lexical idiosyncracies of these two items (as Rooth 1 98 5 supposed), since it can be replicated in Dutch and German with their semantic counterparts (which, by the way, are not cognates). (8) a.
De PAUS heeft zelfs dit gebruik. toegestaan.
b. OUe PAUS heeft aileen dit gebruik toegestaan. (9) a. Der PAPST hat sogar diesen Brauch erlaubt. b. OUer PAPST hat nur diesen Brauch erlaubt.
2. 1 .2 c-command Next, observe that in the preceding examples even does not c-command its associated target, assuming a non-controversial view of Dutch and German phrase structure. Given the ubiquity of the notion of scope defined by the c-command relation, this is a remarkable finding. Paradoxically, the next set of
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
forward-looking and backward-looking functors, with some items belonging to both classes. However, when sentences are considered in which the focus adverb is separated from its target, the situation becomes more complex.
Jack Hoeksema and Frans Zwans
55
examples shows that nevertheless the c-command relation is relevant to the way even is associated with focus. (The contrast between sentences such as (10a) and (I ob-c) was first noted in Jackendoff I 972). ( 10) English a. EvenJOHN likes Bill. b. �venJohn LIKES Bill. c. �venJohn likes Bll.L.3
German g. Selbst iiber HANS wurde geredet. h. "'Selbst iiber Hans wurde GEREDET. These sentences exemplify the use of even and its counterparts as phrasal operators. Their scope is restricted to the phrase they attach to. Elements outside that phrase are not c-commanded by the focus adverb and cannot serve as the focus expression associated with it. So again we see that it is necessary to distinguish a phrasal use from another, sentential, use. There is a parallel here with negation, for which traditionally a distinction is made between sentential and phrasal negation. We propose that this distinction is not a peculiarity of negation, but a more general characteristic of a subclass of the focus adverbs. Sentential scope of even and negation is restricted to clausal domains. Thus, sentence (I I) may be interpreted as equivalent with (I 2), but (I 3) does not have the interpretation corresponding to {I4). (I I) EVERYBODY does not want to leave East Germany. (1 2) Not everybody wants to leave East Germany. {I 3) EVERYBODY thinks I do not want to leave. {I4) Not everybody thinks I want to leave. Similarly for even : ( I 5) DIETER has even left East Germany. {I6) Even Dieter has left East Germany. (17) DIETER thought I'd even leave. (I S} Even Dieter thought I'd leave.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Dutch d. Zelfs JAAP vind ik leuk. EvenJaapfind I nice 'I find even Jaap nice.' e. "'Zelfs Jaap vind IK leuk. £ "'Zelfs Jaap vind ik LEUK.
56 Some remarks on focus adverbs
2. 1 .3
Stress
Quite similar to the behaviour of even is that of Dutch ook, 'also',and its German counterpart auch. Again, we observe the possibility of phrasal and sentential scope. ( 1 9)
a.
Oook HEIN heeft gezondigd. Also Hein has sinned
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(Similar examples could be given here for Dutch and German.) In making a distinction between sentential and phrasal use, we diverge from the proposals in Jacobs (I98 3). See Hoeksema (I989) for a critical discussion ofJacobs's analysis. We also note that in the semantic analysis of focus adverbs given in Roath (I 985), it is critically assumed that the focus adverb always c-commands the expression in focus. To account for sentences such as (I I), Roath assumes adjunction of the focus adverb to S at the level ofLogical Form. It is not entirely clear to us how such a movement account could prevent iterated adjunction, leading to a wide-scope reading of the adverb in (I 7) which is equivalent to (I 8). Roath's proposal is the mirror-image, in a way, of the quantifying-in solution to the same problem proposed by Jacobs (I983). Jacobs suggests that a focus adverb may combine with its scope domain by means of infixation, such that an adverb which has scope over a sentence may end up between the constituents of that sentence. He also points out that wide-scope negation, as in (I I) above, provides independent evidence for such a mechanism. We are not satisfied either with Roath's raising account or with Jacobs's infixation account. They fail crucially to take into consideration the sensitivity of wide-scope readings to semantic properties of the item taking wide scope and the operator over which it takes wide scope. It is well known from the literature on scope ambiguities (e.g. Kroch I974) that wide-scope readings as exemplified by (I I) are not available when the subject expression is changed into, say, an existential quantifier. Negative polarity items in subject position are never licensed by wide-scope negation, in spite of what the infixation and quantifier raising mechanisms would predict (c£ Ladusaw I98o; Hoekstra, de Hoop and Zwarts (I984).• We conclude that the most popular mechanisms for dealing with scope ambiguities, Quantifying-In and Quantifier Raising, are not well suited for the treatment of focus adverbs, including negation. We emphasize that the variation in scopal be.haviour is not random, but rather stable across the three languages English, German and Dutch, and presumably many more. This suggests a more semantically driven account, which takes into consideration the meaning of the scope-taking elements. Unfortuately, not much has happened since Kroch (I 974) towards the articulation of such a theory.
Jack Hoeksema and Frans Zwarts
57
b.HEIN heeft OOK gezondigd. Hein has also sinned
We have not been able to ascertain why the stress properties of zelfs and ook differ so significantly, in spite of their otherwise striking similarities (as discussed in e.g. Plank 1979). However, they do not appear to be random idiosyncracies, since their German and English counterparts exhibit similar dif ferences in behaviour. 2. 1 .4 Scope ambiguities
Rooth (198 5) points out, phrases with focus adverbs exhibit scope inter actions with quantifiers elsewhere in the sentence, whereas focused expressions linked to a sentential focus adverb do not engage in such interactions. We illustrate this claim with some ofRooth's examples in (22).
As
(22)
We are required to study only physics. b. We are required to only study physics.
a.
Sentence (22a) has two readings. One reading states that the only required subject is physics. The other, perhaps more accessible, reading states that we are
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
In the sentential, but not in the phrasal use, this adverb can bear stress, unlike zelfs , 'even'. In this respect, ook is more similar to niet, the negation adverb and wei, the adverb of emphatic affirmation: (2o) a. HEIN heb ik. NIET gezien. Hein have I not seen 'Hein I did NOT see.' b. FRED heb ik. WEL gezien. Fred have I affseen 'Fred I DID see.' The possibility of bearing stress also makes it possible to use these adverbs as remnants of the gapping construction. Zelfs, being unstressed, does not occur in this function. (21) a. HEIN komt, en FRED OOK Hein comes, and Fred too 'Hein is coming, and so is Fred.' b. FRED komt NIET, maar HEIN WEL. Fred comes not, but Hein aff 'Fred is not coming, but Hein is.' c. "'HEIN komt, en FRED ZELFS. Hein comes, and Fred even 'Hein is coming, and even Fred is.;
58 Some remarks on focus adverbs required to study nothing but physics. Hence if we study both physics and maths, we meet the requirement of the first reading, but not the stronger one associated with the second reading. Sentence (22b), on the other hand, is unambiguous, since it lacks the first reading. Rooth's findings are to be expected, given that quantifier expressions such as only physics may generally take wide scope, even when embedded in a subordinate clause, whereas sentential focus adverbs, as we noted before, do not take scope outside their dause.5
2.2.
As noted in Anderson ( 1 972) and Rooth (1985), a focus adverb such as
even can be associated with more than one focused expression. Anderson illustrates his observation with the following example:
(23) John claims tht he can sell refrigerators to Eskimos, but in fact he couldn't even sell WHISKEY to the INDIANS. There is a question here concerning the scale relative to which these focus constructions are to · be interpreted. For another kind of multiple focus construction, to wit, gapping in conjunctions with let alone, Fillmore, Kay & O'Connor (1988) proposed multidimensional scales, of the kind familiar in sociolinguistics as implicational or Guttman scales. Consider for example sentence (24). (24) John couldn't sell whiskey to the Indians, let alone refrigerators to the Eskimos. In Fillmore et al.'s opinion, such a sentence implies not only that selling refrigerators to Eskimos is more difficult than selling whiskey to the Indians, but also that selling refrigerators is more difficult than selling whiskey and that selling to the Eskimos is more difficult than selling to the Indians. Hence we are justified in positing a two-dimensional scale as the one in (25), such that any slot in the scale is ranked at least as high as any slot below or to the right of it. (25) Two-dimensional Scale of Sales Difficulty Indians Eskimos 2 whiskey 4 3 refrigerators However, while two-dimensional scales might be applicable to the paired contrasts of the gapping constrUction, we feel they are not appropriate for all multiple-focus constrUctions. A sentence such as (23) merely implies, in our
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Arguments against a transformational link between phrasal and sentential uses
Jack Hoeksema and Frans Zwarts
59
opltllon. that the pair (whiskey, Indians) is ranked lower on some one dimensional scale of salesmanship prowess than the pair (refrigerators, Eskimos). In other words, (23) might still be used felicitously when the dif ficulties of selling are as indicated in (26), which is not a perfect Guttman scale.
(26)
Indians whiskey ' refrigerators
Eskimos
4
2 3
account would predict island conditions on the relation between adverb and focus, and such conditions do not appear to exist, as Anderson (1972) has already pointed out For instance, the complex NP condition ofRoss (1 ¢7) can be violated without any problem. One of Anderson's examples is given in (27).
(27) John even has the idea that HE is tall for a Watusi. In this respect, focus adverbs show a different behaviour than other focus constructions, such as gapping and exception phrase constructions, which do exhibit the effects of island constraints.
(28) Tom arrested the guy who raped Suzy and Bill, Sally. (29) Except for him, I've met nobody who likes everybody. Sentence (28) does not have the interpretation according to which Bill arrrested the guy who raped Sally, and sentence (24) does not have the reading 'I've met nobody who likes everybody but him'.
3 · C O N S T R A I NTS O N P H R A S A L F O C U S A D VERBS Phrasal focus adverbs do not attach themselves to appropriate phrases under all circumstances. For instance, evenJim is possible as a subject or direct object, but not usually as an object of a preposition.
(3o)
a.
Even Jim left the ship.
b. This bothered even Jim. c.
1)1 sat next to even Jim.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The possibility of multiple foci is also interesting for another reason. It is possible to develop a transformational account of association with focus along the following lines. Consider the phrasal use of focus adverbs basic and derive the sentential use from the phrasal use by moving the focus adverb away from the focus. This would immediately explain. for instance, why focus adverbs usually c-command their associated focus. However, it would not explain the exceptions to the c-command requirement that we have seen. nor is it compatible with the possibility of multiple foci. Finally, such a movement
6o Some remarks on focus adverbs Both Jacobs ( 1 98 3, 1 986) and Roath (198 5) consider this to be a possible argument against assigning constituent status to the string evenjim. The main difference is that Jacobs takes the argument seriously, while Rooth ( 1 98 5) considers the arguments in favour of constituenthood (some of which were reviewed above; c£ also Hoeksema 1 989) to be sufficiently compelling to ignore the problem raised by (Joe). Jacobs's position is clearly untenable, if only because it cannot explain the marginal possibility of focus adverbs inside phrases in at least some cases in English, such as the occurrence of even in (3 1 ), taken from C. S. Lewis's The Allegory oJLove (p. 1 7). (3 1) . . . the ardent lover even of his own wife is in mortal sin
English sentence such as (J2a) below is grammatical, while its German counter part (32b) is not. (32) a. John will talk to only Mary. b. *John mochte mit nur Mary reden.
John would with only Mary talk Bayer proposes to link this clear distinction between English and German prepositional phrases to another familiar difference between these two languages. English prepositions can be stranded, whereas German prepositions cannot. The deeper explanation for this difference lies, according to Bayer, in the fact that English prepositions govern in the same direction as English verbs, whereas German prepositions and German verbs are disharmonious in this respect. According to ideas advanced by Kayne (1 984) and Koster (1 987), this makes it possible to extract out of English PPs, while German PPs are islands. The details of the Kayne/Koster account of preposition stranding do not concern us here. What is important for our purposes is the fact that this theory predicts the difference in acceptability between (32a) and (32b), provided we assume that expressions with focus adverbs are quantifiers, which have to undergo Quantifier Raising at Logical Form. If they have to undergo move ment, and if movement out of prepositional phrases is possible in English but . not in German. then we predict that the English sentence is grammatical, while the German one is not. However, before we put too much stock in this explana tion, recall that Roath had observed that
in English, too, occurrences of focus
adverbs between a preposition and its object are often bad. In this respect we find a rather marked difference with preposition stranding under WH movement. The latter is entirely natural in English, quite unlike strings such as
plead with even the Fiihrer.6 Perhaps even more problematic is the fact that regular quantifiers following prepositions are fine in German. in spite of the fact that they, too, would be expected to move out of the PP as the result of Quantifier
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
A rather different perspective on the matter is offered in a series ofpapers by Josef Bayer, e.g. Bayer ( 1 988a,b, 1 989). Bayer wants to know why it is that an
Jack Hoeksema and Frans Zwarts 61
Raising. After all, they engage in scope ambiguities just like NPs adorned with focus adverbs. Rooth {1985) already mentioned this as a problem for a move ment account, and in our view it is still a valid objection to the proposals of Bayer. Finally, it seems to us that Bayer's theory makes wrong predictions in a number of cases for reasons which we do not quite understand ourselves. Thus, for instance, while focus adverbs directly following prepositions are as bad in Dutch as they are in German, there are isolated cases where such sequences are possible. An illustration is given in (3 3). {33) Met aileen goede bedoelingen kom je er niet.7 With onlygood intentions come you there not 'You won't get there with good intentions only.'
(34) The closing off of East Berlin without interference from the West and with the use only of East German, as distinct from Russian, troops was a major Communist victory, which dealt West Berlin a deadly, possibly a fatal, blow. The position in which we find only here is not one from which extraction may proceed; witness the following example. (3 s ) •of which troops did they close off East Berlin with the use? Finally we note that a movement account of focus adverbs predicts all kinds of island effects which we do not find in the data. Consider e.g. the example in (36), which ought to violate the Conjoined Structure Constraint ofRoss ( 1967), since extraction of the second conjunct could never be across-the..:board. (36) He deals with thieves and even with killers. The sentence in question is no doubt correct. The prediction, therefore, is not. What conclusions can we then draw concerning the placement of focus adverbs? The two major theoretical accounts we are aware of, those ofJacobs (1983, 1986) and Bayer (1988a,b, 1989), are not satisfactory. Besides the empirical shortcomings reviewed above, they fail to bring out individual differences between focus particles, such as observable distinctions between sogar and nur, or to address the rather different distribution of negative-polarity focus adverbs, such as auch nur in German and ook maar in Dutch (c£ note 6). In this respect, a recent proposal by K.rifka (1989a,b) seems to us more promising. He suggests that negative-polarity items such as auch nur irgendein jungein German or ook maar iinjongen in Dutch are similar to expressions in focus in so far as they come with a set of alternatives. In fact, this explains why such constituents are typically accented in much the same way that focus expressions
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
An English example which does not fit Bayer's theory is given in (34). It is taken from the Brown corpus of American English and clearly bespeaks its age.
62 Some remarks on focus adverbs
4·
C O N STRA I NT S O N THE ARGUMENTS O F PHRASAL F O C U S ADVERB S
It has frequently been observed that focus adverbs do not attach themselves to all phrases. Expressions such as only someone and only everyone, for instance, do normally not give rise to an acceptable sentence. {37) a. *Only someone objected to the proposal. b. *Only everyone was present at the meeting. On the other hand, if we add a restricting phrase to someone and everyone, as in {38), or replace both by the composite expressions some woman and every woman, then the resulting sentence is perfectly acceptable. {38) a. b. (39) a. b.
Only someone from Middletown objected to the proposal. Only everyone from Middletown was present at the meeting. Only some woman objected to the proposal. Only every woman was present at the meeting.
These patterns are by no means restricted to the adverb only. If we select even instead, we find that the judgements are essentially the same. The following two sentences, for example, are both unacceptable.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
are. As Krifka notes, however, there exist two important differences between negative-polarity items and consituents in focus. Whereas the alternative set of focused expressions is typically given by the context, the alternative set of negative polarity items is typically given by the lexical knowledge of the language user. For instance, the set of alternatives for the polarity expression eine miide Mark ('a red dime') consists of amounts of money, and the set of alternatives for einen Fingerriihen ('lift a finger') of acts of labour. Secondly, where� the alternative set of focus constituents may be unordered, the alterna tive set of negative polarity items is ordered. Thus, the alternatives of eine miide Mark are ordered by their monetary value, and the alternatives of einen Finger rnhren by the physical effort they involve. The negative polarity item itself represents of course the bottom element in this order. Obviously, as Krifka points out, these ordered alternative sets are closely related to Fauconnier's notion of a scale. Contrary to what Fauconnier suggests, however, we need not assume that the alternatives are ordered linearly; it is sufficient to require that they exhibit a pre-order. Insightful as Krifka's remarks may be, we feel that our observations have just scratched the surface of the complex distribution of focus adverbs. Clearly, there is much more work to be done.
Jack Hoeksema and Frans Zwarts 63
{40) a. 0Even someone disagreed with the speaker. b. 0Even everyone appeared at the convention. However, when we add the restricting prepositional phrase from Middletown to someone and everyone or replace both by the composite noun pharses some senator and every senator, the result is again perfectly good. {4 1 ) a. b. (42) a. b.
Even someone from Middletown disagreed with the speaker. Even everyone from Middletown appeared at the convention. Even some senator disagreed with the speaker. Even every senator appeared at the convention.
(43) a. De slager heeft ook iemand gehoord. The butcher has too someone heard 'The butcher heard someone too.' b. De minister heeft ook iedereen voorgesteld. The secretary has too everyone introduced 'The secretary introduced everyone too.' Superficially, the expressions iemand and iedereen seem to act as the arguments of the focus adverb ook. Yet when we take the interpretation into account, it becomes immediately clear that the associated focus constituents of ook are not iemand or iedereen , but rather de slager and de minister. This is corroborated by the fact that the occurrence of ook iemand or ook iedereen in subject position invariably produces an ungrammatical sentence. (44)
00ok iemand heeft de slager gehoord.8 Too someone has the butcher heard 'Someone too heard the butcher.' b. 00ok iedereen heeft de minister voorgesteld. Too everyone has the secretary introduced 'Everyone too introduced the secretary.' a.
On the other hand, if we add a restricting phrase to iemand and iedereen , as in (40), then the resulting sentences are acceptable. {45) a. Ook iemand uit Groningen heeft de slager gehoord. 'Someone from Groningen too heard the butcher.' b. Ook iedereen uit Groningen heeft de minister voorgesteld. 'Everyone from Groningen too introduced the secretary.' What these examples suggest is that a focus adverb only leads to an accept able result if the semantic interpretation of the constituent in focus is such that
c
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Indeed, the Dutch counterpart of the focus adverb too displays a similar behaviour, even though the next two sentences may seem to point in the oppo site direction.
64 Some remarks on focus adverbs
it allo�s for a set of proper alternatives. Thus, for instance, only someone and only everyone are ill formed because someone and everyone cannot be contrasted with quantifiers of a similar type and therefore have no proper alternative. But when we take an expression such as someonefrom Middletown, it is immediately clear that the range of alternative quantifiers includes someonefrom New York, someone from Berlin, someonefrom Amsterdam , etc. For that reason, the phrase only someone from Middletown produces an acceptable result. It seems to us that an account along these lines may ultimately also explain why focus adverbs do not in general attach to monotone decreasing noun phrases, such as nobody or nothing.
C O N STRAINTS O N C O M B I NATIONS O F F O C U S ADVE RB S
One of the least-studied aspects of focus adverbs is their compatibility with other focus adverbs. Yet it is clear that there are important questions here waiting to be investigated. One may observe, for instance, that focus adverbs, just like modal adverbs, can be stacked up and sometimes appear to form idiomatic clusters. On the other hand, one finds that certain combinations of focus adverbs are ungrammatical or semantically anomalous. An example of an idiomatic cluster is German auch nur (lit. 'also even' - 'even, so much as'; c£ also note 6), which consists of two focus adverbs that would be incompatible in a non-idiomatic interpretation, since auch presupposes that there are alternatives for the element in focus which could be substituted for its salva veritate, whereas nur asserts that there are no such alternatives. Thus the unacceptability of the combinations in (46) (where both also and only are interpreted as applying to the direct object) is not at all surprising. (46)
a. #Margaret also denounced only CATHOLICISM. b. #Margaret also only denounced CATHOLICISM. c. #Margaret denounced also only CATHOLICISM.
It is therefore interesting from a historical-semantic point of view how the German cluster auch nur and its Dutch counterpan ook maar received their idiomatic properties and negative-polarity behaviour. Of some interest also is the difference: between English on the one hand and Dutch and German on the other via a vis the possibility of stacking up negation with the word for 'even'. Compare the examples in (47). (47)
Not even MOSES was without weak moments. b. �icht sogar MOSES war ohne schwache Momente. (German) c. 0Niet zelfs MOZES was zonder zwakke momenten. (Dutch) a.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
S·
Jack Hoeksema and Frans Zwai-ts 65
In German and Dutch, focus adverb sequences corresponding to not even are not acceptable. Instead, one uses idiomatic strings such as German nicht einmal {lit. 'not once') and Dutch niet eens (id.), or a combination of'even' on the subject with 'not' on the predicate, as in (48 a,b). The latter possibility also exists in English (48c). (48) a. Sogar Moses war niche ohne schwache Momence. b. Zelfs Mozes was niet zonder zwakke momenten. c. Even Moses was not without weak moments.
(49) a. (Not NP) VP b. NP (not VP Given that even Moses likewise denotes an ultrafllter, we expect [not [even tn be ill formed in subject position. Hence the sequence not even is presumably a complex focus adverb. Additional evidence for this analysis comes from the possibility, in Dutch (but not in German), of the sequence corresponding to even notin adphrasal position:
Moses]]
( so)
a. Zelfs neit de slager was op de hoogte.
Even not the butcher was in the know 'Not even the butcher knew about it.'
Here a complex adverb analysis is all bur unavoidable, since niet de slager is not itself a possible NP. The fact that English has fixed the order negation < even, whereas Dutch has fixed the opposite order, may be due to the existence in English, but not in Dutch, of a negative-polarity variant of even which syntactically takes narrow scope with regard to negation and semantically takes wide scope (c£ Rooth 1 9 8 5 for a motivation of this claim). It is attractive to make a leap now and to propose the following generaliza tion: (51) Multiple Adverb Constraint
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
It is clear that in (47a) even has scope over not, in spite of the fact that normally the leftmost operator has scope over any operators to its right in English. It can be argued that the proper analysis of not even Moses is [[not even] Moses] rather than [not [even Moses]]. The reason is that not does not attach to subject-NPs which denote ultrafllters, i.e. the set of properties generated by a single individual (c£ Barwise & Cooper 1 98 1). Thus, for instance, not John left is ill formed. Barwise & Cooper note that the ultrafilters are the self-dual quantifiers, that is, they are the class ofNPs for which sentences of the form in (49a) are equivalent with sentences of the form in (49b). If we make the plausible assumption that predicate negation is preferred over phrasal negation, provided that the meaning is not affected, we can explain why it is that the class of ultrafilters resists phrasal negation.
66 Some remarks on focus adverbs
If more than one focus adverb is associated with a focus expression, then we are dealing with a complex focus adverb, not a free syntactic combina tion. If we take a variable-binding approach to focus adverbs, then the constraint in (51) could be reduced to a, more general constraint against multiple binding of a single variable. In the framework ofRooth ( 1 98 5 ), this constraint does not seem to follow immediately. It may be too early to tell if the Multiple Adverb Constraint is correct, but it seems worthwhile to pursue it further. In addition, we emphasize the need for detailed studies of focus adverb clusters. What are the mechanisms involved in producing such clusters? What kinds of meaning do they develop?
In the preceding pages, we have reviewed a number of issues in the study of focus adverbs. We have found that there is considerable variation among focus adverbs along several dimensions, including the syntactic relation between the focus adverb and its associated focus, the placement of focus adverbs in various positions, and the interaction between several focus adverbs. The literature on focus adverbs contains a number ofvaluable theoretical approaches, but at this point they must be considered as no more than avenues along which we may slowly and rather cautiously proceed towards, one hopes, that ultimate destiny, the truth. We have found Roath's (1985) semantics for focus adverbs to be particularly useful. Less satisfactory, in our view, are current approaches to the syntax of focus adverbs, which fail to address the many individual differences among focus adverbs in sufficient detail and do not cover the cross-linguistic variation that must be accounted for. Finally, we want to point out the need for an integrated theory of focus which takes into account its prosodic, syntactic, semantic and pragmatic aspects and relates focus adverbs with properties of other focus constructions, such as Gapping, Stripping and exception-phrase constructions. Acknowledgements. This
research was supported by the Netherlands Organization for Research, NWO (Grant B 3o-3 1 1). We would like to thank Manfred Krifka for valuable comments on an earlier version of this paper. JACK HOEKSEMA and FRANS ZWARTS Centre for Behavioural, Cognitive and Neurosciences and Neuroscienrific Studies University of Groningen 9712 EK Groningen Netherlands
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
6 . S U M MARY AND C O N C L U S I O N S
Jack Hoeksema and Frans Zwarts
67
N O TES 1
As a matter of fact, this view, which is
clearly
the
'received
view' on
focus
adverbs, has been challenged by Vallduvi
such as (i), which appear to be acceptable.
(1 989).
(i) John and Mary know the Amazon quite
For example, he notes that a focus
adverb, together with its associated con
well, but only John has been to the
stituent, may be deleted in the gapping
CITIES in Brazil.
construction. (i) Mary (only scratched) the Mercedes, and John t:he Bentley.
We agree that this sentence, with CITIES, not John, as the intonationally most prominent i tem, is fine. However, we take this example to show only that the notion
what difficult to get for us, is what Mary
'intonational prominence' is gradient, but
does to the Mercedes in merely scratching
not that John is not in focus in (i). In fact, it
it and similarly for John and the Bentley.
appears to us that John must be stressed,
Taking the usual point of view that the
even if this stress is weaker than the stress
remnants ofGapping are focus expressions,
in CITIES. In the absence of phonetic
whereas the deleted part belongs to the
measurements, this can be shown perhaps
non-focus or presupposition part, Valluvi
most clearly by considering similar data
suggests that the phrase associated with
from Dutch, in which there is a lexical
only is not itself a focus expression.
distinction between 'weak', or unstressable,
However, it seems possible to us that the
pronouns and 'strong', or stressable, ones.
requirement on Gapping is merely that the
In Dutch examples comparable to (i), in
remnants be focused, but not that the left
which the constituent in construction with
out material be non-focused. In this way
the focus adverb does not bear the main
we might explain the possibility of (i). In
intonational peak, it is nevertheless a fact
addition, we might explore the possibility
that this constituent cannot be a weak
that a gapped stting
pronoun. For example, in (ii) below, the
may contain a
ij 'me' is grammatical, but not
relatively weakly stressed focus expression.
pronoun
Note that if the intonational expression of
its weak variant
sentation such as that of Jacobs (this
m
me.
(ii) Ze toonden Pier en mij de Amazone,
maar aileen mij (•me) toonden ze ook
volume), it is automatically predicted that
de STEDEN.
not all focus expressions will be equally
'They
prominent prosodically. See also note 3 for some related discussion.
By 'quasi-universal', we mean that infer
showed
Pier
and
me
the
Amazon, but only me they also showed the CITIES.' The possibility of more than one focus in a
ences can be drawn about all elements in
sentence, and rhe fact that one focus may
the set of alternatives higher on the scale.
be much more prominent, intonationally,
We disagree with Roorh
(1985),
who
than the other, ofcourse may create some
argues for a weaker characterization of even , according to which even John left
what confusing data, but we would not want to alleviate the confusion by simply
merely implicates that someone other than
denying focus status to expressions that are
John left. Roath's characterization, taken from Karmmen and Peters
(1979),
fails to
provide a basis for distinguishing also from even ,
and does not correspond to our
intuitions.
less than maximally prominent.
4 Thus, sentence (i) is impossible, in spite of
the possibility of wide-scope negation in
the corresponding sentence (ii). (i) •Anybody didn't leave.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The intended reading here, which is some
focus is mediated by a mettical tree repre
2
3 Vallduvi (1989) denies that such sentences are ungrammatical. He points out cases
68 Some remarks on focus adverbs
(ii) Everybody didn't leave.
(i) Das schaffi ein Normalbegabter mit nur Studieren nie. That achieves a normally gifod with only studying never 'A normally gifted person will never achieve this with studying alone.' According to Bayer, the presence of an additional adverb, nie, makes this sentence acceptable. Since both the . quantified expression nur Studieren and the quan tificational adverb nie must find their scope by raiSing at LF, a configuration arises in which the operator-trace path for nie properly contains the path relating nur Studieren with its trace. Given Pesetsky's (1982) Path Containment Condition, an extension and modification of Kayne's ( 1 984) Connectedness Condition, this containment is what makes the otherwise impossible extraction acceptable. In our opinion, this explanation is convoluted, and it seems to us that equally acceptable sentences can be given without the presence of an additional adverb. For instance, we find the following sentence acceptable: (ii} Wurde ein Normalbegabter so erwas mit nur Studieren schaffen konnen? 'Would a normally gifted person be able to achieve such a thing with studying alone.' However, when we replace nur by sogar, 'even', we get a much less acceptable sentence: (iii) •wurde ein Normalbegabter so etwas mit sogar Studieren schaffen konnen? Here, the only acceptable order is the one in which the focus adverb precedes the preposition: (iv) Wurde ein Normalbegabter so erwas sogar mit Srudieren schaffen konnen? 8 We are here considering the case where ietrUJnd has regular focus intonation. There is another possibility, which is still mar ginal but better, where not ietrUJnd but the focus adverb ook itself is the intonationally more prominent item. This phenomenon deserves further study.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(i) also shows that negative-polarity any patterns with the existential quantifiers and not with the universal quantifiers. 5 If it is assumed that infinitival comple ments are not clauses but VPs, then the restriction on scope has to be modified somewhat. In any case, it should be noted that focus adverbs inside an infinitival complement do not take scope over elements outside it, e.g. example (i) is not interpreted as saying that even Plato ttied to argue this point. (i) PLATO ttied to even argue this point. 6 While focus adverbs between prepositions and their objects are sometimes possible but usually disfavoured in English, we note that they are also sometimes possible in German, though perhaps more marginally than in English. Bayer mentions a number of cases, and we mention here the case of the polarity item auch appearing inside PPs, c£: (i) Ich glaube nicht, daB er mit auch nur einer Person geredet hat. One might object that auch nur is a scalar adverb (c£ Kri£ka 1989a,b for some discussion of this point and its relevance for a general understanding of a sizeable class of polarity items), and that scalar adverbs in general may freely occur inside PPs, presumably because they are not quantificational in nature. This is ttue enough. However, we point out that auch nur phrases may take wide scope over other quantifiers and so have to undergo the rule of Quantifier Raising at LF according to the assumptions of the frame word adopted by Bayer. An example illustrating wide scope is (ii). (ii) Niemand kennt aile Freundinnen auch nur eines dieser Herren. 'Nobody knows all girlfriends of even one of these gendemen.' Here the phrase auch nur eines dieser Herren has scope over aile Freudinnen. 7 Bayer (1988a) cites a similar example from German, attributed to Ad Foolen:
Jack Hoeksema and Frans Zwarts 69
REFERE N C E S Jacobs, Joachim (1983), Fokus und SkGlen: Zur Syntax und Semantik von Gradpartikeln im Dtutschm, Niemeyer, Tiibingen. Jacobs, Joachim (1986), 'The syntax of focus adverbials in German', in W. Abraham & S. de Mey (eds), Topic, Focus and Configura tionality, Benjarnins, Amsterdam/Phila delphia, IOJ-27. Jacobs, Joachim (1990), 'Focus ambiguities', this volume. Karrtunen, Lauri & Stanley Peters (1979), 'Conventional implicature', in C.-K. Oh & D. Dinneen (eds), Syntax and Semantics 1 1 , Academic Press, New York. Kayne, Richard ( I984), Connectedness and Binary Branching, Foris, Dordrecht. Koster, Jan (I 987), Domains and Dynasties: The Radical Autonomy of Syntax, Foris, Dor drecht. Krifka, Manfred (1 989a), 'Some remarks on polarity items', in D. Zaefferer (ed.), Semantic Theory and Linguistic Universals, Foris, Dordrecht. Krifka, Manfred (1 989b), 'Polarity pheno mena and alternative semantics', to appear in the Proceedings of the Seventh Amster dam Colloquium. Kroch, Anthony ( I 974), The Semantics ofScope in English , doctoral dissertation, published I975 by the Indiana University Linguistics Club, Bloomington, and I979 by Garland Press, New York. Ladusaw, William (I 980), Polarity Sensitivity as Inherent Scope Relations, doctoral disserta tion, University of Texas at Austin, published by the Indiana University Lin guistics Club, Bloomington. Pesetsky, David (I982), 'Paths and categories', doctoral dissertation, MIT Plank, Frans (1 979� 'Zur Affiniriit von Selbst und Auch', in H. Weydt (ed.), Die Partikeln der deutschm Sprache, De Gruyter, Berlin. Pullum. G. K. (I988), 'Topic . . . comments: some lists of things about books', Natural Language and Linguistic 'Thftlry, 6: 28 3-90· Rooth, Mats (1985), 'Association with focus', .
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(1976� Dit Gradpartikeln im Dtutschm: Untersuchungen zu ihrer Syntax, Stmantik und Pragmatik, Niemeyer, Tiibingen. Altmann, Hans (1978), Gradpartikel-Probleme, Gunter Narr, Tiibingen. Anderson, Stephen (1972), 'How to get even ', Language, 48: 893-906. Barwise,John & Robin Cooper (I981), 'Gen eralized quantifiers in natural language', Linguistics and Philosophy, 4: 1 59-219. Bayer, Josef (1988a), 'Interpretive islands: evidence for connectedness and global harmony on logical form', in G. Grewen dorf & W. Stemefeld (eds), Scrambling and Barriers, Benjamins, Amsterdam. Bayer, Josef (1988b), 'Rightward movement and the syntax of quantificational particles in German', unpublished paper, Max Planck lnstitut, Nijmegen. Bayer, Josef (1989), 'Fokussintensive Parti keln: Eine Kontrastive Umersuchung', talk given at the University of Constance, 2o April 1989. Fauconnier, Gilles (1975), 'Pragmatic scales and logical structure', Linguistic Inquiry, 6: 353-75· Fillmore, Charles, Paul Kay & Mary Cather ine O'Connor ( 1 988), 'Regularit}r and idio maricity in grammatical constructions: the case of let alone . . .', Language, 64: 501-38. Grosz, Barbara (1977), 'The representation and use of focus in dialog understanding', Technical Note 15, SRI International, Menlo Park. Hoeksema, Jack (1989), 'Only in Dutch: a comparison of three adverbs', in M Jack, L Koening & A. Taylor (eds), Penn Review oJLinguistics, No. 1 3: 1 o6-22. Hoekstra, Eric, Helen de Hoop & Frans Zwarts (1988� 'Lineaire Restrikties op Polaire Uitdrukkingen?', TABU: Bulletin voor Taalwetenschap, 18: 226-36. Jackendof:t Ray (1972� Semantic Interpretation in Generative Grammar, MIT Press, Cam bridge, Mass. Altmann, Hans
70 Some remarks on focus adverbs doctoral dissertation, University of Massa chusetts, Amherst. Ross, John Robert (1967� 'Constraints on variables in syntax', doctoral dissertation, MIT.
Sidner, Candance (1979), 'Towards a com-
putational theory of definite anaphora comprehension in English discourse', doctoral dissertation, MIT. Vallduvi, Enric (1989), 'Only and focus', paper given at the 1989 Annual LSA meeting.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
joumtJI of&mantics 8: 71-<)0
© NJ.S. Foundation (1991)
Focus: An Idea in Motion ULRI CH F. G. KLE I N University of Cologne Absuact This paper discusses stress patterns concerning the relation between intonation and the mean ing of an utterance. It is argued that this should be done within a linguistic approach. While the first pan of this paper deals with differences in meaning that are caused by differences in and Jacobs. Within this discussion the interrelation is shown between the syntactic feature [+F]
(+ focus) and the phonetic and semantic properties of an utterance.
1
I NTRO D U C T I O N
We could describe prosody as a vocal observational-introspective fact of speech originating in the mind-body that is biologically-culturally acquired, cognitively-emotionally motivated, and propositionally-nonpropositionally meaningful in a symbolic-gestural fashion.
The ironical definition by ArndtiJ"anney (1988: 83) makes a caricature of the risks of a strictly phenomenon-orientated work in linguistics. This especially applies to highly sensitive areas such as stress and intonation. Even if only viewed under communicative aspects, they can be an expression of, for good reasons, well-differentiated systems of knowledge. The following functions of prosody shall be stated here with examples usually found in the literature. In tone languages prosody continually serves as a differentiation of lexemes (in Chinese [wa] depending on the intonation means mother, hemp , horse or to scold). Intonation distinguishes modes of sentences. The German sentence Stellt ihr euch an can be interpreted as an interrogative sentence (Are you queueing up?) as well as an exclamation ( What afuss you are making!) or as an imperative (Queue up! Get in line!), depending on the different Fo-contour. Furthermore, the functions of prosody prevent erroneous syntactic analyses (Hans spielt mit Clara und Paul mit dem Ball bleibt zuriick-Hansplays with Clara and Paul with the ball stays behind ); they disambiguate meanings (Sie will ihn kurz halten vs. Sie will ihn kurz halten-She wants to hold him ( in her arms) for a short while vs. She wants to keep him on a tight rein.) They can also 'serve to direct the discourse, indicate a special emphasis of the speaker, be used for rhetorical and ritual means'; Wunderlich (1988: 3) etc. Although Paul ( 1 88o) had already pointed out in different words that sentences with sentence stress in the same position can have a different
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
intonation, the second pan gives an overview over the focus theories of Hohle, Rochemont,
72
Focus: an idea in motion
structure of information, thereby having realized the connection between pro sodic and semantic facts, Klein (I 98o: 3) could still justly say:
that most of the descriptions [of the German sentence intonation), for example those in the accepted grammars are based on von Essen (1956) only in so far as they are not completely impressionistic anyway. This alone is worth mentioning because this booklet mainly served didactic purposes. By no means was it intended to be a systematic scientific study of fundamentals.
analysed by a linguistic approach. I will try to prove this statement in the first part of my paper and show how one can determine, by means of a test, the various meanings of sentences with otherwise identical lexemes and identical word order, which are produced by a different prosody. Not because of a historical interest will the second part show the development from the Focus Theory by Hohle (I 982) to Rochemont (I 986) and to the Relational Focus Theory byJacobs. This argumentation rather serves to prove my main thesis (in accordance with Rochemont and Jacobs) that a syntactic feature [+F) (= +focus) is spelled out phonologically and can be inter preted semantically.
2 C O N CE R N I N G P R O S O D Y The prevalent lack of interest about the prosodic phenomena which, at this point, I want to narrow down to the aspect relevant here, namely stress, is, among others, certainly due to the following reasons: in no other area of performance are fewer mistakes made (or noticed) ; hardly any other area provides such heterogeneous intuitions o f the speaker;
no other phenomenon is presented so sparsely in written language.
However, the minimal pairs in (I }-(4) show that different word stresses are sufficient for different interpretations of otherwise identical sequences of phonemes (capital letters characterize the stressed syllable):•
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
To agree with Lieb's (I 98o: 3 4) view-' Whoever raises his voice in anger does rhis neither in English nor in German, but just in anger'-would be correct in so far as the treatment ofprosody would not be a linguistic task but rather a task of psychology, sociology, etc. However, in my opinion, which is not only due to the abundance of linguistic literature dealing for the last decade with the interplay of intonation and meaning, one cannot agree with Lieb. More recent studies, especially in connection with Jacobs's Relational Focus Theory, indicate that the patterns of stress in sentences are, among other constituents, also relevant for their meaning in a more narrow sense. Therefore they have to be examined and
Ulrich F. G. Klein 73
(r )
AuGUST vs. AUgust
August vs. german proper name
(2) umFAHren vs. UMfahren
to drive round vs. to run down (3) unterSTELLen vs. UNterstellen to impute vs. to place under (4) missFALLEN vs. MISSfallen2
to displease vs. displeasure
(s)
a. Hilde kennt nur EINEN Roman von Thomas Mann.
Hilde.knows only ONE novel by Thomas Mann.
b. Hilde kennt nur einen ROMAN von Thomas Mann.
Hilde knows only one NOVEL by Thomas Mann . (6) a. Der Abteilungsleiter wurde enclassen, weil er die KOLLEGIN nach DienstschluB gekiiBt hatte.
The head of the department was discharged because he had kissed his COLLEAGUE after office hours. b. Der Abteilungsleiter wurde enclassen, weil er die Kollegin nach DIENSTSCHLUSS gekiiBt hatte.
The head of the department was discharged because he had kissed his colleague after OFFICE HOURS. Jacobs (1988b: 4), where (s) was taken from, points out that (sa) 'describes circumstances which allow that Hilde knows all other narratives by Thomas Mann, whereas the meaning of (sb) excludes just this'. In other words, the different propositions of the sentence pairs are subject to different conditions of truth. 2.1
A test to determine the literal meaning
This can be tested with an expanded application of the method suggested by Reis (r98o) to determine the literal meaning.3 The method differentiates between the implicated (B;) and the literal (Bj) meaning of an expression. Reis (1980: 2) demands that the meaning Bi contained in the expression S in context K must be seen as part of the literal meaning of S, provided that: 1.) Bi is not variable (that is Bi appears in all contexts of speech of S); 2..) B cannot be eliminated (that is the connection of S with the negationlnonretention of Bi i leads to a contradiction);
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The reason why (r)-(4) are different words is that the different word stresses create different meanings. If different stresses of otherwise identical words create different meanings and therefore different words, this should also he true of sentences. Thus each sentence of the pairs (s) and (6) is a different sentence, although each consists of the same lexemes in identical order.
74 Focus: an idea in motion
J.) The appearance of Bi cannot be reconstructed
as
a combination of the original literal
meaning of S, principles of conversation, and circumstances of the contexts of speech K.
Let us now apply the test to example (sb): (s)
b.
Hilde kennt nur einen ROMAN von Thomas Mann. Hilde knows only one NOVEL by Thomas Mann.
(According to the intuition the stress should be rising-falling here.) The context K1 is given:
K1 : Two literary scholars (A and B) discuss if it would be sensible to let their colleague Hilde participate in a new Thomas Mann edition. A utters (5 b).
B 1 : Hilde knows nothing about the study of literature. B2: Hilde's participation does not further our proj ect. B3 : Hilde knows one novel by Thomas Mann, but apart from this she has no knowledge about his other work. One can easily imagine contexts in which B 1 and B2 are irrelevant, therefore violating condition 1 ofReis's method. Here (sb) could be the answer of Hilde's brother to his mother's question, if it would be sensible to give Hilde, a scholar of German, a Thomas Mann edition as a Christmas present. According to condition 3 , B 1 and B2 are excluded as candidates for Bj in contexts in which they can be reconstructed pragmatically. This is the case for example, if A had previously made derogatory remarks about Hilde's ability (B1), or if he takes only accepted Mann critics into consideration for project K1 and B knows this (B2). The sentence (sb ') *Hilde kennt nur einen ROMAN von Thomas Mann, was aber nicht heiBt, daB sie nur einen ROMAN von Thomas Mann kennt. *Hilde knows only one NOVEL by Thomas Mann, which does not mean that she only knows one NOVEL by Thomas Mann. is definitely unacceptable because of condition 2. If one now also takes into account the prosodic circumstances of (sb), one reaches the conclusion which has already been anticipated by the second part of B3 : I assume that stressing novel (apparently in connection with the particle only) in (sb) leads to the conjunction of two propositions (P1 and P2):
P 1: Hilde knows one novel by Thomas Mann. P2 : Hilde knows nothing else by Thomas Mann apart from one novel.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
In K1 (sb) can now have, amongst others, the following (and many more) meanings.
Ulrich F. G. Klein
75
A conjunction already takes the truth conditionfalse if only one of such state ments becomes false. Consequently (sb "), in which the second part of the sentence (sb ") contradicts P2, is unacceptable: (sb ") *Hilde kennt nur einen ROMAN von Thomas Mann und alle seine Erzahlungen. *Hilde knows only one NOVEL by Thomas Mann and all his narratives. The possible objection that the conjunction of the propositions solely rests on the particle seems not convincing to me if one looks at the sentence (sb ) which is at least highly questionable: "'
(sb ' ) 11Hilde kennt einen ROMAN von Thomas Mann und alle seme Erzihlungen. ??Hilde knows one NOVEL by Thomas Mann and all his narratives. "
"'
(s)
a.
Hilde kennt nur EINEN Roman von Thomas Mann. Hilde knows only ONE novel by Thomas Mann.
means B4: Hilde knows one (and only one) novel by Thomas Mann. This sentence does not rule out that, with otherwise identical lexemes (includ ing only !) and identical word order, Hilde knows other work by Mann, as the completely correct sentence (sa ' )-used as a test of negation for (sa)-proves: (sa ') Hilde kennt nur EINEN Roman von Thomas Mann, aber aile seine Erzihlungen. Hilde knows only ONE novel by Thomas Mann but all his narratives. It must be stated that the sentence stress is, among other aspects, relevant for the literal meaning of sentences and that different stresses can therefore lead to different meanings of sentences. If this is indeed so, then other explanations must be ruled out a priori which, simply put, state that one stresses only the contents one wants to emphasize (and only these). The certainty with which persons unambiguously interpret stresses shows that emotions, knowledge of the world, etc. do not create the relevant paradigms for the phenomena of stress. But on the contrary, the analyses of sentences such as (s) suggest that stressing obeys strict rules, as is usually assumed in connection with syntactic phenomena.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
If the reader thinks that this example is grammatical, the reason is that he has changed the pattern of intonation.4 (s b ) only becomes acceptable if one uses the intonation typical for enumerations. This would, however, contradict the above-mentioned restriction that the stress is rising-falling. But it is even more convincing that (sa), which also contains the particle only, does not at all implicate P2 (sb), as the analogous analysis of (sa) shows:
76 Focus: an idea in motion
3
S TRES S AND THE S T RUC TU RE O F IN FO R MA TION
As shown above, the changing of certain prosodic characteristics of a sentence S
makes it necessary to treat the so changed sentence S ' as an independent one. If one accepts, for example, according toBierwisch (198o: s), that any grammar G and its language L are subjects to the following condition For any Is there is nothing in sem that is not related by syn to some part of conftguration in pt ',5
3.1
Nonnal stress and unmarkedness
Apparently _ speakers/listeners have definite intuitions of how sentences are usually stressed, that is to say they know whether or not a sentence is stressed normally? Evidently this intuition of German8 speakers/listeners is described by Heidolph, Flamig & Motsch (1981 : 868) as follows: Here a well-known tendency is expressed, namely that-apart from exceptions which can be more or less defined-the unmarked emphasis appears either at the end of a sentence or directly before the end of a sentence. That is, it appears there in a part of a sentence which is lexically occupied.
But there are many cases which contradict the tendency or regularity of stressing the right-hand side. The examples under (6) have to be stressed at the position outlined in order to be unmarked. (When a shift of the stress occurs to a position further to the right of the given place, the examples become slightly more marked-)9 (6)
a.
b. c.
d.
Ein BRIEF kommt an. A LETTER arrives. Nicht PETER hat sein Lineal vergessen. It was not PETER who hasforgotten his ruler. Ich suche den KLOGS TEN unter den Studenten. I am lookingfor the BRIGHTEST ofthe students. Ich SAGE dir was. I TELL you something.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
that all parts ofsem are related through syn to pt, then it must be possible to explain the phenomena of stress through syntax. Otherwise one would violate the above-stated condition because stress is, among other aspects, relevant for the meaning of sentences.6 Consequently the structure of information of two sentences S and S ' is also different. It will now be shown how stressing changes the structure of information in sentences.
Ulrich F. G. Klein
77
On the other hand, sentences with stresses at the end sometimes appear to be marked in some sense; respectively they can only be accepted within limits (cf. (7a)). And there are sentences which contain several stresses of equal strength (whereby in these sentences no part shall be stressed more strongly than that part marked by capital letters); cf. (7b) and (7c) : 1 0 (7)
Ich schickte an einen Bewerber den FRAGEBOGEN. I sent the QUESTIONNAIRE to an applicant. b. KLAUS wohnt in MONCHEN und ICH in BERLIN. KLAUS lives in MUNICH and I in BERLIN. c. Ich kiiBte erst RITA, dann ihre FREUNDIN auf die Wange. First I kissed RITA, then her FRIEND on the cheek. a.
=
,
'
3 .2
Focus
Despite some reservations, Hohle uses the orthodox question test going back to Paul (188o) to determine the focus of a sentence. The respective focus of (8),
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
These examples clearly show that the assumed tendency to stress the right hand side as indication of the normal stress in German is highly deficient as a descriptioiL Hohle (1982} tries to determine the definition of normal stress with the definition of unmarkedness. But the description of normal stress as unmarked stress is only then helpful (and not circular), if one can state conditions when the sentence is marked or unmarked. Hohle (1982: chapter 1) critically names different explanations: firstly, those which are based on the normal stress which is not explained further (for example, Lenerz 1 977).1 1 Secondly, he names those which consider sentences as stressed normally, if no limitations arise for the previous or following sentences because of the stress. Thirdly, he names those which regard only these sentences as stressed normally which are possible as a beginning of a text or discourse; fourthly, those which consider sentences as unmarked when no part of the sentence is emphasized by stress. Hohle (1 982: 8 5) demonstrates the problems in these explanations and he then tries to offer an axiomatic explication: 'A sentence is stressed normally, if the speakers consider the stressing as being stylistically normal; it is not stressed normally if the speakers think of the stressing as being stylistically unusual.' A sentence is called stylistically normal if it is 'relatively unmarked in context with regard to stressing' (Hohle (1982: 103) ). This means that it (contrary to a differently stressed sentence) 'can appear in most types of context' (Hohle (19!!2: 102)). The number of contexts depends on the focus of the sentence ( FK (Si) ) which is defmed as a distinct part of the sentence Si. 'The function of this part in Si cannot be determined by the relevant context. (The other parts of si form the topic Tk (Si) ) (Hohle ( I 982: 86) ). 1 2
78
Focus: an idea in motion
suggested by the various questions, can be realized by the speaker through different stressing:1 3 (8)
a.
Karl fahrt: morgen nach BERLIN.
Karl will travel to BERLIN tomorrow.
b. Karl fahrt MORGEN nach Berlin.
TOMORROW Karl will travel to Berlin. c.
Karl FAHRT morgen nach Berlin.
Karl will TRAVEL to Berlin tomorrow. d. KARL fahrt morgen nach Berlin. Tomorrow KARL will travel to Berlin.
(9)
a. Wohin reist Karl morgen?
Where will Karl travel to tomorrow? b. Was macht Karl morgen?
What will Karl do tomorrow? c.
Was macht Karl?
What does Karl do? d. Was wird geschehen? What will happen? e. Was ist geschehen?
What has happened?
f. Warum bist du so traurig?
Why are you so sad? The contexts in (9) direct the attention to another expression relevant here, namely the focus projection , meaning that an emphasis made manifest by stress need not remain restricted to the stressed constituent (Fk1 (Si) - Fk, (Si) ) : (8a) can answer all the questions in (9) but (8b), for example, cannot (certainly not (9a+b) ). (8b) is highly questionable as an adequate answer to (9c-f). That is so even if fictitious contexts can be constructed in which this perhaps seems possible if a great number of prerequisites of the situation are given. Whereas (Bird) give examples for different foci, (Sa) is a candidate for focus projection because the stress on Berlin permits one to regard even more than the preposi tional phrase including the word as new information (as (9) shows). (9a) also is an example for maximal focus according to HoWe (1982: 99): 'Of all possible foci of a sentence Si those foci are maximal Fk.r, (SJ which contain the most constituents.' When the stressing of an element allows for focus projection, HoWe (1982: 99) calls this element focus exponent: 'Within a complex constituent K;, the constituent Kj is the focus exponent of K; for which is valid: If Kj is a possible
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
On closer inspection of (Sa) one can see that this sentence is not only an adequate answer �o (9a) but also to (9lrd) (and also to (9e+f) ).
Ulrich F. G. Klein
79
But if stressing generally is part of Is in the sense that it enables the speaker to express different meanings and to evoke different meanings in the listener; and if-as we can now say-different foci evoke different meanings of otherwise identical sentences, then it seems strange to take the mechanisms accounting for the normal stressing out of Is. This is strange because the normal stressing is, after all, only one special case of stressing in generai.t6 Hohle (1982: 94) warns not to put too much information on LC : 'It would mean to obliterate an essential distinction, if one wanted to describe those kinds of conditions of usage [meaning conditions of appropriateness of sentences with a different focus structure, UK] in the same way as conditions of truth.' I agree with him, but I think that Hohle puts too much emphasis on PC because it was proved that particularly stress influences the truth conditions of sentences. To summarize: a theory which transfers the syntactic and!or semantic (but in any case basic linguistic) phenomenon focus17 to pragmatics is in my opinion theoretically unacceptable, and will therefore give factually wrong predictions.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
focus of K;, then K; is also a possible focus. This means that there is a focus projection from � (K;) = � to FK, (K;) = MK. (K;).'14 Hohle's approach to this subject deserves attention for various reasons. His approach and his deftnition are more specific than various other theories. Indeed, they are admirably exact and formal. Moreover, their application gives mostly empirically correct predictions. The detailed description of Hohle's explanation is also justified because his treatment of these aspects, which up to now were just treated with common sense, permits a theoretical approach. Moreover, he has brought about an extensive standardization of terminology in the work published after his. Besides another basic point of criticism which I will refer to later on, Hohle's approach has a most striking theoretical deftcit: he (1982: 93) thinks that sentences have, beside certain other characterizations (such as phonological, morphological, syntactical and logical), a pragmatic characterization (PC) which he '[identiftes] with stating the number of the possible foci of Si . . . The PC is not part of the possible logical characterization (LC)'. This does not agree with the results presented in the ftrst section. There the argument was that the variants of stress and the resulting differences in meaning manifest themselves in·that part of the sentence which Bierwisch (1980) calls linguistic structure Is = (pt, syn , sem) . However, if one would agree with HoWe's theory, they would have to be placed in ct ( = context) or in cs ( = communicative sense) of Bierwisch's (1980: 7) description: mu = ((ins, pt, (pt,syn,sem), ct, m), ias, cs) 1 5
So
Focus:
an idea in motion
3·3
FocUs in generative grammar
Rochemont's (1986: r ) main thesis is 'that focus has a uniform grammatical identification only as a syntactic element with, in English at least, a certain systematic phonological interpretation and, presumably universally, a range of semantic interpretations'. He bases his theory on the NSR by Chomsky!Halle (1968 : 6 If) , already characterized as deficient in note 8. This is repeated here with Rochemont's informal description:
·
·
(19) NSR: Assign nuclear stress (= accent) to the rightmost lexical category in S.
(2o)
I didn't say he was getting a new blackBIRD, I said he was getting a new black BOARD. (2 1) Archie would be a bitch of a man to WORK for.'8 (22) A LETTER arrived for you today. {23) a. A: Who followed Ralph into the bedroom? B: LAURIE followed Ralph into the bedroom. b. A: Did Nancy follow Ralph into the bedroom? B: No, LAURIE followed Ralph into the bedroom? (24) A: I finally gave in to my desire to splurge and went out and bought something new today. B: Oh, really? What did you BUY? The relations of stressing and focusing demonstrated by the examples, lead Rochemont to rephrase NSR which he calls NSR" , and which will be explained here: {25) NSR " : Assign an accent to the rightmost lexical category (in a [+focus] con stituent) in 5}9 The NSR" integrates two autonomous components: a phonological component which stresses, and a syntactical component which establishes the feature [+F] [F = focus]. The interpretation of this syntactic feature is the task of a semantic component. Consequendy Rochemont (1986: 36) extends his adaptation of the T-model20 by the following items:
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
With the following examples Rochemont proves that stressing via NSR does not have to be focus projective (2o), that stresses which are not caused by NSR do not necessarily produce minimal focus (2 1) or have to be interpreted contrastively (2 3a), and that stressed elements do not have to be focused (24).
Ulrich F. G. Klein 8 1 (26) 21
1 . a.
D-structure
b. Assignment of prominence to X0 (OPT) c. FA (OPT) d. S-structure
2. a. Accent Placement b. PF
3· a. FR b. LF
4· Focus Interpretation Rules The coordination of these systems of knowledge is responsible for the hetero
geneous picture which (2o)-(24) show. First, the most frequently appearing
(27) John was hit over the head with a black BOARD.
)
Example (2 I shows that in English as well as in German there are elements which cannot be stressed normally. They do not behave according to rules but can be listed and therefore contained in a finite list. can only be predicted if NSR has access to chis list.
(8)
and comparable cases
Examples like (2 3a) cause Rochemont to extend NSR by the condition in a [+focus] constituent. Since the NSR " first tries to realize chis condition, the stress
can be placed in case of success on a different position than at the end of the
sentence (success meaning when an element has the feature [+F] ). Although the
stressing in (23 a) is not focus-projective, it is cenainly the only adequate one,
because a sentence such as
(23aB ' ) Laurie followed Ralph into the BEDROOM. would cenainly be worse in the context of (2 3aA). Firstly, the stressing at the beginning of a sentence does not inevitably have to signal contrast (cf. (23 aB))
(as forced in (2 3bB) by context (2 3bA) ), and, secondly, it can certainly be
�aximal focus-projective as (22) shows.
In 9rder to collect systematically the data described by (22) and (23ab),
Rochemont formulates rules for focus interpretation (cf. 28). Compared to
Hohle, the main advantage of Rochemont's thesis precisely between the conditions of focus and their
is that he differentiates interpretation , so that Roche
mont assigns different tasks to different components: (26 ,Ic) (together with ( I b) ) is responsible for the extension of the focus which is decisive for the semantic contents of a sentence22 and which in Hohle's terminology must therefore be pan ofLC, and not of PC as Hohle assumes. (26.4) determines the pragrnatic-
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
case will be presented again: in English normal stress (here meaning that it allows for maximal focus) puts the accent on the lexical category on the right most side of the sentence (cf. (27) ):
82 Focus: an idea in motion
{28) a) Presentational Focus: An expression P is a Presentational Focus in a discourse �. � = (q> 1 , , <Jln} , if, and only if, (i) P is an expression in q>;, and (ii) at the time of utterance of q>1 in 6, P is not c-construable. b) Contrastive Focus: An expression P is a Contrastive Focus in a discourse �. � = (q> 1 , , <Jln}• if, and only if, (i) P is an expression in <Jl;, and (ii) if P/q>; is the result of extracting P from q>; , then P/q>; is directly c-construable, and <Jl; is not c-construable. Example (22) is to be seen as an out-of-the-blue expression: according to {28) A letter is focus. The rest, however, can be indirectly c-constructed (cf. (29) ) and gets no stress. Because (28a) is applied to (22) no contrast-reading develops. In (23bB) LAurie cannot be c-constructed, whereas the rest of the sentence (that is <Jl; without LAurie) is identical with the previous expression and is therefore directly c-construable. The term c-construction (= context-reconstruction) is central to the pragmatic interpretation of the feature [+F] and is defined by Rochemont (1986: 62£) as follows: •
•
•
•
•
•
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(29) a) An expression P is c-construable in a discourse a if, and only if, P is either directly or indirectly c-construable in a. b) An expression P is directly c-construable in a if, and only if, (i) p has a SemantiC antecedent p I in a, Or (ii) the intended antecedent ofP in () has been brought to the attention of the participants in a. c) An expression P is indirectly c-construable in a if, and only if, (i) P is a member of a lexically specified class of scenesetters, or (ii) P is an acceptable scenesetter in a by virtue of the participant's anticipated familiarity with speaker's discourse setting. (29) names the conditions where an expression p can be c-constructed. (Indirectly) c-construable are per se certain linguistic elements (such as verbs of appearance, deictic elements). Also there are words which are directly c-construable because of their semantics.24 Furthermore, elements can be c-construed through conventional and conversational regularities. Rochemont's suggestion shows how one can emphasize the differentiation of conditions ofusage and conditions oftruth, as {rightly) demanded by Hohle. This can be done in such a way that a sentence receives the information structure which it has to have according toBierwisch ( 1980). At the same time the intuitions (contrast) which we have at times about certain stresses are not neglected. But on the contrary this differentiation explains the intuitions according to some rules.
Ulrich F. G. Klein 8 3
Further interesting examples which can be explained by Rochemont's inter pretative component, and their derailed description, unfortunately have to be omitted here.25 Instead, I will turn to the above-mentioned explanation of the examples (2oH23) and therefore to the focus problems. In the correcting remark (2o), board receives a stress through NSR,. and by the rules stated in (30) to identify the focus exponent and -range {Rochemont 1986: 84, there called Focus Assignment) which are interpreted contrastively according to (28b).
By applying (3ob) to {3 1 ) Mary cooked LASAGNA for Bill. (a grammatical structure according to Chomsky (198 1) is applied to (3 1 ) ) it can be explained how the feature [+focus] is transmitted: from lasagna to the NP (3obi), from there to V (3obii), from V to VP (3obi), from VP to I (3obiii), from I to N {Mary) (3obiii) and I ' (3obi), and from I ' to I " (3obi). The PP for Bill cannot obtain [+focus] by any of the rules of (3o) but it is contained in a constituent (VP) marked [+focus]. But in (24) buy is certainly not focus according to Rochemont, because he defines it byJackendoff's conception of new information. Rochemont (1 986: 25) solves this problem with the postulate that 'the wh -specifier acts as a focus whether it is accented or not'. Rochemont's elegant solution can also be applied to German: (32) A: Ich war gestern in der Stadt und habe mir ein paar schone Sachen gekauft. Yesterday I was in town and bought some nice thingsfor myself. B: Was hast du dir denn GEKAUFT?26 What didyou BUY? I will not attempt further transfers to German because of certain problems already mentioned in Lenerz & Klein ( 1 98 8). For instance, a transfer of NSR" seems questionable not only because of the much less regulated word order in German. In my opinion Rochemont's approach separates linguistic and general-communicative competence much more exactly than Hohle. But Rochemont's definition offocus {necessary for the explanation of (25) ) as new information points to a deficit in Hohle's theory, whichJacobs can solve with his concept of Relational Focus Theory. This will now be discussed.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(3o) a) If ex, is prominent, then ex is [+focus]. b) (i) If ex is [+focus] and ex is X0, then xn is [+focus]. (ii) If ex is [+focus] and ex is an argument of X0 contained in x n , then X0 is [+focus]. (iii) If X0 is [+focus] and ex is an adjunct ofX0, then ex is [+focus].
84 Focus: an idea in motion 3·4
The Relational Focus Theory
The theories by Hohle and Rochemont described up to now defme focus as new information (in some sense). Together withjacobs (1988a), I will call such a focus conception Traditional Focus Apprehension {TFA), which is based on {asJacobs 1984: 29 calls it) Absolute Focus-Conception:27
The AF is certainly adequate for German declarative sentences where elements of certain word classes do not occur. Problems arise, however, with different sentence modes or, for example, with negations of declarative sentences or degree words, cf. (3 4)-(40) which were taken from different papers byJacobs and are set out as the following examples: (34) (35) (36} a. b. (37) (38)
a.
b. (39) (40) A: B:
Nicht GERDA hat Ottheinrich geheiratet Ottheinrich did not marry GERDA. Sogar PETER hat sein Lineal vergessen. Even PETER hasforgotten his ruler. Peter besucht sogar GERDAS Schwester. Peter visits even GERDA'S sister. Peter besucht sogar Gerdas SCHWESTER Peter visits even Gerda� SISTER. Wenn mich doch GERDAS Schwester besuchen wiirde! Ifonly GERDA'S sister would visit me! Peter nahm den GROSSEREN Schraubenzieher. Peter took the BIGGER screwdriver. Peter, nimm den GROSSEREN Schraubenzieher. Peter, take the BIGGER screwdriver. ??Ich kenne den Mann , DER dich beleidigt hat. ??I know the man WHO insulted you. Was hat Peter dir iiber Gerdas GEBURTSTAG erzahlt? What did Peter tell you about Gerda's BIR TilDA Y? Er hat erzahlt, daB sie diesmal ungeheuer viele GESCHENKE be kommen hat,ja daB sogar LUISE ihr etwas geschenkt hat He said that this time she hadgot a tremendous amount ofPRESENTS and that even LUISE had given her a present.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(33) Absolute Focus-Conception (AF}: In sentence S by interpretation of l is the part X focus of S, and the part Y is background of S only then if for I applies: X and Y are complementary in S, X comprises 'new' information (in a sense which has to be explicated) and Y comprises 'old' information (in a sense which has to be explicated accordingly).
Ulrich F. G. Klein 85 Example
(34) clearly shows that the element emphasized by stress cannot be new information, because the 'most natural context for (34) is a previous expression of the corresponding affirmative sentence, an expression where Gerda is mentioned beforehand' Qacobs I988a: 97). Gerda also cannot be Contrastive Focus in Rochemont's sense, because it can definitely be c-constructed (although one certainly has a kind of 'contrast intuition' at the reception of this sentence. This can also be seen because it can be continued with sondern Kiltharina ( but Kiltharina) ). In (3 s ) even more clearly any contrast reading is excluded, because in a possible previous sentence such as
Peter (who in this context is supposed to be very attentive) is an element of the quantity of children. If one does not want to develop a specific focus theory-which would lessen the heuristic value of the first conception-for sentences which comprise elements of the mentioned word classes,28 then a theory is needed which covers the so-called normal case as well as examples such as (34) and (35). Sentences like example (36}, which is analogous to ( s ) , have already been discussed in detail. In the optative sentence (37) the speaker can at the most assume that he is visited by somebody's sister, but one can 'hardly say that it can be presumed that the speaker assumes that the listener believes that the speaker is visited by somebody's sister, [because] a typical situation of communication for these sentences is one where no listener . . . exists' Qacobs I984: 39). The connection of focus structure on the 0n� hand and sentence mode and illocution type on the other is made even more clearly by (38} because the focusing in (38a) causes the listener to take something, in (3 8b} to do something, cf.Jacobs (I 988b: 20).29 Jacobs solves the problems which occur in these examples for TFA by formulating the Relational Focus Theory (RFT) of which the Relational Focus Conception is the basic idea: (42) Relational Focus-Conception: In sentence S by interpretation of I is the parr X of S focus of the part XI of S, and the part Y of S is background of Xr only then if for I applies: X and Y are complementary in the area of XI , X is in an XI -specific way of Xr affected particularly with regard to contents (in a sense which has to be explicated further) and Y is in an XI-specific way of Xr a prerequisite with regard to contents (in a sense to be explicated accordingly). It is remarkable that the RFT can completely integrate the results and predictions of TFA (cf.Jacobs 1 984) by not limiting the 'sense to be explicated'
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(4I} Aile Kinder waren heute so unaufmerksam. Today all the children were so inattentive.
86 Focus: an idea in motion to the new information, but by providing it with the concept of alternative relation (from Jacobs 1988b: 2 1 ) : If an element of X of a sentence S is focused, then all possible usages of S are somehow concerned with alternative placements of the position of X
(43) "'Ich kenne den Mann, das dich beleidigt hat. I know the man which insulted you . With the RFT also complex foci can be expressed as in (40), where focus and background of the particles lie in the focus of ASS (the focus binder receives an index above, the focused element below): (4o ) '
ASSi Er hat erzahlt, idaB sie diesmal ungeheuer viele GESCHENKE bekommen hat,ja daB sogar1 1LUISE1 ihr etwas geschenkt ha�. ASS He said that ;that this time she hadgot a tremendous amount ofPRESENTS and that even 1 1LUISE1 had given her a present;.
Another decisive advantage of RFT is that it has developed a process to ascertain the boundfocus (as the non-free focus is called by Jacobs) because the question test and the contradiction test (or negation test) can only be applied to declarative sentences. Pars pro toto I want to demonstrate (cf. Jacobs 1 988a, 1988b) the following tests here: an ascertaining test for interrogative sentences (44), a should request test ((4S), also c£ (3 8b) ), and a test to exclude alternatives for the particle auch ( too) (46) which functions as a focus binder: -
(44)
A: Ich weiG, daB sich jemand verliebt hat. I know that someonefell in love. B: Tatsachlich? Hat sich JUITA; verliebt? Really? Did JUITAJall in love?
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
X is therefore an element of a paradigm. If there appear degree words or negations in a sentence containing X, they act asfocus binders. Their semantics is decisive for the interpretation of the elements emphasized by stress. Jacobs (1984, 1988a) can integrate the TFA by introducing an illocutionary operator which has scope over the whole sentence: in the cases where there is maximal (or as in Jacobs's terminology, free)30 focus, it acts analogous to particles as focusing element in the sense of a focus binder. This approach can be used further: it also predicts the contexts in which an (actually questionable or even ungrammatical) sentence such as (39) can be adequately uttered, namely, then, when there is an alternative to the emphasized element who in (39). For example, this alternative can be created by the context of the possible preceding sentence (4 3), in which (39) is mentioned as a correction.
Ulrich F. G. Klein 87
{45) A: Welchen Schraubenzieher soli ich!Peter nehmen? Which screwdriver shall 1/Peter take? B: Peter, nirnm den ;GROSSEREN; Schraubenzieher. Peter, take the ;BIGGER; screwdriver. {46) A: Gerda will nur den Prado besuchen. Gerda only wants to visit the Prado. B: Nein, (sie will) aud1; die ;ALMERIA; {besuchen). No, (she) also; ( wants to visit) the ;ALMERIA;. The advantages ofRFT over TFA may be summed up here: 1.
This contribution has attempted to show how the considerations about focus have changed from an intuitive approach to a theory which can be embedded in the current modular model of language because it assigns the correct tasks-a prerequisite for empirically correct predictions-to the different levels of language. I have omitted the presentation of the rules of stress, because the latest version is presented in this volume by Jacobs. Acknowledgements I would like to thank Jiirgen l.enerz for putting me on the right track; I also would like to thank Priyamvada Bondre and especially Andrea Lepper for translating my pidgin into English. However, I take sole responsibility for any mistakes occurring, in particular in the quotations. ULRICH F. G. KLEIN Institut fiir Deutsche Sprache und Literatur Universiriit zu Koln Albenus-Magnus-Platz sooo Koln 30 Germany
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The RFT provides a coherent structure to interpret emphasized linguistic material in all possible sentences. But for TFA all those sentences cause problems which are not declarative sentences and/or comprise focus binders, such as degree words, negations, etc. which are realized as morphemes. 2. This is also true for different kinds of emphasis which are not presented here, as for example, topicalization in German. J. The RFT includes all the explanational potential of TFA. Terms such as stylistically normal (Hohle) or unmarked {Lenerz) remain pragmatically conversationally relevant Those terms acquire their meaning through the fact that the conditions of usage of sentences emerge from their semantic structures which can be analysed via RFT. In my opinion this can be seen most clearly in the so-called contrast-stressing which probably is the purest case of alternative relation. 4· The RFT can also express complex focus structures.
88 Focus: an idea in motion
N OTES
=
=
9 Example (6a} from K.iparsky (I¢6: 89), (b) from Jacobs (1988b: 20), (c) and (d) from Klein {I990). IO (7a) from Lenerz (I977: 66), (b) from Bierwisch (I¢6: I s r ) , and (c) from Klein (1990)· I I His criterion to determine the unmarked sequence of parts of sentences (Lenerz (I977: 27) could also, without difficulty, be transferred to stressing; for example, like this: if a sentence can be stressed on an element A as well as on an element B, and when the stressing of B can only occur under certain testable conditions to which A is not subject, then A is stressed normally and B is the marked part. Hohle's explanation of normal stress, which will be mentioned later, in my opinion is, then, only paraphrasing a description implicitly underlying Lenerz (I977)· 12 For a formal definition see Hohle (I982: 87). IJ (8) can be found, for example, in Paul (r88o) or in Uhmann (I 987). I4 MK (K;) means the number of all con stituents of K;; cf. Hohle (I982: 92). I 5 mu = meaningful utterance, ins = in scription, p = person, t rime, ct = con text, m = meaning; cf. Bierwisch (I 98o: 7). I6 Most authors, however, accept the normal stressing as standard and would, for example, classify the contrast-or emphasis stressing-as an exception. However, my decision for an approach which considers normal stressing as a special case is not based only on a simple logical thought (per genus proximum et differentiam speciflcam ), but also on Jacobs's Relational Focus theory, which I shall discuss below. I7 Evidence for this is also the fact that in other languages focus is marked by its own focus morphemes, such as in Japanese (cf. Uhmann I 987: 6), in the =
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
I Along with Wiese (I988: chapter 1.4) and also Vennemann (I986: chapter 2) particularly in differentiation to Chomsky & Halle (I¢8}-I accept the existence of a linguistic unit syllable and its relevance for prosodic phenomena. 2 One could possibly see the different stressing in (2H4) as other differences of these word pairs spelled out phonetically, but pair (r ) can defmitely only be distinguished through the word stress. TEnor vs. TeNOR (tenor vs. tenorist ) shows that {I) is not a singular case; and MOtor vs. MoTOR (engine) shows that it does not necessarily lead to different words. 3 cf. also Posner (I 979). 4 Besides, this seems to me to provide some piece of evidence for the thesis that persons principally expect correct and meaningful expressions of their opposite partners. If these expressions nevertheless are deficient in some way (incomplete or partially wrong according to linguistic rules), they will automatically be corrected by the listener. This happens because the listener wants to keep up the assumed correcmess of the speaker's utterance. 5 This formula describes the relations between the elements which consrirute the tripie linguistic strucrure Is (pt , syn , sem) (pt = phonetic structure, syn mor phosyntacric structure, sem = semantic structure). 6 By the way, 'wrong' stresses can also lead to ungrammatical sentences: Auch THOMAS istJaul vs. "AUCH Thomas ist Jaul ( Thomas too is lazy) . 7 Stressed ROrmally, unmarked, etc. should be understood intuitively for the time being; a more precise defmition follows. 8 This applies to English in a similar way; cf., for example, the Nuclear Stress Rule by Chomsky & Halle (I 968) or the revised edition by Rochemont (1986).
Ulrich F.
I
8
=
2 1 The difference in (26) between (1h) and ( 1 c) is that ( 1 b) operates through lexical and (1c) through phrasal categories. The " NSR probably operates on (za) (Roche mont omits to explain this). zz This is determined in LF with the rule of 'Focus Raising' (FR); cf. Rochemont
(1986: 8s):
. . ex . . .) -+ [s '
cx;[s . . . � . . .] )
[+F) 23 Rochemont's adaption of the T-model suggests a dangerous misunderstanding: because the component (26,4) is optically
principle
of Generative
To illustrate this, the standard example bachelor and unrruJrried is given. 25 cf. Lenerz & Klein (1988). 26 An alternative response would be Was
HAST du dir denn gekauft? ( What DID you buy?) . Rochemont, however, does not
examine this variant of stressing. In my opinion it can be explained, however, by Hohle's notions of VERUM focus because this paraphrase is possible: Hinsichtlich wessen ist es wahr, daB du es gekauft hast? (s Concerning what is it true
that you bought it?).
=
=
every
24
=
=
[s' .
with
Grammar.
Thema RherruJ , Given-New, Kommentar-Topic (comment-topic) , logisches Priidikat-logisches Subjekt ( logicalpredicate-logical subject) , etc. 28 Hohle (1 982: 108) did see the problems of
27
The same is rrue for pairs such as
such sentences, but excluded their examination completely. 29 I do not wholly agree with this analysis, because in my opinion (s) should be realized as optative with the stress on Schwester (sister) in order to be not marked. 30 See Jacobs in this volume, where he can show there is no such thing as Focus Projection.
RE FERE N CES H. (198 8) (ed.), Intonationsfor schungen , Niemeyer, Tiibingen. Arndt, H. & R W.Janney (1988), 'Some para Alrmann,
doxes of prosodic research and what these
may suggest about speech as multimodal
communication', Per Linguam , 4: 62-89.
Bierwisch, M (1966), 'Regeln fiir die Intona tion deutscher Siitte',
in Untersuchungen
studio gramrruJticll VII)
(3. Aufl. 1 973),
uber Akzent und Intonation im Deutschen
Akademie-Verlag, Berlin: 99-199.
(
Bierwisch, M (1980), 'Semantic srrucrure and illocutionary force', in Searle, J. R, F. Kiefer & M Bierwisch (1980), Speech Act Theory and PragrruJtics , Reidel, Dordrecht: 1 -J s . Chomsky, N. (1981), Lectures on Government and Binding, the Pisa Lectures (Studies in Generative GramrruJr 9 ), Foris, Dordrecht.
& M Halle (1968), The Sound Pattern of English , Harper & Rowe, New
Chomsky, N. York.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
19 'The notation of the rule corresponds to the conventions in Chomsky & Halle (1968: 6 1f): the condition in brackets is a more specific case: if it occurs, the condi tion in brackets is to be applied, if it does not occur then the matrix condition is " valid. Thus NSR is factually comprised of two extrinsically arranged rules' (Lenerz & Klein 1 988: 2of). zo (OPT) optional; FA Focus Assign ment; FR Focus Raising; PF Phonetic Form; LF Logical Form.
89
placed under LF (even if set off by its own number), one could think that it had a status similar to LF which would disagree
Iroqois language Onondaga (cf. Jacobs 1984: 27), or in the African language Swahili (Krifka ( 1 983)); cf.Jacobs (1988a: 1 14)· Here the last lexical category is the preposition for. This c:>se of preposition stranding is not included in the NSR
G. Klein
90 Focus:
an idea in motion
Essen , 0.
-
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
von (1956), Grundzuge der hoch Lenerz, J. & U. F. G. Klein (1988), 'Fokus deutschen Satzintonation , Henn, Ratingen. Glasnost', S&P-Arbeitsberichte, 9: I 6-3 5· Heidolph, K. E., W. Fliimig & W. Motsch Lieb, H. H. (1980), 'Intonation als Mittel (1981) (eds), GrundzUge einer deutschen verbaler Kommunikarion', Linguistische Grammatik , Akademie-Verlag, Berlin. Berichte, 68: 34-47. Hohle, T. N. (1982), 'Explikationen fur Paul, H. (r 88o), Prinzipien derSprachgeschichte "nonnale Betonung� und "normale Wort (8. Aufl. 1 970), Niemeyer, Tiibingen. stellung�·. in W. Abraham (1982) (ed.), Posner, R (1979), 'Bedeutung und Gebrauch Stazglieder im Deutschen , Nart, Tiibingen, der Satzverkniipfer in den natiirlichen 75-153· Sprachen', in G. Grewendorf (1979) (ed.), Hohle, T. N. (r988), 'VERUM-Fokus', S&P Sprechakttheorie und Semantik , Suhrkamp, Arbeitsberichte, S: I -7. Frankfurt-on-Main, 345-85. Jacobs, J. (1984), 'Funktionale Satzperspek Reis, M E. (1980), 'Grundbegriffe der tive und lllokutionssemanrik', Linguistische Semanrik', MS, Cologne. Berichte . 91: 25-58. Rochemont, M S. (1986), Focus in Generative Jacobs, J. ( 1988a), 'Fokus-Hintergrund Grammar (Studies in Generative Linguistic Gliederung und Grammatik', in H. Alt Analysis 4 ) , John Benjamins, Amsterdam/ mann (ed.), 89-134· Philadelphia. Jacobs, J. (r988b), 'Akzenruierung', MS, Uhmann, S. (1987), Fokussierung und Intona Wuppertal. tion: Eine Untersuchung zum Deutschen an Kiparsky, P. (1966), 'Ober den deutschen Frage/Antwort-Sequenzen in experimentellen Akzent' in Untersuchungen uber Akzent und Dialogen , Diss., Konstanz, appeared as Intonation im Deutschen ( studiagrammatica Fokus-Phonologie (1991), Niemeyer, VII) (3. Aufl. 1973), Akademie-Verlag, Tiibingen. Berlin: 69-98. Vennemann, T. (1986), Neuere Entwicklungen Klein, U. F. G. (1 990), 'Fokus und Akzent', in der Phonologie , Mouton de Gruyter, Cologne (Kolner Linguistische Arbeiten: Berlin/NewYork/Amsterdam. Germanistik 19 ). Wiese, R (1988), Silbische und lexikalische Klein, W. (r98o), 'Der Stand der Intonations Phonologie: Studien . zum Chinesischen und forschung', Linguistische Berichte, 68: 3-3 3· Dtutschen (Linguistische Arbeiten 2 1 1 ), Krifka, M. (1983) Zur semantischen und prag Niemeyer, Tiibingen. matischen Motivation sntoktischer Regular Wunderlich, D. (1988), 'Der Ton macht die iliiten , Fink, Miinchen. Melodie: Zur Phonologic der Intonation Lenerz, J. (1977). 'Zur Abfolge nominaler des Deutschen', in H. Altmann (ed.), 1-40. Satzglieder im Deutschen' (Studien zur dtutschen Grammatik 5 ), Nart, Tiibingen.
© NJ.S. Foundation (1991)
Journal of&rtuJntics 8: 91-1 o6
'Only' as a Determiner and as a Generalized Quantifier
SJAAK DE MEY University ofGroningen Abstract
1 'O NLY' I S N O T C O N SERVATIVE It has been observed in the literature on GQs (compare e.g. van Benthem 1986: 8) that ONLY, the denotation of'only' in an arbitrary model, lacks the property of Conservativity. We define ONLY in the following way.1 ONLY (A, B) iffB � A A relation R is called conservative iff R (A, B) iff R (A, A n B) Now we have that ONLY (A, B) .... ONLY (A, A n B) that is, (B � A) .... (A n B � A) is logically true, since its consequence is. However, (A n B � B) .... (B � A) is not logically true, in spite of the fact that its antecedent is. Hence, ONLY so defmed is not a conservative relatiotL The denotation assigned to an expression is meant to be a correct model of the meaning of that expressioiL So, in order to check whether our model of the meaning of an expression e is correct we should check our intuitions about the meanings of the sentences where e
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Two types of linguistic theories have been particularly concerned with the analysis of 'only': pragmatics, in particular focus theory and presupposition theory, and generalized quantifier (GQ) theory, the latter in the negative sense that it has been eager to show that 'only' is not a GQ.Judging from such analyses, then, it would appear that the analysis of'only' is not at home in the grammar of narural language. The main negative point of the present article is to dispute this. The main positive point is the observation that there are strong relationships between 'all', 'the' and 'only'. We propose a way to account for them.
92
'Only' as a detenniner and generalized quantifier
occurs. If a correct model of the meaning of 'only' would be conservative then the sentences { 1) Only willows weep and (2) Only willows are willows that weep
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
should, according to the defmition of Conservativity, have the same meaning, which, obviously, is not the case. The modelling of the meaning of'only' by the non-conservative relation ONLY seems .to be correct then, at least in this aspect. Conservativity is generally believed to be a defining property of determiner denotations. Compare Keenan & Stavi (1986), van Benthem (1986). Moreover, determiner denotations are taken to be the primary examples of quantifiers relevant for natural language (NL) quantifiers. Hence, the conclusion seems straightforward: ONLY is not conservative because 'only' is not a determiner. Without further background, the issue may seem to be of little importance. In order to give the matter more substance we may look at it in the following way. It is the aim of formal semantics to characterize the meanings of NL expressions in formal terms, e.g. in terms of GQs. From a formal point of view, GQs are just higher-order predicates, that is, predicates or relations taking sets as arguments. Obviously, there are, for a given domain, far more GQs than there are NL expressions whose denotations can be modelled by GQs. However, when it is essential, and not just a question of incidence, that the meanings ofNL expressions can be neatly described in formal terms, then what we may expect is that there is a very proper subset of the set of GQs, the NL quantifiers, which, moreover, admit of a neat axiomatization. One of the primary tasks of formal semantics in this view, then, is to axiomatize the set of NL quantifiers. Conservativity seems to be an appropriate axiom. ONLY fails this axiom, and hence is outside the set of NL quantifiers. However, we should realize two points. Clearly, whether the set of NL quantifiers is axiomatizable crucially depends on the way we establish which NL expressions are NL quantifiers. In particular, the issue hinges on whether 'only' is a determiner and on how important it is to be a determiner. Moreover, even if one accepts, as we do, that such an axiomatization is indeed among the primary tasks of formal semantics, one should realize that it is not apriori certain that the set of NL quantifiers is axiomatizable. Maybe only a subset of the set ofNL quantifiers is. So the fact that ONLY does not seem to be in the axiomatizable part of a subset of NL quantifiers need not be taken to be decisive. There may be other subsets of the set of NL quantifiers that do include ONLY and that are axiomatizable. What is at stake, then, is the appropriateness of Conservativity as an axiom, and, in the long run, the
Sjaak de Mey
93
2
B O OLEAN Q UANTI FIERS
Algebras are sets closed under one or more operations. Put differently, an algebra is a set of entities a subset of which are compound or structured. The structure inherent in the compound members can be described in terms of the operations of the algebra. As languages have many compound expessions, algebras are suitable models for pieces of language and their grammars. A lattice is a more highly structured type of algebra. A lattice is a set which is closed under two operations, commonly called the meet (o ) and the join {V) operation and over which a partial order (�) is defmed. A binary reflexive, transitive and anti-symmetric relation is called a partial order.3 Moreover, the order in a lattice is defined in such a way that each pair of elements a and b has both a greatest lower bound (glb) and a least upper bound Qub). For each pair of elements a and b, the meet of a and b is the glb of a and b, whereas the join of a and b is the lub of a and b: (o -GLB) (V-LUB)
X �a & X � b - X � (a 0 b) a � x & b � x - {a V b) � x
A Boolean algebra is a commutative, distributive, complemented lattice. Accordingly, the meet and join operations in a Boolean algebra are com mutative and distributive. Moreover, a complementation relation is defined assigning to each member a unique complement. Characteristic of Boolean algebras is the fact that, for an arbi� element a, (a o -.a) is an absolute
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
relation between determinerhood and being an NL quantifier. These tWo points determine the course of the discussion in the next two sections. In Sections 2-3 we create a context in which we can evaluate the appropriateness of Conservativity as an axiom. Our proposal is to pay due respect to the Boolean character of 'only'. In Section 4 we discuss what it is for an NL expression to be a determiner. Sections s-6 deal with the relations between 'only', 'all' and 'the'. There is, then, in this paper heavy emphasis on Boolean structure.2 Boolean structure is very prominent in the ways we reason. Consequently, it may be demanded of a grammar of a natural language that it properly describe the Boolean nature of natural language. We reason with meanings in a way com parable to the way in which we calculate with numbers. Boolean semantics is to formal semantics what the theory of integers is to number theory. And although the algebra of integers is not a Boolean algebra (e.g. addition and multiplication of integers are not idempotent operations whereas union and intersection of sets are), there are striking resemblances between the algebra of integers and the Boolean grammar of meanings.
94
'Only' as a determiner and generalized quantifier
minimum for the partial order, whereas
(a V -.a) is an absolute maximum for
the partial order. That is: (IRRELEVANCE1 ) (IRRELEVANCE2)
(a o -.a) � b b � (a V -.a)
'
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
It follows, from Antisymmetry, that, for all elements a and b, (a o -.a) = (b o ...., b), and also that (a V -.a) = (b V -.b). Therefore, for all a, (a o -.a) is written as o whereas (a V -.a) is written as 1 . It has been proved that the logic of Boolean algebras is representable in set theory as each power set of a given set D is a Boolean algebra. That is, the logic of Boolean algebras is, at the same time, the logic of power sets: it is part of intuitive set theory. In particular, the meet operation is representable as inter section (n) of sets; the join operation is representable as union (v) of sets; complementation is representable as set theoretic complementation ( ); whereas the partial order is representable as the subset relation (s;;;); . o is representable as the empty set, whereas 1 is representable as D. Note that in this representation, (IRRELEVANCE1 ) says that o is a subset of any subset of a given set D. What we are particularly interested in are quantifiers for a given set D (the domain) because such quantifiers are suitable models either of the meanings of certain NL expressions or of properties of such meanings. We define a quantifier in a very general way: any m-ary operation taking n1-adic, . . . , n, adic relations over D as arguments and mapping them to a �-adic relation over D is a quantifier for D. Moreover, we extend the notion of a quantifier by sti pulating that any set of quantifiers for D or any relation among quantifiers for D be a quantifier for D. In this way, properties of quantifiers are themselves quantifiers. As, quite generally, GQs are operations taking relations as arguments and having relations as values, we systematically distinguish between the Latin terms 'unary' , 'binary' , etc., and the Greek terms 'monadic', 'dyadic' , etc. We use the Latin terms for the 'arity' of quantifiers and the Greek terms for the ; 'adicity' of the arguments or values of quantifiers. If the value of a quantifier Q for a given n-tuple is a ttuth value (ttuth values are monadic relations as they are subsets of 1 , the set having just the empty set as its sole member), then we say that Q itself is a relation rather than an operation. If all the arguments of a m-ary quantifier Q have the same adicity n, we can characterize Q as an m-ary n-adic operation or relation.
Sjaak de Mey 2.1
95
The Boolean operations and the Boolean order as Boolean quantifiers
2.2
Internal negation and duality
In order to represent the meanings of yet other connectives and determiners we introduce also, alongside Boolean complementation or external negation (EN), a family of operations of Boolean internal negation (IN) and Boolean duality (DU). The operations of internal negation and duality are defmable in any Boolean algebra that has sets as members. If Q is any set of subsets of a given set D then
0
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
As the power set of a given set D is a Boolean algebra, and both the three Boolean operations (i.e. intersection, union and complementation) and the Boolean order (i.e. the subset relation) are clearly quantifiers, we have with each set D a set ofBoolean quantifiers for D. Many NL expressions tum out to have a Boolean meaning. Propositional connectives, and a number of determiners and determined NPs take their denotations from among the Boolean quantifiers. They are moreover the NL quantifiers par excellence .4 We distinguish between Boolean algebras according to the type of their members. The following Boolean algebras are specifically relevant for NLs: the truth value algebra (propositional calculus), and the property algebra, the determiner algebra and the NP-algebra for a given domain D. These algebras are sets of truth values, properties ( = subsets of D, hence, monadic relations over D), determiner denotations and NP-denotations, respectively. They provide models of the Boolean meanings of words such as 'and' , 'or', 'not' , connecting sentences, nouns or intransitive verbs, determiners and determined NPs, respectively, and of the words 'if' and 'all', denoting relations between truth values and properties, respectively. In spite of the fact that the GQs modelling aspects of the meaning of any of these words are formally different in that they take arguments of differing categories and, hence, belong to different Boolean algebras, they are really the same relations, that is, they are insensitive to the precise character of their arguments. Their Boolean nature guarantees uniform semantic behaviour. Of course, this should not be taken as implying that such words cannot have other than Boolean meanings as well. In what follows we restrict attention to quantifiers that are relations,5 although everything we say on them can be readily generalized to cover also the case of quantifier operations. Hence, although the operations of intersection, union, complementation, and their Boolean compositions are Boolean quantifiers as well, they are of less relevance for our purposes. For a more general treatment we refer to de Mey (forthcoming).
96
'Only' as a determiner and generalized quantifier
EN(Q) IN(Q) DU(Q)
= = =
POW(D) - Q {X � D I (D X) E Q} EN(IN(Q) ) = IN(EN(Q) ) -
If R is a binary relation over POW(D) then, of course, EN(R) is the com plement with respect to POW(D) x POW(D). As to internal negation, we should distinguish between left internal negation (L-IN), right internal negation (R-IN) and left and right internal negation (L&R-IN): L-IN(R) R-IN L&R-IN
= = =
{(X, Y) E POW(D) = POW(D) I (X Y) E R} {(X, Y) E POW(D) X POW(D) I (X, Y ' ) E R} {(X, Y) E POW(D) X POW(D) I (X' ' Y ' ) E R} I '
DU{Q) = IN(EN{Q) ) EN{IN{Q) ), IN(Q) ) = L-IN(EN(Q) ) L-DU(Q) = EN(L= R-IN(EN(Q) ) EN(R-IN(Q) ) R-DU(Q) L&R-DU(Q) = EN(L&R-IN(Q) ) = L&R(IN(EN(Q) ) In what follows we use the signs 'IN' and 'DU' also as dummies for arbitrary members of the families of internal negations and duals. Looking upon POW(D) as a Boolean algebra, then, the subset relation and all the Boolean compositions by intersection, union, external negation, internal negation and duality are (instances of) Boolean quantifiers for D as the following definition says:6 =
=
·
1. The subset relation over POW(D) is a Boolean quantifier for D 2. If Q1 and Q2 are Boolean quantifiers for D, so are EN(Q), IN(Q), DU(Q), (Q1 1'""\ Q2) and (Q1 u Q2) J. If Q 1 , . . . , Qn are Boolean quantifiers for D, and ifR is a n-ary Boolean relation defined for Q I , . . . , Qn, then R is also a Boolean quantifier for D Note, furthermore, that the truth value algebra has a fixed domain: 2, which is the set of truth values, the power set of 1. A property algebra and, hence, a quantifier algebra depend on a given domain D, however. In order to reach a more general definition we should be able to drop the reference to a given set D. We accomplish this by introducing the following two axioms of Extension and Quantity*, and requiring that all Boolean quantifiers obey these axioms: If A, B � D � D* then R0(A, B) iff R0.(A, B) (EXT) (QUANT) If D � D* and f is a I- I function mapping D into D* then Ro(A, B) iff �l(f(A), f(B) ) So, if a quantifier is a Boolean quantifier for D it is also a Boolean quantifier for any D* which is either a superset ofD or the permutation product of a superset of D under a permutation of D.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
We can furthermore defme the operations DU of duality:
Sjaak. de Mey
97
From these two principles, a more common principle of Quantity is derivable: (QUANT) If 1t is a permutation of D then Ro (A, B) iff R0 (1t (A), 1t (B) ) A quantifier obeying (QUANT) is called logical. All Boolean quantifiers are logical, then. Given these axioms, we can also look upon connectives and Boolean deter miner denotations as denoting essentially the same relations. That is, if we write the denotation of if as IF, and the denotation of 'all' as ALL, IF is the same relation as ALL, disregarding the character of the entities these relations are defined over. Moreover, we also have the following correspondences: '
R-IN
L-IN
L&R-IN L-DU
R-DU
AND ONLY-IF NOT-OR IF NOT-IF NOT-AND OR ALL NOT-ALL NO ONLY..NOT ONLY NOT-ONLY..NOT SOME
What is especially nice about the approach we have taken here is that we can calculate the properties of the other Boolean determiners if we take the properties of the order as a starting point and look at how they are transformed under the operations yielding compound Boolean quantifiers. Here is an example. ALL is a partial order and, hence, it is reflexive, antisymmetric and transitive. ALL has also some other properties, e.g. it is monotonic on both sides. More in particular: ALL is L-decreasing and R-increasing.7 Note that this follows from the fact that ALL is transitive. Hence, the logic of monotonicity of Boolean quantifiers is part of the logic of Boolean algebras, more in particular it is derivable from the transitivity of the subset relation. It is an easy exercise now to establish which properties are preserved under EN, IN or DU. For example, NO is non-reflexive. As NO is the right internal negation of ALL we have that for all X � D: NO (X, X) iff ALL (X, X'). There is only one case in which a set X � D is a subset of its own complement: that is so if X is empty. This is in accordance with (IRRELEVANCEI ).
3
C O N SERVAT I V I TY A N D RELATED PROPERTIES
Let us go back now to Conservativity. Actually, for two-place relations we should distinguish between L-Conservativity, R-Conservativity and L&R Conservativity. What is generally called Conservativity is R-Conservativity. A quantifier is
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
EN
'
98 'Only' as a determiner and generalized quantifier
L-Conservative iff R (A, B) iff R {A "' B, B) R-Conservative iff E.(A, B) iff R {A, A "' B) L&R-Conservative iff R (A, B) iff R {A "' B, A "' B) ALL is R-conservative. This follows from some general lattice properties of the meet {intersection) operation and the order (the subset relation): 1.
2.
A � {B "' C) - {A � B) & {A � C) {A � B) & {A � C) - A � {B "' C)
[Transitivity ('"'-GLB
and taking C A.8 Hence R-Conservativity is certainly a Boolean property of ALL. ALL is neither L-conservative nor L&R-conservative, however. Yet, ALL has another interesting left-hand property instead: it is L-progressive. We define ==
Again, that ALL is L-Progressive follows from some general lattice properties of the union operation and the subset relation: 1. 2.
(B u C) � A - {B � A) & {C � A) (B � A) & (C � A) - (B u C) � A
[Transitivity [u-LUB
and taking A C.9 The following schema shows the distribution of Conservativity and Progressivity over the connectives and Boolean determiners: =
L-CONS
AND
L-PROGR
R-CONS
R-PROGR
SOME
IF
ALL
IF
ALL
ONLY-IF
N-AND
NO
N-IF
N-ALL
N-IF
N-ALL
N-ONLY-IF N-ONLY
ONLY-IF
ONLY
AND
SOME
OR
ONLY..NOT
N-ONLY-IF N-ONLY
N-AND NO
N-OR N-ONLY..NOT
ONLY
OR
ONLY..NOT
N-OR
N-ONLY.NOT
From this schema it might appear that L-CONS and L-PROGR are preserved under L-IN, whereas R-CONS and R-PROGR are preserved under R-IN. Also, it would appear that L-CONS and L-PROGR are reversed under R-IN whereas R-CONS and R-PROGR are reversed under L-IN. Yet, this would not be an appropriate thing to say. Let us notice Hrst that ALL has two further properties that we will call 'L complemented Conservativity' {L-C-CONS) and 'R-complemented Progres sivity' (R-C-PROGR). We defme R is L-C-Conservative iff R {A, B) iff R (A "' B ' , B) R is L-C-Progressive iff R {A, B) iff R {A u B ' , B)
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
R is L-Progressive iff R {A, B) iff R (A u B, B) R is R-Progressive iff R (A, B) iff R (A, A u B) R is L&R-Progressive iff R {A, B) iff R (A u B, A u B)
Sjaak de Mey
99
R is R-C-Conservacive iff R (A, B) iff R (A, A ' r� B)
R is L-C-Progressive
iff R {A, B) iff R (A, A ' u B)
ALL is L-C-Conservacive: 1 0 I.
{A r'l B
I
) � A & A � B .... A
2. A r� B � B J. A r'l B = (A r'l B 4· A � B I
I
I
)
r'l
r'l
B �B I
B=0
[Transitivity [Hyp [see note I O
[3 [2,
S· A r'l B � B - A � B I
ALL is R-C-Progressive:
s
[Hyp
3· A = (A r� A ' ) u (A r� B) 4· A = A r� B
[2, Distribucivity
[2, note I O [3, Contraction [4
[3 [4, note IO
[Hyp
S· A � B
[I , note I O
[ I , note I O
The proof can be read in both directions. Furthermore, L-PROGR and L-CONS are transformed into L-C-CONS and L-C-PROGR, respectively, under L-IN, and vice versa, whereas R-PROG and R-CONS are transformed into R-C-CONS and R-C-PROGR, respectively, under R-IN, and vice versa. Furthermore, L-PROGR and L-CONS are transformed into L-C-PROGR and L-C-CONS, respectively, under R-IN, and vice versa, whereas R-PROGR and R-CONS are transformed into R-C-PROGR and R-C-CONS, respectively under L-IN, and vice versa. To give just one example: suppose R is R-CONS. Then R (A, B ' ) iff R (A, A r� B ' ). Then R-IN(R) (A, B) iff R-IN(R) (A, (A r� B ' ) ' ) = (A, A ' u B). The conclusion may be clear. It is not the case that L-CONS (R-CONS) and L-PROGR (R-PROGR) are preserved under L-IN (R-IN), nor is it correct to say that L-CONS (R-CONS) and L-PROGR {R-PROGR) are reversed under L-IN (R-IN). The fact that, for example, NO, which is the R-IN of ALL, is R-conservacive just as ALL, follows from the fact that ALL is also R-C-PROGR and R-C-PROGR is transformed by R-IN into R-CONS. Given this framework we are now in a better posicion to evaluate the claim that the den9tacions of all determiners are R-Conservacive. In this framework
we derive the properties that determiner denotations have from some more basic facts, in particular the properties of Boolean algebras. Any grammar of
account for the Boolean character of NLs. In a Boolean algebra, however, there is no privileged posicion for R-Conservacivity.11 Note that if non-conservative determiner denotations are not NL quan tifiers, the corresponding connectives are not NL quantifiers either, because
NL should
they are essentially the same relations. So, AND would then be a NL quantifier,
but OR would not. This result is very implausible indeed.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
1. A � A' u B 2. A = A r� (A ' u B)
100
'Only' as a determiner and generalized quantifier
As for 'only', ONLY is the L&R-IN of ALL, or, equivalendy, ONLY is the converse of ALL,the superset relation. As we have shown, ONLY has respect able Boolean properties. Hence, ALL is an NL quantifier iff ONLY is. Some thing similar holds of the connectives 'IF . . . THEN' and 'ONLY IF', which are the subset relation and the superset relation among truth values, respectively.
4 T HE A U T O N O M Y O F SYNTAX
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Having said this, it may still be relevant to check whether the claim that 'only' is not a determiner is tenable, and, if it is, to discuss the question of how important it is to not be a determiner. So let us first look at the reasons why 'only' is said not to be a determiner. These have to do with the fact that 'only' can combine with proper names ('only John'), definite descriptions ('only the pilot who shot at it') and numerical NPs ('only three pilots who shot at it'). So, 'only' combines with NPs whereas genuine determiners should combine with Ns only. Therefore, 'only' is said to be an adverbial. Now, in fact, we should ignore deflnite descriptions and numerical NPs, otherwise the case is immediately lost: also 'all' combines with them ('all the pilots', 'all three pilots'). What should be shown, then, is that in 'only pilots' 'only' is not a determiner. No argument to support this claim has been put forward, as far as I know. It is of interest to note that we do not have '"'no the pilots', '"'some the pilots' or '"'each the pilot', although we do have 'all the pilots'. So, taking the argument seriously, we should conclude that 'all' is not a determiner either. Clearly, the whole quest becomes dubious. There is another consideration. It is well known that many determiner words may occupy two different positions in the sentence, an NP-internal position as in 'All the boys were dancing', and a 'floated' position as in 'The boys all danced'. Clearly, the two occurrences of 'all' do not essentially differ in meaning and we should apply a GQ analysis to both occurrences of this word. So, even if we had to conclude that 'only' is not a determiner because it does not occur in NP-internal position, nothing seems to follow. There are many quantifier words occurring only in one of the two positions discussed. This may also be a language-specific matter in the sense that there may be languages in which what apparendy correspond to English NP-internal quantifler words may appear in floated position only. So what we would have to prove is not only .that 'only' is not NP-internal (in English, that is) but that it is neither a floated quantifler. Now the second issue. How important is it to be a determiner? Clearly, this is an issue that demands much more discussion than we can give it here. We merely emphasize the following theoretical consideration. If we take deter-
Sjaak de Mey
101
minerhood to be important we may do so because we are inclined to believe that syntax dictates to us the relevant distinctions. In this context I woud like to cite a passage from Montague's paper 'English as a Formal Language' (1974: 2 10).
There are two reasons to pay special attention to this passage. First, the last
sentence can be taken to be a formulation of the principle of Compositionality,
which says that the meaning of a compound expression is composed from the
meanings of the compounding expressions. Second, it also says that syntax and semantics should go hand in hand, but that syntax cannot be the guide. If this is
the right track we clearly have to get rid of the idea that syntax is autonomous, that is, that syntax dictates the correct distinctions.
Strangely enough, there are many people working in the tradition of Montague grammar who nevertheless defend a much narrower version of the principle of Compositionality. According to such narrow versions, syntax does lead semantics. Therefore it is important to emphasize that this is not what Montague himself defended. Moreover, the spirit in which we approached the
analysis of 'only' in the previous section is in complete agreement with the
passage cited above. 'All' and 'only' are closely related from a semantic point of view, and whatever differences there may be in syntactic behaviour, they are accidental rather than essential. Syntax cannot dictate the proper analysis of 'only'.
In the remaining two sections of this paper I shall elaborate on some similar ities between 'all', 'the' and 'only'.
5 EXI STE NCE CLAI M S The relation ONLY as defined above is the denotation o f the word 'only' as we fmd it with bare nouns such as in 'only pilots'. An interesting question now is
whether there is an existence claim on the second argument of 'only'. That is to say, a sentence such as
(3) Only pilots who shot at it escaped the Mig that chased them
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Some linguists roughly sharing the main goal of the present paper, that is, to define the notion of a true sentence of English (or English sentence true with respect to a given interpretation), have proposed that syntax-that is, the analysis of the notion of a (correctly formed) sentence-be attacked first, and that only after the completion of a syntactic theory considera tion be given to semantics, which would then be developed on the basis of that theory. Such a program has almost no prospect of success. There will often be many ways of syntactically generating a given set of sentences, but only a few of them will have semantic relevance; and these will sometimes be less simple, and hence less superficially appealing, than certain of the semantically uninteresting modes of generation. Thus the construction of syntax and semantics must proceed hand to hand.
102
'Only' as a determiner and generalized quantifier
states that any pilot who escaped the Mig that chased him was a pilot who shot at the Mig that chased him The question is, does it also imply that there were pilots who shot at the Mig that chased them? As ONLY is the superset relation, that is, the converse of the subset relation, it does not enforce this. This analysis of 'only' is in keeping with the one Geach gave. I refer for a short survey of the various analyses of 'only' to Atlas (this volume). This is, moreover, precisely what we should expect given the fact that ONLY is the converse of ALL. Yet 'all' is often understood with an existence implicature on its first argument rather than an existence claim. 'All students read books' does not necessarily imply that there are students, although it may suggest it. So, we could argue that in the case of 'only' it is also an implicature rather than an implication that there are pilots who escaped the Mig that chased them. An argument to support this can be derived from sentences such as .
If this is a grammatical and non-contradictory sentence then we should conclude that the definition we gave of ONLY is a correct model of the meaning of 'only'. I would say that (4) is, indeed, grammatical and non-con tradictory. With an eye on what follows, it might be wise to emphasize once more that with 'only' in the sense we analysed here the existence implicature clearly rests on the second argument. To be sure, in models where there are no students, (4) can only be true if there is no one who reads books. Yet even in models where there are students, (4) can be true although there is no one who reads a book.
6 OTHER F O R M S O F ' O NLY' Let us now tum to other collocations of 'only': the 'proper name-only' (as in 'only John'), the 'definite description-only' (as in 'only the pilot'), and the 'numeral-only' (as in 'only three pilots')P We start with the definite description-only as we find it in
( s) Only the pilot slept Let us first point to a technical difficulty that arises from the type of frame work we use. Perhaps it would be natural to analyse ( s ) as expressing that the set of sleepers is a subset of the set denoted by 'the pilot'. However, this would be a category mistake as 'the pilot' is an NP, and, hence, not the kind of set which SLEEP can be a subset of. This may seem to suppon the point of view of those who have denied 'only' the status of a generalized quantifier. I do not accept this. In reaction I would point to the case of 'all the'. With the latter, the solu-
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(4) Only students, if anybody at all, read books
Sjaak de Mey 103
cion is to defme a quantifier ALL-THE. So let us try and work out this solution for 'only' as well: ·
ONLY-TlfE.s (A, B) iff iBI = I & B � A Note that this quantifier ONLY-THE is the converse ofTHE,g: THEsg (A, B) iff IAI = I & A � B
ONLY-THE,s (A, B) iff !AI = 1 & B � A Note that this definition leaves the contribution of THE untouched. Yet it would be completely parallel to the case of 'all the' if we look upon ALL-THE as the composition of ALL and THE,g where the latter is defined as follows: THE,s (A, B) iff !AI - I & A � B Whether this account of 'only the' is correct depends on whether sentences such as
(6) Only the pilot, if anybody at all, was asleep are grammatical and non-contradictory. I would say that (6) is an acceptable sentence. We should extend the analysis to proper names now. Proper name denota tions are entities, members of D, not generalized quantifiers. However, proper names can be raised to generalized quantifier level. JOHN, then, is the set of John's properties: JOHN = {X � DljE X} We now defme a generalized quantifier for each proper name a ONLY-PN. = {X � DI X = {a} and defme further ONLY-PN. (A,B) iff A - {a} & B � A ONLY-PN., basically, is the set of singleton properties that a has. Still , the second argument may be empty. Again we have to check our intuitions about sentences such as (7) Only John, if anybody at all, slept
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Note further that the quantifier THE,g is closely related to ALL: the difference is in the cardinality claim on the first argument with THE that is lacking with ALL. Yet there is something unsatisfactory about this solution: the definition of ONLY-THE does not require that the first argument be a singleton set as well. Another natural definition is
I04 'Only' as a determiner and generalized quantifier
Still a different case is the 'numeral-only' as we find it in sentences such as (8) Only three pilots are sleeping Again we introduce a quantifier ONLY-THREE (A, B) iff iAJ = 3 & B = A This definition runs parallel to the definition of ONLY-THE. Again, the crucial question is whether sentences such as (9) Only three pilots, if anybody at all, slept
(ro)
As
few as three pilots slept
in the sense of: three pilots slept and that number is smaller than we may have expected. Here the set of pilots is a subset of the set of sleepers, and not vice versa. Furthermore, it is not unreasonable to hold that we find this kind of 'only' only with numerals. In the other three cases we discussed, the bare plural 'only', the proper name 'only', and the definite description 'only', there was no numeral. We cannot replace 'only the pilot' in (3 ) and 'only John' in (6) by 'as few as the pilot' and 'as few as John'. SJAAK DE MEY Institute for General Linguistics Oude Kijk in 't Jat Straat 26 97 I 2 EK Groningen Netherlands
N O TES I We use the sign 'b' for the subset relation. Later on (Section 3 ) we will also use this sign for the partial order in a lattice. No confusions should arise from this systematic ambiguity. 2 The Boolean character of NL has been extensively demonstrated in the literature. Compare Keenan & Faltz (I984), Zwarts (I986}. We trust that the
3
reader is familiar with the concepts of lattices in general and Boolean algebras in particular and, hence, we will not discuss the structure of such types of sets in detail A relation R is reflexive iff for all X � D: R (X, X). A relation R is antisymmetric iff for all X, Y � D: R (X, Y} & R (Y, X) - X Y. A relation R is transitive iff for all X, =
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
are accceptable. I must admit that my intuitions here are not as certain as in the other cases. My feeling is that (9) is a possible, but not very prominent reading of (8). But then, we should certainly not exclude the possibility that there are more senses of 'only', senses in which it is not the converse of 'all'. Note that instead of (8) we could also say
Sjaak de Mey I05
R
Here is a proof. From left to right: 1. A � B (HYP A � B & A � A (� is reflexive, PC 3· A � B & A � A - A � A n B (nGLB 4- A � A n B (MP 2,3 5· A n B � A (('\..LB 6. A e A n B [� is anti-symmetric, 4,5 2.
From right to left: 1. A '= A n B (HYP 2. A !; A [� is reflexive 3· A !; A n B [I ,2, substitution 4· A � A n B - A � A & A !; B (('\..GLB 5· A � B IFF A !; A & A � B (PC 6. A � A n B - A � B (4,5 7· A !; B (MP 3 ,6 8. A e A n B - A � B (I ,7 I I We already noted that the often discussed monotonicity properries of determiner denotations are also derivable (and, hence, should be derived) from the transitive character of the order in a Boolean algebra. I2 To be sure, there are more collocations than we discuss here, e.g. we also have 'only' as a VP-modifier as in ' We can only pray'. We cenainly do not claim that in all cases where 'only' can be used only one analysis of 'only' can be given. Nor do we claim that one and the same analysis should hold for all possible occurrences of 'only'. The only sense of 'only' we are concerned with here is the Boolean ONLY which is the converse of ALL. Note that in 'we can only pray' 'only' also seems to be the converse of 'all'; compare 'all we can do is pray'. That is,
ALL ( WE-CAN-DO, PRAY) = ONLY (PRAY, WE-CAN-DO)
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
R (B, C) - R (A, C). is non-reflexive if not for all X � D: (X, X) e R, and irreflexive iff for no X � D: (X , X) e R Similarly for the other properties. 4 Determined NPs denote the image set of the denotation of their Detenniner for their head noun denotation: � frn.Det N) � = (X � Dl DET(N , X) ) and, hence, are in a sense derivable from determiner denotations. Put differently, the logic of detennined NPs is a proper pan of the logic of detenniners. Subject detenniners have quantifier relations as denotations whereas object determiners are quantifier operations. 6 Compare what was said at the end of Section 2.1. Note funhermore that in this definition EN, IN, DU, n and u are operations defined over quantifiers, not properties. They are operations of Boolean quantifier algebras, not propeny algebras. Due to the Boolean character, however, they are really the same operations as the corresponding operations of propeny algebras. 7 A relation R is L-decreasing iff for all X, Y, Z � D: R (X, Y) & Z � X - R (Z, Y) and R-incteasing iff R (X, Y) & Y !; Z R (X, Z). These properries themselves are sets of relations, hence quantifiers. 8 More fully written out we have in the first line that A � (B n C) & (B n C) !; B - A � B. and A � (B n C) & (B n C) !; C - A � C. 9 That is, B � (B v C) & (B v C) £:; A - B £:; A, and C � (B v C) & (B v C) !; A - C f:; A I O Use is being made of the theorem saying that A � B iff A = A n B Y, Z � D: R (A, B) &
1o6 O nly '
'
as
a determiner and generalized quantifier
RE FERE N CE S determiners', Linguistics and Philosophy, 9, 3 = 2 5 3 -)26.
Mey, S. de (1990), Determiner Logic, or the Grammar of the NP. PhD. diss, University of Groningen. Montague, R (1974), 'English as a Formal Language', in: R Thomason (ed): Formal Philosophy, Selected Papers of Richard Mon tague , Yale University Press, New Haven and London. Zwarts, F. (1986), 'Categoriale grammatica en Algebrai:sche Semantiek', Ph.D. diss., University of Groningen.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Atlas, J. (1989), 'Topic/comment, presup position, logical form, and focal stress implicatures', article read at the Focus Workship at the Fraunhofer Institut fiir Arbeitswirtschaft und Organisation, 2830 June 1989. Benthem, J. van (1986), Essays in Logical Semantics, Reidel, Dordrecht. Keenan, E. & L. M Faltz (1984), Boolean Semantics for Natural Language , Reidel, Dordrecht. Keenan, E. & J. Stavi (1 986), 'A semantic characterization of natural language
© Nl.S. Foundation (1991)
journal ofSmtantics 8: 107-us
External and Verb Phrase Negations in Actual Dialogues R OB T. P. WI C H E University ofLeiden Absuact
1 EVENT L O G I C In this section I will say a few words about event logic. My version of event logic
is heavily inspired by a paper of Parsons (198 ;) and, to a lesser degree, by some papers by Marrin (198 1a,b). Martin says that the explicit need for linguistic purposes for variables for
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Negations play an important role in actual dialogues. If one participant of a dialogue is negating an utterance of the other participant (or is uttering a sentence that entails a negation of an utterance of the other participant), there occurs a verbal conflict between the parti cipants} This conflict is resolved as soon as a participant is forced (only by verbal means, of course) to give up a sentence so that there is no longer a conflict. In recent theories about.negation a distinction is rather often made between two different kinds of negation.2 Gabbay & Moravcsik (1978) and Hoepelman (1979) make a distinction between (sentence) negation and denial. Gabbay and Moravcsik motivate their distinction as follows: 'That the negation of a true proposition takes us to a false one, is one of the early . lessons in elementary logic. Sentence negation is important for logic, for it gives us general ways of characterizing contradictoriness, and thus helps us formulating such basic laws as the law of non contradiction. In everyday discourse however, negative sentences are used to formulate denials of various sorts. In fact, even the notion of a denial is too narrow; denial, objection, criticism, etc. are all everyday activities the point of which is to say: "No, it is not like this; rather, it is like that." '3 Jacobs (1982) distinguishes between 'kontrastierende Negation' and 'nicht kontrastierende Negation', and Hom (1985, 1989) between Truth functional (descriptive, logical) negation and Metalinguistic (non-descriptive, non-logical) negation. Barrh & Wiehe (1986) distinguish between three kinds of negation: Exclusion negation, Choice negation and Discrepancy negation. I will try to combine all these approaches. In the style ofJacobs I will make a distinction between Contrasting Negation (Cneg) and Non-Contrasting Negation (NCneg). Cneg has to be divided in Choice negation and Metalinguistic negation. NCneg has to be divided in Exclusion negation and Verb Phrase negation. Verb Phrase negation corresponds roughly with the Discrepancy negation from Barth & Wiehe. NCneg is characterized formally by the negation operator '-' from classical, two-valued logic, Cneg by the metalingnistic operator '-KORR'. 'KORR' is Jacobs's correcmess operator. Event logic will be used as a valuable instrument for characterizing formally the difference between the different types of negation that are discussed in this paper.
108 External and verb phrase negations in acrual dialogues
events was first noted by C. S. Peirce.• Parsons states that the .idea that many sentences can be assigned logical fonns that make reference to, or quantify over, events, states and processes was first proposed by Hans Reichenbach, and was worked out in some detail by Donald Davidson.5 'One phenomenon that Davidson's account handles in a nice way has to do with certain adverbial modifiers (certain adverbs together with certain prepositiotial phrases functioning adverbially). Davidson's idea was that these modillers appear in logical fonns as predicates of events.'6 One advantage of event logic is that it accounts in an easy way for the fact that each of the following sentences
is generated by7 (4) Mary walks slowly in the back yard If we treat (4) as an untensed sentence, its logical form is {Ee}[(walk0Xe) & Agent(Mary0,e) & slowly0(e} & in°((the back yard)0, e) ]
(s)
The logical fonns of (1), (2) and (3) are the same as the logical form associated with (4) with one or more conjuncts dropped out. With the given method of symbolization, the inferences are valid in the ordinary predicate calculus. The brackets '(' and ')' around 'walk' are not used by Parons. I borrow them from Martin and use them to indicate that in ( s ) only '(walk0)(e)' represents an event; the event of walking. If we would not use these brackets, 'slowly0(e)' could be interpreted as 'e is an event ofslowly'. The variable 'e' in ( s ) ranges over events. 'Agent(Mary0, e)' means that Mary is the agent of the walking event, 'in 0((the back yard)0, e)' means that the location of the walking event is in the back yard. 2
C O O PERATIVE AND C O MPETITIVE DIALOGUES
Carlson (1984} makes a distinction between cooperative (agreement-seeking} and competitive (truth-seeking} games. The aim of the players in a cooperative game is 'to make their private lists match each other.'8 The aim of the players in the competitive game is 'optimally satisfied if the opponent is forced to unilaterally give up his conflicting assumptions and accept the other's view'.9 There is, however, one thing that both kind of dialogues have in common:
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
( 1} Mary walks slowly Mary walks in the back yard (3) Mary walks
(2)
Rob T. P. Wiehe
109
Def. I p{p ') is forced with verbal means by p ' (p) to give up a sentence S in a dialogue D := p '{p) shows {proves) in D in a correct way that S is false or that the state ment S is not correct, or a person q of which the authority with respect to the truth value ofS is acknowledged by both p and p ' , claims that S is false. We all know that it is often not possible (for practical or theoretical reasons) to end a competitive phase of a discussion by forcing an opponent to give up a sentence. In that case the participants can decide to disregard their conflict for the moment and to continue the dialogue. In that situation there is a transition, too, from a competitive to a cooperative phase.
3
EXTERNAL NE GAT I O N {Extneg)
If there is an explicit verbal conflict between a statement 'S' of a participant p and a statement 'Not S' of the other participant p ' , and 'S' is uttered in the dialogue before 'Not S', we call the statement 'Not S' an external negation (Exmeg). If the only function is the conveying of new information, we call it a
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
'Whether by cooperative information sharing or by competitive debate, the players attempt to make their sets of assumptions coincide.'10 The basic distinction I prefer to make is not between cooperative and competitive dialogues, but between cooperative and competitive puts of dialogues. Competitive parts are parts in which there is a 'conflict of avowed opinion'.1 1 Cooperative parts are parts in which there is no such explicit verbal conflict. Unless the participants decide that their dialogue will be a competitive one from the start, it seems a good working hypothesis to me that dialogues start as cooperative dialogues. As soon as there is a conflict of avowed opinions, the dialogue becomes competitive. If there is no longer such a conflict, the dialogue becomes cooperative again. The verbal conflict is resolved as soon as a participant gives up a sentence so that the conflict disappears. There are two possibilities here: ( I ) a participant is forced with verbal means to give up a sentence; (2) a participant voluntarily gives up a sentence. Negation plays an important role in all this. If one participant negates an utterance of the other participant (or utters a sentence that entails a negation of an utterance of the other participant), there is definitely a conflict of avowed opinions. This conflict is resolved as soon as a participant is forced with verbal means to give up a sentence so that the conflict disappears.
1 10
External and verb phrase negations in actual dialogues
Verb Phrase negation (Section 4). In this case there is no disagreement between the participants at all. There are three types of Extneg: Exclusion negation (Section 3-1 ), Choice negation (Section J.2), and Metalinguistic negation (Section J.J). 3.1
Exclusion negation ( Excluneg)
Kohl doesn't negotiate (7)
Kohl negotiates (6}
According to Mannoury, Exclusion negation is the oldest 1 3 function of negations. If we choose the following symbolization for (6): (8) negotiate*(Kohl*) the symbolization of (7) is (9) - negotiate*(Kohl*) The negation operator in (9) is as usual placed in front of rest of the sentence. The conjunction of (6} and (7) is, according to their symbolizations, an explicit contradiction. R1 A negative statement 'Not F' of a participant p (p ' ) in a dialogue D is an Excluneg, with symbolization '- F' if and only if I . 'Not F' is a reaction of p (p ) upon an earlier statement 'F' in the same dialogue D of p ' (p). 2. p (p ) does not suggest or present an alternative for F. '
'
Sometimes a person explains why he or she makes a certain (negative) state ment. Consider, for example, the following dialogue: p p' The king of Romania is rich ( 1 o) The king of Romania is not rich (ua), for Romania does not have a king ( I I b)
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Gerrit Mannoury (1 867-19S6}, Dutch philosopher and intellectual leader of the Signifies Movement in the Netherlands, was the Hrst to make an explicit distinction between Exclusion negation (from now on: Excluneg} and Choice negation (Choiceneg). An Excluneg rejects something without putting an alternative in its place. The attention is concentrated entirely on the possibility which is excluded.12 We see an example of an Excluneg in the following dialogue: p' p
Rob T. P. Wiehe
I II
an interpretation of ( u ) , and other sentences of this type, I choose (with '- R' and '- K' as abbreviations for 'The present king of Romania is not rich' and 'Romania does not have a king'). 14
As
(12) - R BECAUSE!FOR - K My symbolization of (12) is 15
{I 3) - K & (- K - - R) 3 .2
Choice negation ( Choiceneg)
Kohl negotiates (6) Kohl doesn't negotiate (7) Kohl bargains (I 4) The utterances of (7) and (I4) together form an example of Choiceneg. In this paper I will not choose (9) as an interpretation of (7). Instead I will choose another way to symbolize (7) and (I 4) which is borrowed from event logic. 17 The symbolizations are
(I s ) - (Ee)[During(e, now0) & (negotiate0Xe) & Agent(Kohl0, e) ] (I6) (Ee)[During(e, now0) & (bargain°Xe) & Agent(Kohl0, e) ] In the classical, two-valued, logic, we may replace (I s) by
(I7) (Ve)[- During(e, now0) v - (bargain°Xe) v - Agent(Kohl0, e) ] The following sentence (IS} follows logically from (I S) and (I6) (Ee)[During(e, now0) & -(negotiate0Xe) & (bargain°Xe) & Agent(Kohl0, e) ] There is a great advantage in this way of symbolizing sentences. We translate (7) as an external negation and can deduct from the symbolizations {I S) and {I6) another sentence in which the negation operator does not have the maximum scope, and that can be seen as a translation of (7) and (I 4) together, or as a trans lation of the sentence
{I9) Kohl doesn't negotiate, but bargain
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
A Choiceneg presupposes the distinction of two alternatives. By rejecting one of these alternatives, one simultaneously afHrms the other. The attention of speaker and hearer is divided over both alternatives.16 The following dialogue offers an example of a Choiceneg. Metalinguistic negations are treated in the following section. p' p
I 12
External and verb phrase negations in actual dialogues
If a participant of a dialogue can choose between Excluneg and Choiceneg, the Choiceneg is the most cooperative choice.
R2 A negative statement 'Not F, (but} G' of a participant p in a dialogue D is a Choiceneg, with symbolization '- F & G', if and only if r . 'Not F, (but} G' is a reaction of p upon an earlier statement 'F' of the
2. 3·
other participant p', and The interpretation '- F & G' doesp't lead to a contradiction and is not counterintuitive. The interpreter of the statement 'Not F, (but) G' believes that the speaker presents 'G' as an epistemic alternative for 'F'.
Metalinguistic negation (Mlneg)
Not all negations of the type 'Not-F, (but} G' are examples of Choiceneg. Consider, for instance, the following sentences
(20) It is not possible, it is necessary that Havel is right (2 1 } Gorbachev did not solve some problems, he solved all of them (22) Ceaucescu did not pass away, he was killed Van der Sandt's comment18 on sentences of this type is:
On the standard translation all these sentences amount to plain contradiction. And since the implicarures denied do uncontroversially not contribute to the semantic content and thus cannot be sensitive to logical operators, it is difficult to see how the prediction of contra dictoriness can be avoided. The problem is of course that none of the above sentences is con tradictory. They are all easily interpretable as a refusal to accept a previous utterance because of the implicarures invoked.19 Hom calls these negations 'metalinguistic negations' (from now on: Mlneg). This kind of negation is 'not reducible to a truth functional one place connective with the familiar truth table for negation, nor is it definable as a separate logical operator; it represents, rather, a metalinguistic device for registering objection to a previous utterance (not proposition} on any grounds whatever, including the way it was pronounced'.20 Jacobs knows this type of sentences, too.21 Choiceneg and Mlneg are in his terminology both contrasting negations (KN). Jacobs uses the operators KORR (for 'korrekt'} and AD (for 'adaquat'} in his characterization of KN. All these negations have as their general formula22
(23) - KORRa & E1 �1 .E1 �n KORRallt...lln •
•
The operator 'KORR' (an abbreviation of 'korrekt'} is defined as follows:
(24) KORR a = [ a & AD a ) .
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
3·3
Rob T. P. Wiehe
1 13
'ADa' is true 'genau dann . . . , wenn a im angedeutenden (und dann nariirlich zu explizierenden) Sinn inhaltlich adaquat ist'. The first conjunct of (23) is true if .{\!)a is true, but also if - a is t;r!le.23 jacobs admits that the left parts of sentences like (20) , (21), (22) can be represented by '- ADa', but he adds 'Damit ergibt sich allerdings noch keine allgemeine logische Repriisencation fiir KN, denn diese Negationsart kann ja durchaus auch wahrheitsfimktional gebraucht werden, wie z.B in (4.54d)'.24
Nicht Lulse bewunden Peter, sondern Gerda
(25)
- CORR{I-F) & G
Instead ofJacobs (24), I suggest (24 ) CORR(I-F) = F & AD(I-F) '
Sometimes, as in (2o) and (21), it is known that the following sentence is also true
From (25) and (25 '), the next sentence follows in classical two-valued logic: (25 ) (F & G) & - AD(I-F) '
So not only (25) but al�o (25 •') holds for (20) and (2 1).
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
For this reason, Jacobs takes (23) as the general formula for KN. l am not sure that there is such a general logical representation of KN as Jacobs suggests. In my opinion (4.54d) is not about correctness at all, but about truth. If we disregard the internal structure of (4.54d), this sentence is of the type '- F & G'. Horn says that Dummett 'is on the right track in characterizing this use of negation (Mlneg, RW) as "a means of expressing an unwillingness to assen 'A', without necessarily constituting a willingness to deny 'A'. But 'Dum men's neo-Fregean representations, utilizing scope distinctions to account for the difference between the two ways in which negation can be understood, may not be sufficiently general or generalisable. While "(1-(not A)" may be unobjectionable for descriptive (propositional) negation, it is not dear that a representation like "not (1-A)" can be interpreted coherently for all cases in this chapter'.25 Pragmatical adequacy and (pragmatical) correctness are not predicates of sentences, but predicates of statements. For this reason, I prefer 'ADI-F' above 'AD F' and 'KORRI-F' above 'KORR F'. As a formula for (20), (21) or (22) ), I suggest, combining ideas ofJacobs and Dummett ('CORR' corresponds with Jacobs' 'KORR'· operator):
1 14 External and verb phrase negations in actual dialogues As a rule for Mlneg, I
suggest:
R3 A negative statement 'Not F, (but) G' of a participant p (p ) in a dialogue D is a Mlneg, with the symbolization '- (CORR(rF} & G' if and only if 1 . The statement is a reaction of p (p ) upon an earlier statement 'P of the other participant p (p ); and 2. The s�tement is not a Choiceneg. 1
1
1
The symbolization of (2o) is (disregarding tense)26 (26) - CORRrPoss (Ee) [(righteo) (e) & Agent(Haveleo, e) ] & Nee (Ee) [(righteo)(e) & Agent(Haveleo, e) )
( I I a) The king of Romania is not rich
The external Russelian symbolization of ( I I a) is, as everybody knows
( I 1 a ) - (Ex) (Kx & ( 'Vy) (Ky - y=x) & Bx) 1
The ordinary symbolization of (ub) Romania does not have a king lS
(ub i ) - (Ex) (Kx) It is a well-known fact that in a classical, two-valued logic ( I 1 a ) follows from ( I I b ) So it is not true in this case that 'the descriptive reading self-destructs'. Consequently, there is no need to call (ua) a metalinguistic negation. In my opinion, (na) is an Excluneg (see Section 3.I). 1
1
.
4 VE RB PHRASE NEGAT I O N (VP neg) Consider the following start of a dialogue p
I
p Kohl doesn't negotiate
(7)
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Horn makes a distinction between Truth functional (descriptive, logical) and Metalinguistic (non-descriptive, non-logical) negation. About the relation between these two kinds of negation, he makes the following true remark: 'there is a procedural sense in which the descriptive use of negation is primary: the nonlogical metalinguistic understanding is typically available only on the "second pass", when the descriptive reading self-destructs.'27 Because of this remark, I fmd it a bit surprising that Horn considers sentences of the following type Metalinguistic negations.28
Rob T. P. Wiehe
II5
The only function of (7) in this case is the conveying ofnew information. As the interpretation of (7) in this situation I choose the following formulation. (27) (Ee) [- {During(e . now>�<) & (negotiate>�<Xe) } & Agent(Kohl>�< , e) ] This symbolization expresses the fact that (7) gives new negative information about Kohl. In (27) only the (linguistic) predicate is negated. Therefore (27) is called verb phrase negation (VPneg). A disadvantage of these symbolizations of (7) is that (z8) does not follow directly from (27) and (16). To make sure that (18) follows from (27) and (16), I introduce the next meaning postulate for VPneg.
MPI states that an Excluneg is materially implicated by a VPneg. The opposite does not hold. My rule for VPneg is: R4 A negative statement '1t>�< doesn't (didn't do, shall not do) ()>�<• of a participant
p (p ') in a dialogue D is a Verb Phrase Negation, with the symbolization (Ee) [ - {P(e, t) & (6°Xe) } & Agent( 1t0, e) ] , if and only if the statement is not a reaction of p (p ') upon a statement '1t0 does (did, shall do) 6°' of the other participant p ' (p).
5
S O ME SPECIAL K I N D S O F NEGAT I O N S 5.1
Sentences with accentuation and cleft sentences
Carlson says 'that a sentence with special emphasis will always be construed as a to some related move, either explicit or assumed.29 He continues: 'This seems the right generalization intuitively. Focused sentences are curious discussion openers; they indicate that some implicit assumptions were enter tained by the speaker which motivate the emphasis.'3° Carlson's conclusion that 'no sentence with focussing emphasis can be an initial move of a well formed dialogue'31 seems correct to me. According to Carlson, the following sentence, with accentuation on 'Annie' response
(28) Annie does not want tea 'has a wide range of possible shades of meanings. When an appropriate context is provided, it can be understood at least in any of the following ways'.32 (29a) It's not Annie who wants tea, bur somebody else, if anybody.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
MP 1 (Ee) [-{P(e, t) & (6°Xe) } & Agent( 1t >�< ,e ) ] - (Ee) [P(e, t) & (d>�<)(e) & Agent( 1t0 ,e ) ]
1 16
External and verb phrase negations in actual dialogues
(29b) Annie, for one, is not one who wants tea. (Joa) It's Annie who does not want tea, not someone else (Job) At least Annie is one who does not want tea Appropriate contexts for (29) and (Jo) to arise are, in the same order,�� (Jia) (J 1 b) (32a) (32b)
Annie wants tea Everyone wants tea Jane does not want tea Who does not want tea?
(28 ' ) Annie does not want tea has two interpretations. These are (3 3a) - (Ee) [During(e, now"') & (want"'Xe) & Object(tea"', e) & Agent(Annie"', e) ] (3 3b) (Ee) [-(During(e, now"') & (want"')(e) & Object(tea"', e)} & Agent(Annie"', e) ] (J Ja) is an Extneg and (3 3b) is a VPneg. (28) has in my opinion the two same interpretations. {3 3a) is the symbolization of the left part of (29a) and (29b); (J Jb) is the symbolization of the left part of (Joa) and (Job). The following cleft sentence {J4) It's Annie who wants tea has two negations (29) It's not Annie who wants tealAnnie is not one who wants tea. and (Jo) It's Annie who does not want tealAnnie is one who does not want tea With these two negations correspond, as we saw, the symbolizations (3 3a) (wide scope) and (3 3b) (narrow scope). Consider also the cleft sentence (J 5 ) It's tea what Annie wants
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The difference between (29) and (Jo) is 'patently a matter of scope': the negation belongs to the main clause of (29a) and (29b), while it is in a sub ordinate clause in (Joa) and (Job).�4 This scope ambiguity explains the ambigu ity of (28) in the two contexts (J I ) and (32).�5 From the previous sections we know that the following (unaccentuated) sentence
Rob T. P. Wiehe
I I7
The two negations of this sentence are {36) It's not tea what Annie wants (37) What Annie doesn't want is tea {36} has the same symbolization as (29); (37) has the same formal interpretation as (30). The differences between (29) and {36), or between {30) and {37), are pragmatical ones. (30) and (37) can be seen as plausible answers to the following questions, in that order (32b) Who does not want tea? (38) What doesn't Annie want? Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The following combinations however seem counter-intuitive: {39) p: Who doesn't want tea? p ' : What Annie doesn't want is tea. (40) p: What doesn't Annie want? p ' : It's Annie who doesn't want tea 5 .2
Morphologically incorporated negations
Seuren gives the following characterization of morphologically incorporated negation: 'Many lexical items contain a morphologically incorporated negation, expressed trough a bound morpheme, usually a prefix: un-, in-, a-, dis-, etc., as in uncooperative, indirect, asymmetrical, discontinu ous.'36
Sentences with morphologically incorporated negations do not seem to present any difficulties. As a symbolization of the next sentence with a morphologically incorporated negation (and of other sentences of this kind) (41) Kohl negotiates untirable I suggest (42) {Ee) [During(e, now0) & (negotiate0Xe) & untirable0(e) & Agent{Kohl0, e)] Morphologically incorporated negations can be introduced as a part of our lexicoiL Sentences like (4 1) are translated in the same way as sentences without negation elements.
.,
I I8
External and verb phrase negations in actual dialogues
5·3
Negative polarity items
Sentences with negative polarity items are discussed in detail by Seuren and Zwarts. What I shall attempt now is to show, with a few examples, that at least some of these sentences can be handled by event logic. The sentence37 . {43) Viktor Kortchnoi doesn't live in Russia anymore
(44) Viktor Kortchnoi lives (does live) in Russia any more A plausible interpretation of (44) seems to be {45) Viktor Kortchnoi lived in Russia, but doesn't live in Russia The symbolization of (4S) is (46) {Ee) [Before(e, now*) & (live*)(e) & in"'(Russia*, e) & Agent{Viktor Kortchnoi*, e ) ] & - {Ee) [During(e, now*) & (live*)(e) & in*(Russia*, e) & Agent{Viktor Kortchnoi*, e) ] The following sentence follows from (46) {47)
(Ee) [Before(e, now*) & -During(e, now*) & (live*Xe) & in°{Russia*, e) & Agent{Viktor Kortchnoi*, e) ]
So (47) seems in the end a good translation of (43). The part 'Before(e . now*) & - During(e, now*)' can be considered as the translation of the words 'not anymore' in (43). Another sentence with a negative polarity item is38 {48) Modrow doesn't mind Honnecker's lying Just like (43), {48) cannot be interpreted as the (external) negation of another sentence. So the symbolization (49)
- (Ee) [During(e . now*) & (mind0Xe) & (Ee ) [(ly0Xe ) & Agent{Honnecker0 , e ) & Object{e ' , e)] & '
'
'
Agent(Modrow*, e) ]
is not to be considered as a symbolization of {48). As a translation of {48), I suggest Qeaving out the time operator)
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
contains the negative polarity item 'anymore'. A characteristic of sentences of this kind is that, if they have a normal intonation, they cannot be interpreted as negations of other sentences. (43), for example, cannot be interpreted as a negation of
Rob T. P. Wiehe 1 19
( so) (Ee) [ - ((mind•Xe) & (Ee ) [(ly•Xe ) & Agen�Honnecker0, e ') & Object(e ' , e)] ) & Agen�Modrow0, e) ] '
5 ·4
'
Sentences without a literal meaning
Consider the sentence from the Dutch language (5 1) Jan is geen licht
This sentence does not have its literal meaning. The meaning of (51) is roughly as follows: Sentences of this type will not be analysed in this paper.
6 Q UANTI F I CATI O N AND NEGAT I O N Consider the sentence (5 3) Some politicians do not admire Lenin, but Trotzky If we use the ordinary way of symbolizing sentences from the predicate calculus, a plausible translation of (53) is (5 4) (Ex) [politician°(x) & - adrnire*(x,Lenin*) & adrnire*(x,Trotzky*) ] The symbolizations from the ordinary predicate calculus seem perfect for sentences of this type. Because of my earlier choice of event logic, I have to keep to this choice for the translation of ( (5 3). As a translation I suggest: (5 5) (Ee)(Ex e (politicianj)[During(e, now*) & (adrnire*Xe) & -Obj ect(Lenin*, e) & Obj ect(Trotzky*, e) & Agent(x, e) ] The expression '(Exe(politicianj)' has to be read as follows: 'There individual x, that is a member of the set of politicians.' As a translation of the sentence (56) Bush doesn't read all newspapers, but some newspapers I suggest (57) (Ee)[During(e, now0) & (read0Xe) & - ('lx e (newspaperj) Obj �x, e) & (Ex e (newspaperj) Obj�x, e) & Agent(Bush0, e) ]
is
an
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(52) John is a rather stupid fellow
120
External and verb phrase negations in actual dialogues
�sensnenegl:orpottaenentioccnresesasted egative
with with
no Figure I
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
ems
polarity
Rob T. P. Wiehe
I2I
' Kohl doesn't walk/ no
>------ VPneg
(Ee) [- [ During (e, now•) & <Walk"> (el & Agent (Kohl•, e))
}
- (Eel [During (e, now•) & <walk•> (e) & Agent (Kohl•, e))
yes
Choiceneg For example, in:
Kohl doesn't walk. but run
- (Ee) [During (e, now•) & <Walk•> (e) & Agent (Kohl•, e)) (Ee) [During(e , now•) & (e ) & Agent (Kohl•, e)) or: (Eel [During (e, now•) & - <walk•> (e) & (e) & Agent (Kohl•, e))
no
MLneg For example, in:
Kohl doesn't walk. but talk
- CORR (Ee) [During (e, now•) & <walk•> (e) & Agent (Kohl•, e)) & lEe) [During Ie. now•) & le ) & Agent IKohi•.• e )J
Figure 2
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
no
>------ Excluneg
122 External and verb phrase negations in actual dialogues
From these examples we see that quantificated sentences, too, can be translated in the symbolic language of event logic.
7 C O NCLUS I O NS The flowchart depicted in Figure 1 can be helpful in discovering which kind of negation a certain sentence/statement is, while Figure 3 gives a fairly complete classification of the different kinds of negation. NEGATIONS
- sentences with
� �
Extneg
NPI - sentences with morphologically incorporated negations. - sentences that do
VPneg
Only non-atomic
Atomic and non
sentences
atomic sentences
1---..
Choice Neg
Mlneg
not have their literal meaning ('Jan is geen Iicht').
I
Excluneg
Figure 3
The main differences between the Mlneg, Choiceneg, Excluneg and VPneg are listed in Figure 4· From the examples we saw the Mlneg always consists from at least two atomic sentences. The same holds for Choiceneg. Only Exduneg and VPneg can be atomic sentences. VPneg is not a correction of a previous utterance, and a fortiori not a sentence that is contradicting a previous utterance. Mlneg is a correction, but not a contradiction, of a previous utterance. I can hardly imagine that (20) (21 ) or (22) are used as opening sentences of a dialogue. Choiceneg and Excluneg are contradicting a previous utterance, and are a fortiori corrections of a previous utterance. The only negation that does not give or suggest new information is Exduneg. VPneg is the only negation in which the negation operator may not be placed in front. In Choiceneg we may place the negation operator in front. ,
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
sentences in which the analysis depends only on the form of the sentence.
Sentences in which the analysis depends on form of the sentence and on the context of the uttering of the sentence
Rob T. P. Wiehe 123
'\�
an atomic sentence
E X
t n e
g
Ml neg
Choice neg Excluneg
VP-
Contradicts previous )Jtterance
Gives or suggests new information
Negation operator has to be placed in front
no
yes
no
yes
yes
no
yes
yes
yes
yes/no
yes
yes
yes
no
yes
yes
no
no
yes
no
Figure 4
However, we can deduct from a Choiceneg another sentence that can be considered also as a Choiceneg and that does not have a negation sign in front The negation operator has to be placed in front in an Exduneg and in an Mlneg. There is a formal difference between Excluneg and VPneg. A VPneg is of the type (Ee)[ - {P(e, t) & (d*)(e)) & Agen�7t0, e) ] , an Exduneg of the type - (Ee)[P(e, t) & (d*)(e) & Agen�7t0, e )]. According to the meaning postulate (MP,), Excluneg follows from VPneg. The opposite does not hold. The formal difference between Exduneg and VPneg has to be analysed as a difference of scope of the negation operator. The most negative sentences (excluding the negations in Section s ) are ambiguous, and can be interpreted as an Exmeg or as a VPneg. If the function of the corresponding statement is that of correcting another sentence, it is an Extneg; otherwise it is a VPneg. ROB T. P. WICHE J. S. Bachstraat 20 3 5 3 3 XC Utrecht Netherlands
N O TES 1
Compare the following short dialogue: p: Mitterand walks p ' : And he doesn't have his shoes on p ' gives new information about Mitre rand and his utterance doesn't contradict or correct p's utterance.
2 And not only in recent theories. See Hom (1989: 140, 141). 3 Gabbay & Moavcsik. (1978: 251). 4 Martin (1981a: 2). 5 Parsons (1985: 23 5). 6 ibid.: 236. 7 See Parsons, l.c.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
neg
Corrects a previous utterance
I24 External and verb phrase negations in actual dialogues
....
·
I9 Van der Sandt (1988). My examples are based on those of Van der Sandt 20 Hom (I98 5 : I21). 2I Jacobs (I982: 309). 22 ibid.: 3 I 3. 23 ibid.: 3 I 3· 24 ibid.: 3 I 3 , 301. 25 Hom (I989: 42I). 26 'Poss' and 'Nee' are short for 'possible' and 'necessary'. 27 Hom (I989: 443 , 444). 28 ibid.: 444- Hom's own example is (what else should one expect?): The king of France is not bald. As an example of descriptive negation, he ofers: The queen of England is not bald. 29 Carlson (I984: 308). 30 I.e. 3 I Carlson: 309. 32 Carlson: 321. 3 3 I.e. 34 I.e. 35 I.e. 36 Seuren (I98 5: 230). 37 Seuren: 235. 38 Seuren: 232.
RE FE RE N C E S and pragmatic ambiguity', Language, 61: Barth, E. M & E. C. W. Krabbe (I982), From 12I-74· Axiom to Dialogue, de Gruyter, Berlin. Barth, E. M & R T. P. Wiehe (I986), Hom, L. R (I989), A Natural History of Negation , Chicago UP, Chicago. Problems, Functions and Semantic Roles, de Jacobs, J. (I982), Syntax und Semantik der Gruyter, Berlin. Negation im Deutschen , Fink, Munich. Carlson, L. (I984), 'Focus and dialogue games', in L. Vaina and J. Hintikka (eds), Krabbe, E. C. W. (I985), 'Argumentatie in formele discussies', in W. K.. B. Koning Cognitive Constraints on Communication , Reidel, Dordrecht (ed.), Taalbeheersing in theorie en praktijk , Foris, Dordrecht, I 2o-8. Gabbay, D. M & J. M Moravcsik (I978), 'Negation and denial', in F. Guenthner & Mannoury, G. (I934). 'Die signifi.schen Grundlagen der Mathematik', Erkenntnis C. Rohrer (eds), Studies in formal semantics, North-Holland, Amsterdam. 4: 288, 309 and 3 I 7-45· Hoepelman, J. Ph. (I979), 'Negation and Mannoury, G. (I943), 'La question vitale: "A denial in Montague grammar', Theoretical ou B" ', Nieuw Archief Wiskunde, a i : Linguistics , 6: I9I-209. 16o-7. Hom, L. R (1985). 'Metalinguistic negation Martin, R M (I98u), 'On Logico linguistics:
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
8 Carlson (1984: 301). According to Carl son, a dialogue is represented by a 'two column list of sentences, each side listing sentences successively uttered or written down by one of the dialogue participants'. In addition, each player has a private list, not seen by the other player. In this list 'again on one side, are entered the player's own assumptions at each stage of the game. On the opposite side, the player enters assumptions that he takes the other player to make or have made in this game' (3oo). 9 ibid.: 301. IO Carlson, I.e. 1 1 See Barth & Krabbe (1982: 32). The notion 'conflict of avowed opinions' is important in this book. I2 ibid.: 171. 1 3 Mannoury (I943: 3 3 3). I4 See Krabbe (1985: 1 24) and Wiehe (I988: I89)· I 5 See the previous note. The operator ' ' has to be read as 'If . . . , then'. I6 Barth & Wiehe (I986: I74). I7 See Section 1. I8 Jacobs (I982: 308) makes a similar com ment
Rob T. P. Wiehe 125 structure and transformation', Logico Linguistic Papers, Foris, Dordrecht, I-I 9· Martin, R M (198Ib), 'On the analysis of action sentences', Logico Linguistic Papers, Foris, Dordrecht, 1 55-69. Parsons, T. (1985), 'Underlying events in the logical analysis of English', in E. Le Pore & B. P. McLaughlin (eds), Actions and Events, Blackwell, Oxford, 23 5-67. Seuren , P. A. M (1985), Discourse Semantics , Blackwell, Oxford.
Van der Sandt, R A. (1988), 'Discourse systems and echo quotation', Nijmegen (internal report), I-33Wiche, R T. P. (1988), 'Over het onderscheid russen argumentatieve en formele discussies', Tijdschrift voor taalbeheersing, 10: lli- 3 I .
Zwarts, F. (1986), 'Categoriale Grammatica en Algebraische Sexnantiek', Ph.D. thesis, University of Groningen.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Journal oj&mantics 8: 127-147
© N.l.S. Foundation (1991)
Topic/Comment, Presupposition, Logical Form and Focus Stress Implicatures: Tl?-e Case of Focal Particles only and also JAY DAVID ATLAS Pomona College, Claremont and The Institutefor Advanced Studies, Princeton Abstract
PETE R GEA C H ' S EXC L U S I O N A N ALYS I S A N D I T S DEFECTS Sherwood's semantic intuition was straightforward. The statement Only Socrates is running [R (ONLY s)] seems to assert Socrates is running and no one other than Socrates is running [R(s) & ..., (� )(x�s & Rx) ]. Thus Only Socrates is running
E
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
In Chapter I 2 of the thirteenth-century Oxford logician William of Sherwood's Treatise on Syncategorematic Words ( Syncategoremata) , Sherwood discusses the word only ( tantum) , which in the example Only Socrates is running indicates, according to Sherwood, 'how much of the subject is under the predicate-viz. that the subject Socrates and no more is under it. In that case it is an exclusive word' (Sherwood 1968: 8 I ). In Chapter 7 of the twentieth-century English logician Peter Geach's (I 9621! 980) Reference and Generality , Geach discusses the words only and alone , remarking that medieval logicians 'were gready interested in exclusive proposi tions, but their treatment of them was on the whole superficial. This comes out in their having generally accepted the idea that exclusive propositions were exponible as conjunctions �socrates alone is wise", say, as �socrates is wise and nobody besides (other than) Socrates is wise" . . . If the force of che exclusive proposition is to exclude everything other than what is named in or by the subject-term from �sharing in the predicate", that is no reason for reading in an implication that something named by the subject-term does "share in the predicate" ' (Geach I9621I98o: 208-9). This dispute between English logicians across seven centuries has been echoed in recent and influential work by the Anglo-American philosopher H Paul Grice and by linguists, notably Laurence Hom in his (I 969) 'A Presuppositional Analysis of ONLY and EVEN', in his (I989) treatise A Natural History ofNegation , Lauri Kamunen and Stanley Peters in their (I 979) 'Conventional Implicature', and Josef Taglicht in his (I 984) book Message and Emphasis: On Focus and Scope in English . In this paper I shall argue that neither Sherwood, with his conjunction analysis of Onlyx is F, nor Geach, with his non-conjunctive analysis, nor Hom, with his presuppositional analysis, nor Taglicht, with his conjunction analysis of only and his conventional implicature analysis of also and even , have accounted for the semantic and pragmatic facts, for their analyses have failed to integrate linguistic facts about topic and focus, about entailments, and about Gricean (I975, 1989) 'implicatures'. By reconsidering their views I hope to show how a more coherent account can be achieved. In the course of this paper I will offer my own analysis, building on what I have learned from theirs and, 1 hope, improving on them.
128 The case of focal particles only and also
entails Socrates is running [R (ONLY s) II- R( s) ]. And on the usual, classical assumptions about individual constants, and the usual expectation about proper names, Socrates is running will entail Someone is running [ R( s) 11(l:x) ( Rx) ]. So, by transitivity Only Socrates is running entails Someone is running. It is to this claim that Peter Geach objects (Geach 1962!1980: 208): It is formally much more convenient to treat the exclusive proposition as having precisely the exclusive force of its supposed second component, and not to read 'F(only 8)' as implying 'F(some 8)' (i.e. in the degenerate case where 'El' is taken to be a single proper name, as implying 'F(El)') . . . 'F(only 8)' will thus be true when 'F( )' is true of nothing at all; for 'F(x) ' will then not be true for any interpretation of 'x' as a proper name, let alone its being true for some interpretation in which 'x' names something not named in or by '8'.
names something that is named in or by 'El'.
This means that 'F(only a)' is true iff ....., (l:x) (x;Fa & Fx), i.e. nothing other than a is F. So if nothing is an F, nothing other than a is F, and 'F(only a)' is true. Suppose that our domain of quantification D is finite, e.g. D = {a , b , c , . . . z} , and let us suppose that No one is running is true. Now consider the statement Only a, b, c, . . ., z are running. According to Geach's analysis, that statement is true iff no one other than a , b , c , . . . , z is running [....., (l:x) ( x;Fa & x;Fb . . . x;Fz & Rx) ]. Of course, if no one is running, Only a, b, c, . . ., z are running will be true, even though a , b , . . . , z are everyone, each of whom is not running! And that is surely an absurd assignment of truth-value to Only a, b, c, . . ., z are running. The problem does not arise for Sherwood, because on his analysis, Only a, b, c, . . ., z are running is true iff a , b , c , . . . , z are running and no one other than they is running. When not one of a , b , c , . . . , z is running, the first conjunct is false, and so Only a, b, c, . . ., z are running is false. The problem that Geach and Sherwood fail to note is one independent of the size or character of the domain of quantification. It is a linguistic fact. Consider the sentence *Only everyone is running. This is an ungrammatical sentence of English. If one states Only a, b, c, . . . are running, one 'implies' that not everyone is running, i.e. that someone is not running. One 'implies' that 'a , b , c , . . .' does not list everyone. Likewise the sentences, *Not only everyone is running, *It is not the case that only everyone is running, are ungrammatical English sentences, and likewise if one states It is not the case that only a, b, c, . . . are running, Not only a, b, c, . . . are running , one 'implies' that 'a , b , c , . . .' does not list everyone. To simplify, what is GRAMMATICALLY PRESUPPOSED by Only a is F is a is not everyone [ (l:x) ( x;Fa) ] , that, on the assumption that 'a' is non vacuous, there are at least two things in the domain of quantification D. We may put the PRESUPPOSITION this way: There is someone other than a. So far, we have:
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Geach's (1962!1980: 207) truth-condition was simply: F(only 8)' is true iff no interpretation of 'x' as a proper name makes 'F(x) ' true unless 'x'
Jay David Atlas 1 29
Only a is F GRAMMATICAL PRESUPPOSillON: There is someone other than a. ASSERTION: (i) No one other than a is F (ii) a is F (?)
2
WHAT I S ' a is F '?
A Horn� defense ofthe presupposition analysis Hom (I¢9) has provided a paradigmatic example of a presuppositional analysis. His view is this: {HORN)
Only a is F PRESUPPOSITION: a is F [Fa ] ASSERTION: No one other than a is F [....., (l:x) ( x:la & Fx) ]
He has one claim and one argument in favor of this view. The first is a claim about two negative sentences, (I) and (2) below: (I) It's not true that only Muriel votedfor Hubert. (2) Not only Muriel votedfor Hubert. These sentences, Hom (I¢9: 99) claims, are mutual paraphrases. Hom then argues that the following continuations of {I)/(2) are acceptable or unaccept able as marked: (3)
a.
. . . Lyndon did too. b. . . . Somebody else did as well, but Iforget who. c. . . . �she didn't. d. . . � The election never took place. .
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Geach and Sherwood have disagreed whether (ii) a is F belongs in the assertion, but they are agreed that (i) No one other than a is F does belong. I havejust argued that on the standard logical interpretation of the truth-conditions for (i), as given by Geach, absurd consequences can follow. It is not possible for the truth-conditions of Only a is F to be specified by (i) alone. Something else is required, contrary to Geach's view. The question is, what is it? Is it a is F ? If it is, is the proposition asserted, as Sherwood and Taglicht (I984: 87-8) believe, or 'presupposed', as Hom {I969: 99) believes, 'conventionally implicated' as Kart tunen and Peters {I979) believe, or none of these, as I believe?
I 30
The case of focal particles only and also
B
Problems with the presupposition analysis
Consider the statement Only Socrates is Socrates, a logical, but not very interesting thing to say. On Horn's analysis, we would have:
Only Socrates is Socrates. ASSERTION: No one other than Socrates is Socrates (-. (l:x) (x�s & x=s) ] PRESUPPOSITION: Socrates is Socrates [s=s]
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Now, I admit that there is something very odd about ?? Not only Muriel votedfor Hubert. The election never took place. And the explanation is that the Hrst statement presupposes that there was an election, which the second statement then denies. There is also something odd about ?Not only Muriel votedfor Hubert. She didn't. Horn believes that the explanation of the oddity is the same: the Hrst presupposes that Muriel voted for Hubert, which the second then denies. Since one may naturally infer Muriel votedforHubert from Only Muriel voiedfor Hubert as well, Horn takes Muriel votedfor Hubert to be 'presupposed' by the affirmative and negative sentences. One might Hrst consider whether the data will be as convincing for the other negative sentence: It's not true that only Muriel votedfor Hubert. The election never took place. I confess that I Hnd this much more linguistically acceptable than the 'Not only . . .' sentence. How about It's not true that only Muriel votedfor Hubert. She didn't. ? I don't Hnd that linguistically unacceptable at all. Horn explains the linguistic oddity of (3c) as a case of presupposition violation as in (3d). But the mere oddity of (3c), even in the case of the 'Not only' sentence, does not prove that it is a case of presupposition violation. There might be another explanation of the oddity of (3c). (In fact, there is, and I shall give it later in this paper.) I draw two conclusions. First, (1) and (2) are probably not mutual para phrases, contrary to Horn's intuitive claim. Second, no argument has been provided that the only explanation for the oddity of (3c) is a presupposition violation. The data simply show it to be a possibility. Nevertheless, on the basis of that claim and those data, Horn suggests his presuppositional analysis of Only a is F. Since the ASSERTORIC component alone is held by Horn (1969: 105) to determine entailment relations and truth conditions (which determine the truth-value T or lF when the presupposition is T), his suggestion for the ASSERTION in Only a is F faces the difficulty that I raised against Peter Geach. It would be absurd to take the statement Only a, b, c, . . . are running to be true in the case that I described in Section I . But the view that Only a is F has the presupposition a is F also leads to difficulties.
Jay David Atlas
I3 I
(i) I described myself truly;
(ii) what I said must be false (since I didn't regret that it was); (iii) what I said must be neither true nor false (since by (i) its presupposition is false).
From (i)-(iii) one can infer either that one cannot accept both the presupposi tional analysis of regret and the semantic concept of presupposition or that one cannot accept both the presuppositional analysis of regret and saying truly 'I don't regret that the statement I am now making is false'. If one is a presuppositionalist about verbs like regret, e.g. Hans Kamp, one
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
This analysis makes the assertion of Only Socrates is Socrates into the assertion of a Hrst-order logical truth. Likewise, the assertion of Not only Socrates is Socrates would be the assertion of a Hrst-order logical falsehood Someone who is other than Socrates is Socrates [ (l:x) (x�s & x=s)J, and its presupposition would also be Socrates is Socrates [s=sV Whatever is being asserted in Not only Socrates is Socrates, it is linguistically absurd to identify the assertion with the logical contradiction Someone is identical to Socrates and not identical to Socrates , or with the equivalent Socrates is not Socrates , it being presupposed that Socrates is Socrates. When I assert Not only Socrates is Socrates , I am not asserting Socrates is not Socrates , or asserting Something is both identical and not identical to Socrates , even if what I do assert entails that Socrates is not Socrates. Furthermore, Hom's analysis implies that Not only Socrates is Socrates would possess mutually contradictory assertoric and presuppositional components, as would, on a similar analysis, the statement The king ofFrance does not exist. That statement might be thought to presuppose The king ofFrance exists , which on Strawson's (1 950) original view would make the negative existence statement true only if not true, and so necessarily not true (rather than, as it is in fact, contingently true). So statements whose assertoric and presuppositional parts contradict each other, particularly ones that presuppose their own untruth, are very peculiar statements. Whatever the peculiarity of Not only Socrates is Socrates , its peculiarity does not seem to be one of asserting a logical contradiction while simultaneously presupposing a trivial truth that is the direct denial of what is being asserted. Another interesting case of this kind is The Paradox of the Regretter: I don't regret that the statement I am now making isfalse. That assertion would presuppose that what I am stating in the assertion is false, so it would presuppose that I do regret that the statement I was making, when I made the statement, was false. On the semantic concept of presupposition and on the presuppositional analysis of regret, if in saying 'I don't regret that the statement I am now making is false' I didn't regret that the statement I was making was false, the�:
I 32
The case offocal particles only and also
will
C The essential issue Hom's presuppositional analysis leads us to make false predictions of linguistic anomaly in sentences; the anomalies fail to materialize. We were led to consider the consequences of Only Socrates. is Socrates presupposing Socrates is Socrates because we wished to examine the question whether Only Socrates is running presupposes Socrates is running. The conclusion I draw from my discussion is that Socrates is Socrates is not presupposed by Not only Socrates is Socrates or by Only Socrates is Socrates. Likewise, I doubt that the truth of a is F is presupposed by the statement Only a is F or by Not only a is F. But if it is not presupposed, what is its status? By contrast with Hom's analysis,
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
want to maintain both the presuppositional analysis of the meaning of regret, namely that the regret statement presupposes that what is regretted is true, and the semantic conception of presuppositioiL But these are maintained at the cost of being committed to the claim that some mental descriptions are impossible to apply truly. The statement is paradoxical, i.e. contradictory, if and only if one accepts the presuppositional analysis of the meaning of regret sentences. Depending on one'sjudgement of the statement's contradictoriness, one would have linguistic evidence in favor of, or against, the presuppositional account of regret. Or one might decide just to give up the analysis in order to get rid of the paradox, an appealing solution. The puzzle of the presupposition of The king of France does not exist is a theoretical puzzle; the statement is not intuitively puzzling. The solution is to correct the theory (see Atlas 1 988). The thought that Not only Socrates is Socrates asserts a contradictory of its presupposition is also puzzling. Yet obviously nothing other than Socrates is Socrates-that is not puzzling, and obviously Socrates is Socrates-that is not puzzling. So the falsity of Not only Socrates is Socrates is not puzzling. But if it really asserts a contradictory of what it pre supposes , it should be a linguistically odd statement, e.g. ?The king ofFrance does not exist and he is bald. Yet it is not odd in that way. It just seems unoddly and logically false. On Hom's analysis Only Socrates is Socrates and Not only Socrates is Socrates both presuppose Socrates is Socrates. The former asserts the logical truth No one other than Socrates is Socrates ; the latter asserts the contradiction Something is Socrates and not Socrates. Hom's analysis predicts a linguistic anomaly in Not only Socrates is Socrates that fails to materialize: asserting a logical equivalent of Socrates is not Socrates and presupposing Socrates is Socrates. Whatever is peculiar about Not only Socrates is Socrates , that is not it.
Jay David Atlas
{HORN)
133
Only a is F PRESUPPOSITION: a is F ASSERTION: No one other than a is F.
with Geach's analysis, {GEACH)
Only a is F ASSERTION: No one other than a is F.
with Taglicht's and William of Sherwood's analysis, {TAGLICHT/ SHERWOOD)
my analysis, so far, is {ATLAS)
Only a is F GRAMMATICAL PRESUPPOSITION: There is someone other than a . ASSERTION: (i) No one other than a is F QQ ????????????????????
The fundamental issue facing the analysis at this point is the status of a is F. It is a significant feature of the analyses of Geach and Hom that Only a is F does not semantically entail a is F, while it is a feature of Sherwood's and Taglicht's analyses that Only a is F does entail a is F. In a recent reconsideration of the problem, Hom (1989: 248) takes the 'essential issue' to be this: 'Is only a negative in meaning and positive only by presupposition or implicature, or does it abbreviate a conjunction (only a = " a and nothing (other/more} than a )?' The intuition behind Hom's question is, as we have seen, that in, e.g. Only Muriel votedfor Hubert , the presupposition is the positive Muriel votedfor Hubert , while the assertion is the negative No one other than Muriel votedfor Hubert. On my view the first mistake was the claim that the presupposition is Muriel votedfor Hubert. Hom's arguments do not demonstrate that it is. The second mistake has been in the understanding of the logical form for No one other than Muriel votedfor Hubert. Geach's and Hom's logical form for No oneotherthan Muriel votedforHubert is ....., ('l:.x) (x¥m & V(x ,h) ) . This logical form is logically equivalent to {Ilx) ( V(x ,h) - x=m J ) . This means that at most Muriel voted for Hubert. If no one voted for Hubert, it is still true that at most Muriel voted for Hubert. The logical form is vacuously true, by reason of the truth of --o('l:.x) ( V (x ,h) ) , and so the ASSERTION part of OnlyMuriel votedfor Hubert would be true even if no one voted for Hubert. What would be false would be Hom's proposed presupposition that Muriel voted for Hubert. {How can the presupposition be false and yet the assertion true ? Hom {1¢9) would "
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Only a is F ASSERTION: a is F & no one other than a is F.
1 34
The case
of focal particles only and also
have to give up his logical analysis of the ASSERTION or give up the semantic concept of presupposition.)
3
THE PROBLEM O F ASSERT I O N AND LOGI C AL FORM: T O P I C AND F O C U S
-
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
It seems to me that when one asserts 'Only Muriel voted for Hubert', or its negatives 'It's not true that only Muriel voted for Hubert' I 'Not only Muriel voted for Hubert', one is not presupposing, in the proper sense of that word, that Muriel voted for Hubert. Thus it is that Taglicht (1984: 88) noticed that the example below is 'hard to reconcile with the claim that the positive proposition in only sentences is not part of the assertion'. From this observation Taglicht infers that what is asserted may be represented by the conjunction of Hom's presupposition and assertion. Taglicht observes that asserting Each ofthem stuck to one drink; Bill drank only whiskey and Harry drank only beer would be to 'assert' that Bill drank whiskey as well as that Bill drank nothing but whiskey. Surely there is something correct about Taglicht's observation. His example strongly suggests that Bill drank whiskey is not just presupposed. Whether he is right to say that Bill drank whiskey is 'asserted' is another question, one to which I shall return. There is acrually a fundamental theoretical puzzle raised by Hom's claim that a is F is presupposed by the statement Not only a is F. When we state The queen ofEngland raises the best race horses , we state something different from There is a unique queen ofEngland and she raises the best race horses. Strawson (1950) led us to think that we do not assert the existence of the English queen when we state The queen ofEngland raises the best race horses. In fact, we rarely feel the need to assert sentences like 'There is a queen of England and she raises the best race horses'. The purposes served by direct assertion are better served without it presupposition is adequate. But we do find people saying things like :Jane, and only Jane, got to the party on time'. Yet on Horn's analysis of OnlyJanegot to the party on time , whose presupposition is allegedly Jane got to the party on time, it should be just as unusual to find people stating Jane, and onlyJane, got to theparty on time as it is to find them saying There is a unique queen ofEngland and she raises the best race horses. But that is just an incorrect linguistic prediction. Statements likeJane, and onlyJane, got to the party on time are not unusual. The conclusion of this argument is, once again, that Only a is F does not presuppose a is F. But does it, in the sense of'assert' currently under discussion, assert that a is F as Taglicht (1984) claims? This is, as Hom (1989: 248) remarks, 'an essential question'. But how, in the face of so many conflicting linguistic intuitions and theoretical claims, shall we go about finding a satisfactory answer?
Jay David Atlas
135
(Focal NP Limitation) Focal Noun Phrases in natural language statements cannot be translated by singular terms that are logical subjects in the Semantic Representation of their truth conditions (e.g. Russell-Tarski logical forms). Since only is a focus marker, and the focus NP in Only a is F is 'a', the Focus Limitation Principle has an interesting consequence. William of Sherwood and Josef Taglicht (1984), among others, have taken the logical form of Only a is F to be a is F and no one other than a is F (Fa & {Ilx) [Fx - x = a] ) , in which 'a ' is a logical subject in the first conjunct. But according to my principle, Only a is F cannot have a logical form in which 'a ' is a logical subject Thus the cor� unc tion above is not the logical form of Only a is F ! (I) The Focal Noun Phrase Limitation Principle entails that the conjunction a is F & no one other than a is F is not the logical form of Only a is F.
Since 'a' is a Focal NP in Only a is F, it is not a Topical NP in the statement, as we shall see below. Does that fact about Topic have any consequences for our analysis of Only a isF? In my recent discussions {Atlas 1988, 1 989) of the Topics
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Some years ago, following the lead o f Scott, Davidson, and Harman, and reacting to discussions with Gerald Gazdar and Stephen C. Levinson, I asked myself what the logical form of a cleft sentence (in English) was. Gazdar (1979: 124-5) had proposed one. For It is Sam who wants Fido he had suggested A.x( Wants( x ,Fido) ) (Sam) . Since the 'focus' of the cleft is Sam , and the 'pre supposition' is Someone wants Fido , in Gazdar's logical form the 'focus' corres ponded to the logical subject, the 'presupposition' to the logical predicate. He took his view of clefts with two-place relation symbols because he assumed that in clefts with one-place relation symbols, e.g. It wasJohn who went , the logical structure was similar to the simple subject-predicate John went. He then assumed that the logical structure of It is Sam who wants Fido was similar to that of Sam wants Fido. To a philosophical logician concerned with semantics, and not so much swayed by syntax, this assumption seemed very odd: the pre supposition of the cleft was Someone wants Fido , which was not a presupposition of Sam wants Fido ; it was instead a presupposition of the contrastively stressed SAM wants Fido (compare SAM doesn't want Fido , or Does SAM want Fido?). So I argued that It wasJohn who went paired with the contrastively stressed JOHN went , not with John went, just as I had argued, as had Halvorsen {1978) independently, that It was himselfthatJohn wanted Mary to describe pairs withJohn wanted Mary to describe HIMSELF , not with john wanted Mary to describe himself. So I concluded that Gazdar's identification of logical subject and 'focus' in It is Sam who wants Fido was profoundly mistaken. This was my first formulation of my Focal Noun Phrase Limitation Principle:
I 36
The case of focal particles only and also
of negative existence statements, e.g. Pegasus does not exist, I appealed to work by Gundel (1977) in making the following observations, where 'NP' is a meta variable ranging over proper names and simplex deHnite descriptions (not e.g. 'the F of the G'):
By these criteria the focal NP 'a • in Only a is F is not a topical NP, so that, on the basis of the Linguistic Aboutness Principle: {ABOUT) A statement is about a only if'a' is a Topical Noun Phrase in the statement. the statement Only a is F is not about a. (II) The statement Only a is F is not about a. For the conjunction Fa & cJ> , it is intuitive to say that the conjunction is about a and about the topic(s) of 4>, since the individual conjuncts are about a and about the topic(s) of cJ> respectively. So if the logical form of Only a isF were a is F and no one other than a is F [Fa & (IIx) ( Fx -+ x - a)] , what would it be about? A reasonable answer would be: it is about a and those who F. But by {II), Only a is F is not about a . So, on the basis of the Linguistic Aboutness Principle and of my Criteria for Noun Phrase Topicality, we see, again, that the logical form of Only a is F cannot be Sherwood's and Taglicht's conjunction. {III) The Linguistic Aboutness Principle and the Criteria for Noun Phrase Topicality entail that a is F & no one other than a is F is not the logical form of Only a is F.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(CRITERIA) Criteria for Noun Phrase Topicality ( x ) Topical NPs are not abnormally stressed. (2) Leftmost NPs (in surface forms) are not necessarily topical NPs. (3) If an NP is a topical NP in' a statement (not sentence), the statement is a linguistically acceptable answer to the question What about NP? (4) If an NP is a topical NP in a statement, the asfor NP trans form of the statement is linguistically acceptable (in the same contexts) and makes the same statement. (For example, johnny deceived thegirl '* Asforjohnny, he deceived thegirl.) ( s ) The Grice-Strawson (1954, 1¢4) Condition Statements carry presuppositions of the existence of NP designations only ifthe NPs are topical. (For example, a statementJohnny deceived thegirl carries a presupposition that the designation of johnny exists, but a statement of It wasjohnny who deceived the girl does not.)
Jay David Atlas
i 37
Moreover, the Grice-Strawson Condition shows that Hom's pre suppositional analysis of Only a is F cannot be correct. That analysis calls for Only a is F to presuppose a is F. According to the Grice-Strawson Condition, since 'a' is not a Topical NP in Only a is F, Only a is F cannot presuppose that a exists. And how could the statement presuppose a is F and not presuppose a exists? It could not; so, the statement does not presuppose that a is F, a conclu sion that I also argued for in Section 2 C.
{N} The Grice-Strawson Condition entails that Only a is F does not presuppose a is F.
{l:x) (Brian knew that Mary kissed x) , which entails (under the usual rationality assumptions) Brian knew {l:x} (Mary kissed x) , which in turn entails {l:x) (Mary kissed x) . Thus the cleft statement does not presuppose Mary kissedjohn after all! At most it presupposes Mary kissed someone , and there is no conflict with the reasoning that we have sketched.)
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(There are some subtleties in the argument just given that deserve mention in passing. The intuition to which I appealed in my question 'And how could the statement presuppose a is F and not presuppose a exists ?' is as fundamental a semantic intuition as one can appeal to, but like Bertrand Russell in similar circumstances, I feel the need to try to explain it by more theoretical notions, even if the explanation turns out to be more complicated than what is being explained. I want to hold something like this: Since a is F 'directly entails' a exists , in the sense that it is an entailment of (roughly speaking) a subformula [(l:x} (x=a & Fx) 11- {l:x) {x=a) ) , and presuppositions are preserved under 'direct entailment', since those 'direct entailments' represent part of the information being presupposed, if a exists is not presupposed, neither is a is F. Cleft sentences present an interesting test case. In the statement Brian knew that Mary kissedjohn , Mary kissedJohn is presupposed. And so is John exists. On the reasoning just sketched, ifJohn exists were not presupposed, neither would Mary kissedjohn be presupposed. But in the cleft It wasJohn that Brian knew that Mary kissed , sinceJohn is not a Topical Noun Phrase (it is a focus), by the Grice Strawson Condition john exists is not presupposed by the deft statement. By the reasoning just sketched, that would mean that Mary kissedjohn would not be presupposed in the cleft statement It wasJohn that Brian knew that Mary kissed. Can that be right? Let us consider the question using only the basic properties of deft statements. The 'presupposition' of this deft is:
138
The case offocal particles only and also
4 LEXI CAL M E A N I N G , L O G I CAL F O R M A N D P R A G M AT I C I N FE R E N C E : S O LV I NG T H E P R O B L E M S
A Meaning At this point in our argument, we have eliminated as analyses of Only a is F the proposals of (GEACH), (SHERWOODtrAGLICHT), and (HORN). Never theless we still have to account for two pieces of linguistic data, the intuitive inferences:
When we left off considering my analysis, we were at this stage: (ATLAS) Only a is F GRAMMATICAL PRESUPPOSITION: There is someone other
than a. ASSERTION: (i) No one other than a is F. (ii) ? ? ? ? ? ? ? ? ???? ? ? ??????? I think that there has been omitted from consideration o f the sentence some features of lexical meaning, since the attention in philosophy of language and in formal semantics is always on truth-conditions, as the core of sentence mean ing. We know something about the meaning of 'Only a is F ' that we have so far neglected, namely that it is an analytic entailment of 'Only a is F' that exactly one person is F ! But this proposition is not just a conjunct to be conjoined with No one other than a is F. It is an analytic entailment by virtue of the lexical mean ing of only, a syncategorematic term in the sentence. This proposition is a syncategoll'ematic proposition, and it cannot be simplistically conjoined with No one other than a is F by '&' to yield a logical form for the sentence of the type lJI & <1>. Still, we do want to combine Exactly one person is F with No one other than a is F somehow. The way to do it, thanks to the brilliant invention of multiple quantification by Gotdob Frege (ca. 1 879) and Charles Sanders Peirce (ca. r 883), is the following: (LOGICAL The logical form of Only a is F = FORM) (�x) (ITy) [ (x=y = ry) & (ry - y=a) ] (Now I do not want to be misleading or tricky about this. This logical form is logically equivalent to one that just is the conjunction by '&' of Exactly one person is F (Russell's The F exists) and No one other than a is F, viz.
(�) (ITy) (x=y = ry) & (nx) (Fx - x=a ) .
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(Datum r ) Only a is F. Therefore a is F. (Datum 2) Not only a is F. Therefore, a is F.
Jay David Atlas 1 3 9
(ATLAS) Only a is F GRAMMATICAL PRESUPPOSITION: There is someone other than a. [(�x) (x�a ) ] ASSERTION: Exactly one individual, and no one other than a, is F. (�x) (ITy ) [ (x=y = Fy) & ( Fy - y =a ) ] It's not the case that only a is F GRAMMATICAL PRESUPPOSITION: There is someone other than a . ASSERTION: It's not the case that exactly one individual, and no one other than a, is F. G
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Their equivalence is a theorem of ftrst-order Quantification Theory with Identity, but they are not the same logical form. They are conceptually distinct, and important logical conceptions are involved in the proof of their equi valence, both syntactic conceptions and semantic conceptions.) From the point of view of a theory of speech-acts, in asserting Only a is F, we do not thereby assert a is F, the way we would if the statement were to consist of a conjunction a is F & tP. Rather, what we do assert entails a is F, but it does not 'say' it This feamre of my analysis preserves Aquinas's and Geach's intuition that an 'excluder' like only excludes everything other than what is named by the subject-term 'a' from 'sharing in the predicate' 'F' and need not go on to say that something named by the subject-term does 'share in the predicate' (Geach 1<}62!I98o: 208-9). I differ with Geach's claim that Only a is F need not Qogically) imply a is F. That seems to me to go too far. What we do not need to claim is that a is F occurs as a conjunct in the Semantic Representation/logical form; as we have seen, it certainly is not a conjunct in any such representation. The content carried by a is F occurs only contexmally in Only a is F, not as a logical con stiment (to put it in Bertrand Russell's fashion). My analysis solves Geach's ftnite domain problem, namely that what is ASSERTED in Only a, b, c, . . . are F is true when no one is F and a , b , c , . . . are everyone. When no one is F, my logical form for Only a, b, c, . . . are F is false. My analysis also explains the linguistically anomalous character of # # # Only a is F and a is not F by showing that it is a logical contra-diction, not merely a contra-presupposition (as in Hom's analysis). Compare the less anomalous, contra-presupposition sentences: # #All John's children are asleep and john is childless , # The king ofFrance is bald and there is no king ofFrance I he doesn't exist. Thus I explain Darum I above as an entailment. (Previous analyses, from Sherwood to Geach and Hom, restrict themselves to single quantifters in formulating the truth-conditions. Why? That is just doing semantic battle without even nineteenth-cenmry weapons.) The linguistic facts are all explained by my analysis:
140 The case of focal particles only and also
B
Pragmatic inference
2, namely why there is intuitively such a strong inference from Not only a is F to a is F. Is this an entailment, a pre supposition (Hom 1969), a 'conventional implicature', or a generalized
What remains to be explained is Datum
conversational implicature? As discussed by Hom (1972), Gazdar (1979), Atlas & Levinson (198 1 ), Levinson (I 98 3 ) , etc., scalars like all , some , many, one , have generalized conversational implicata, utterance-interpretations derived from inferences that are required in order to make the utterances satisfy Grice's
(I975, I989)
party.
The assertion of the less informative expression
negation of the more informative expression
all
some
conveys the
Conversely, the assertion of
the negation of the more informative expression conveys the information in the weaker but positive expression: asserting Not all of the boys were at the party implicates Some of the boys were at the party. These are standard, non-focal expression implicata for expressions ordered in a Hom Scale (all, some) by an entailment relation A(all) II- A(some). The case of a Levinson Infonnative ness Scale is similar, e.g. (succeed, try), but the informativeness ordering is not
a tried to open the door implicates a did not succeed , but a succeeded in opening the door does not entail a tried to open the door (Atlas & Levinson (I 98 1 : 3 8-9, ns. 8-9) ). Likewise (only, also) is a Levinson Scale: asserting [Socrates also] is running implicates Not only Socrates is running, and asserting Not only Socrates is running implicates [Socrates also] is running , but Only Socrates is running does not entail [Socrates also] is running. (Since there is JlO generalized conversational implicature: asserting Only Socrates is running implicating Not also Socrates is running, there is no Levinson Scale 0(also , only). The scale is (only, also).) Now consider some data: if I deny Only a is F, I may do so with the emphatically stressed Only a does NOTF. I can paraphrase that statement by Not [Only a is F] (the square brackets indicate scope). By contrast, Only a does NOT F is not paraphrasable by [ [Not only] a] is F. Thus Not [ Only a is F] and [ [Not only] a] is F are not intersubstitutable salva signiflcatione. As I mentioned at the outset of my discussion of Hom ( I 969) in Section 2 A, I do not agree with him that It's not true that only a is F and Not only a is F are paraphrases. Notice also that the word also is quite odd in ?? [ [Also] a] is F, though not in its function as a coordinating conjunction: Frank was kissed by her. Also Brian, and John, were kissed. What is acceptable is [a [also] ] is F, which evidently entails a is F. imposed by an entailment relation; asserting
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
First Maxim of Quantity, that is, 'Make your contribution to the conversation as informative as necessary for current purposes of the exchange'. For example, asserting Some of the boys were at the party implicates Not all ofthe boys were at the
Jay David Atlas 141
The meaning of [a also] is F, Taglicht (1984: 87, �� . 1 56, 164, suggests, is:
n.
8)
(Taglicht' ) [a also] is F CONVENTIONAL IMPLICATUM: Someone other than a is F. ASSERTION: a is F LOGICAL FORM: a is F & Someone other than a is F. while Hom (1¢9) suggests:
[a also] is F PRESUPPOSITION: Someone other than a is F. ASSERTION: a is F. A more informative word S and a less informative word W in a (HORN/ LEVINSON Hom Scale or in a Levinson Scale produce First Maxim of Quantity generalized conversational implicata from the assertion SCALES) of verbal frames A ( ) in which they occur: for (S, W), "A(W)"--> > --.A(S) "--.A ( S )" --> > A(W). (Hom ' )
(1 )
(2) (3) but NOT: and NOT:
" a also is F" --> > Not only a is F "Not only a is F" --> > a also is F "Only a is F" --> > Not also a is F, "Not also a is F" --> > Only a is F.
From observation (3) it is clear that a Levinson Scale will not predict the reinforced implicature: not ALSO a, ONLY a (e.g. John did not ALSO win the Nobel Prize in Physics; ONLY he won it). This Focal Stress implicature must be given another explanation. There are a number of contrasting expressions of this kind in English:
(FOCAL STRESS) not SOME, (but) ALL not ONE, (but) MANY not ALSO, (but) ONLY (e.g. } did not ALSO l . . m . Phystcs . ; f wm the NobeI prrze o hn ALSO did not
I
{
he
I���
ONLY he
1 1·
won it.)
On Hom's (1985) view the Focal Stress inference would allegedly be explained as follows: the first clause of a is not ALSO F; [he ONLY] is F would be a
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The Levinson Scale (only, also) yields:
142 The case of focal particles only and also
a is not ALSO F; [he ONLY] is F. I I
v v
[a ONLY] is F Unlike Hom, Taglicht takes the logical form of [ a also] is F to be the conjunction of the ASSERTION and the CONVENTIONAL IMPLICATUM. In light of the anomaly in non-sarcastic statements of ..It's not true [everyone also] is F and ..[Everyone also] is F, I believe that there is a Grammatical Presupposi tion There is someone other than a in [a also] is F. I also believe that the truth conditions of [a also] is F are given by At least two individuals, one ofwhom is a, is F. Thus I propose: (ATLAS ' ) (a also] is F GRAMMATICAL PRESUPPOSITION: There is someone other than a [ ( l:x) ( x� a) J ASSERTION: At least two individuals, one ofwhom is a, are F. [ ( l:x) ( l:y) ( x=a & y �a & Fx & Fy) ] My ASSERTION, of course, is logically equivalent to Taglicht's a is F and someone other than a is F. I prefer my logical form, again, because of my lFocal NoWl l?hlrase Limitation l?rinciple . Taglicht's logical form makes [a also] is F about a and about those who F. On my view [a also] is F, where 'a' is a Focal NP, is not about a. So I reject Taglicht's logical form. For the same reason I reject Hom's (1969: 105-6) claim that the ASSERTION of [ a also] is F is Fa ,
·
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
'metalinguistic' denial of the presupposition of [a also] is F, viz. the denial of Someone other than a is F, which would be equivalent to No one other than a is F. That proposition is what Hom (1969) takes to be ASSERTED by Only a is F. Thus, on Hom's view, asserting a is not ALSO F; [he ONLY] is F amounts to asserting No one other than a is F; no one other than a is F. I myself do not think that such redundancy is a correct prediction of the linguistic facts involved in the assertion of a is not ALSO F; [he ONLY] is F. Hom (1969) and Hom (1985) cannot both be right. I prefer another analysis, one that Hom himself could well have given on another occasion, in which asserting a is not ALSO F 'metalinguistically' denies a also is F, and so negates the (Levinson Scale) implicatum of also , i.e. negates not only, and therefore entails only (e.g. Was there also sugar in the coffee? NO-ONLY sugar in it). rhis inference is then reinforced by the continuation of the asser tion . . . [he ONLY] is F. This is a different redundancy from the one just considered; it is not the repetition of an asserted content. It is the explicit asser tion of the negation of the implicatum of what was just denied, namely of AL50 [also --> > notonly; DENIAL[ALSO) = DENIAL[not ONLY] = ONLY].
Jay David Atlas 143
C A solution To summarize: (1) Only a is F GRAMMATICAL PRESUPPOSITION: There is someone other than a [ (l:x) (x:Fa) ] ASSERTION: Exactly one individual, and no one other than a, is F. ( (l:x) ( Tiy) [ (x = y = Fy ) & (Fy - y = a)] TOPIC: Those who F (2) It is not the case that only a is F GRAMMATICAL PRESUPPOSITION: There is someone other than a ASSERTION: It is not the case that exactly one individual, and no one other than a, is F. (3) [a also) is F GRAMMATICAL PRESUPPOSITION: There is someone other than a [ (l:x) (x:Fa) ] ASSERTION: At least two individuals, one ofwhom is a, are F. [ (&) (l:y) (x=a & y�a & Fx & Fy) ) TOPIC: Those who F (4) Hom's (1969) alleged presuppositional inference: Only a is F. Therefore, a is F. is an ENTAILMENT. ·
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
which is about a. On my account [a also] is F does not assert Fa , but the statement entails Fa . We are now in a position to explain Datum 2, Hom's alleged presupposi tional inference from Not only a is F to a is F. The inference is two-staged. First, asserting Not only a is F (Levinson Scale) IMPLICATES [a also) is F. Second, [a also] is F 'DIRECTLY' ENTAIT..S Fa (see (Section 3) above). Thus the intuitive inference is explained. Finally, the independently motivated logical forms for Only a is F and a also is F conftrm, by the Popperian (Existential Quantifter) criteria for Informative ness discussed in Atlas & Levinson (198 1 : 4 1 , n. 1 3 , 42, 48), that the logical form of Only a is F is more informative than the logical form of a also is F , as would be expected on the evidence of the generalized conversational implicatures. Thus our semantic account and our pragmatic account of only and also are congruent, and our total account satisfies the desideratum of a unified account of semantic and pragmatic properties that was discussed in Atlas & Levinson (198 1 ). Only is more informative than also in Popper's logico-semantic sense, and it is more informative in Levinson's implicatural-pragmatic sense.
1 44
The case of focal particles only and also
(s ) {LEVINSON SCALE}
5
C O NCLUS I O N
These results are the product of an unusual combination of elements: (i) truth conditional semantics, (ii) Gricean implicatures, and (iii) TOPIC constraints on the structure of Semantic Representations/logical forms. Only all three elements, working coherently together, manage to give an accurate and adequate logico-linguistic explanation of the contribution FOCUS PAR TICLES like only and also make to the truth-conditions, entailments, and pragmatic implications of sentences in which they occur-a better explanation, I suggest than the previous explanations by medieval logicians, contemporary logicians, and some contemporary theoretical linguists. I hope that this case study demonstrates the merit of that unusual combination of philosophical, logical, and linguistic ideas.
Acknowledgements I am indebted to discussion with Dr Stephen C. Levinson and to comments by Dr J. de Mey, Prof. Dr Jacobs, Dr P. Bosch, and Prof. Dr F. Zwarts. I am also indebted to Dr J. Hoepelman and to J. Machate for their invirarion to address the MAFID Focus Workshop, 28-30 June 1989, Fraunhofer-Instimt fiir Arbeitswirtschaft und Organisation, Smngart, Germany. I am
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
There exists a Levinson Scale (only, also) of lexical items also and only that permits First Maxim of Quantity generalized conversational implic atures from the assertion of verbal frames A( ) in which they occur: { I } "A (also)" -> > A(not only ) (2) "A(not only )" -> > A (also) {6} The FOCAL STRESS implicature: "A(notALSO}" -> > A{ ONLY} is explained by Horn's {I98S} 'metalinguistic' denial and the Levinson Scale (only, also). {7) Horn's {I969} alleged presuppositional inference: [ [Not only] a] is F. Therefore, a is F is a two-staged inference, a scalar, generalized conversational implic' ature and a 'direct' entailment: {I} "Nor cnly a is F" ->> [a also] is F (2) [a also] is F I� a is F (8) Only is more informative than also in both Popper's logico-semantic sense and Levinson's implicatural-pragniatic sense. (Thus I have a unified account of the semantic and pragmatic properties.)
Jay David Atlas I 4 5 grateful t o my student Creighton Rosenral for assistance publication.
m
preparing this paper for
JAY DAVID ATLAS Department of Philosophy Pomona College Claremont, California 9 I 7 I I
U.SA
N OTE I To be precise about Horn's comrnirmenr to logical syntax, I would have to rreat 'is
for
=
S', and add
not
as an axiom:
Only Socrates is Socrates , and so I have bothered
with
this
syntactical
complication.
1- (Ilx) (S x = x - s).
RE FERE N C E S Atlas, J. D . ( I 988), ' What are negative exist ence statements ahour?', Linguistics and
Philosophy,
II:
3 7 1 -93.
Atlas, J. D. ( 1 989),
Philosophy without Am biguity, Clarendon Press, Oxford.
Atlas,J. D. & S. C. Levinson ( I 98 1 ), 'It-clefts ,
Horn, L. ( 1 969), 'A presuppositional analysis of
ONLY
of logical operators in English', Ph.D. Angeles.
Pragmatics: Implicature, Pre supposition, and Logical Form , Academic Press, New York.
Geach, P. ( 1 9621 I 98o),
ity, Cornell University Press, Ithaca. Grice, H. Paul (I 97S). 'Logic and conversa
tion', in P. Cole & J. L. Morgan (eds),
Syntax and Semantics J: Speech Acts , Acad
emic Press, New York, 41-58. Grice, H. Paul ( 1 989),
Studies in the Way of
Words , Harvard University Press, Cam bridge.
of
California,
Los
Horn, L. ( 1 98 5 ) . 'Metalinguistic negation and pragmatic ambiguity', 74·
,
Longuage , 6 1 : 1 2 1 -
A Natural History '![Negation ,
University of Chicago Press, Chicago.
Karttunen, L. & S. Peters ( 1 979). 'Conven tional implicature', in C.-K. Oh and D. A. Dinneen (eds),
Syntax and Semantics: Pre supposition , Academic Press, New York, 1 56. Levinson, S. C. (1983), Pragmatics , Cambridge University Press, Cambridge. Sherwood , W. ( 1 986),
Gundel, J. ( 1 977)
Linguistic
University
Hom, L. ( 1 989 )
Reference and General
Chi. Ling. Soc. ,
Horn, L. ( 1 972), 'On the semantic properties
P. Cole (ed.),
Press, New York, 1 -6 1 .
Papers from the
guistic Society, Chicago,
Diss.,
Gazdar, G . ( 1 979),
EVEN',
98- 1 07.
informativeness , and logical form: radical pragmatics (revised standard version)', in
Radical Pragmatics , Academic
and
Fifth Regional Meeting, Chicago Lin
Role ofTopic and Comment in Theory , Indiana University ,
Linguistics Club, Bloomington. Halvorsen, Per-Kristian ( 1 978), 'The syntax and semantics of cleft constructions', Texas Linguistics Forum, 1 1 , Dept. of Linguistics, University of Texas, Austin.
Treatise on Syncategore matic Words rrans. by N. Kretzman ,
University o f Minnesota Press, Minnea polis. Srrawson, P. F. ( 1 9 54) . 'A Reply to Mr Sellars',
Philosophical Review, 63: 2 1 6- 3 1 .
Srrawson, P . F . ( 1 964), 'Identifying Reference and Truth-values',
Theoria 30: 96- 1 1 8.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
'x
accept the logical forms that I have given
'S', for
Socrates' as a predicate symbol 'Socratizes', rather than as
But I have no doubt that Horn would
146
The case of focal particles only and also
Appendix 1
OnlyA is_.E
ATLAS
GEACH
HORN
SHERWOOD
ASSERTION
One individual, k no one other than Lis f.
No one other than a is £
No one other than il is f.
A is £ k no one other than A is f.
GRAM. PRFSUPP.
!is f.
! is f. Someone is E
There is someone other than.A
IMPLICAT: CONVENT.
IMPLICAT: CONYERS.
LEVINSON SCALE !!. .
"not only" ->> also. � also is_.E
11-
PRFSUPP.
il is .El
TOPIC: Those who f
,! is £ [Not only ,il iS f » il is fl
a is E k no one
other than A is £
! is E
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
ENTAILMENT
TAGLICHT
Jay David Atlas
1 47
Appendix 2
�also is_E
ASSERTION
ATLAS
At least two
GEACH
HORN
SHERWOOD
TAGUCHT
� is f.
� is E.
� is f.
� is E.
individuals, one of whom is� are_E.
� is .E.
Someone other than
� is f.
GRAM.
There is
PRESUPP.
someone other than.l!,
Someone
IMPLICAT:
other
CONVENT.
than� is f.
IMPLICAT:
LEVINSON
CONYERS.
SCALE . "'also" ->> not only. FOCAL IMPLIC. "not ALSO"" -» ONLY.
PRESUPP.
TOPIC:
Someone
Those whof
other than � is E
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
ENTAILMENT
jounuJ/ of&mantics 8: 149-165
© N.l.S. Foundation (1991)
Focus and Presupp osition PETER I . BLOK University ofGroningen Ahsuact
I I NT R O D U C T I O N In Linguistics and Philosophy of November I988, Jay David Atlas published an article in which he claims to give a conclusive analysis of sentences like (I):
(I) The present king of France does not exist
()
As, for example, Hintikka has point�d out, 1 I is highly problematic for those
who propose a Strawson-style account of(2) (2) The present king of France is bald
With this approach, (2) has a truth-value gap under the present circumstances in which there is no king of France. But then, it is argued, the same must hold for (I), and so its own truth deprives the sentence of any truth-value whatsoever. Atlas draws our attention to the undeniable fact that (3 ) is simply false in our world:
(3) It is the present king of France who is bald This is e�plained by observing that (3) has the presupposition that there is somebody who is bald, and the sentence states that this somebody is the present king of France. Since the King of France does not exist, the sentence has to be false. This well-known phenomenon was accounted for by Strawson (I¢4), when he claimed that definite descriptions only have existential presupposi tions in sentences which are 'about' this constituent Strawson gives a 'recipe' for aboumess, which is something like 'being able to form a description of the utterance of the form (4)
(4) He (i.e. the speaker) was saying (describing, etc.) who (what, how, which, etc)
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
In this paper traditional approaches to the notion 'presupposition' are criticized. The relation between the concepts 'topic'; and 'presupposition' is discussed in a game-theoretical frame work. It is shown that the concept presupposition has to be defined pragmatically with respect to its dialogical functions.
1 so Focus and presupposition
((4) is Janet Fodor's (1979) paraphrase of Sttawson's argument}. Hence, ( s ) and (6} get the paraphrases (7) and (8} respectively:
( s ) The King of France visited the exhibition today
(6) The exhibition was visited by the King of France today {7) The speaker was saying what has been visited by the King of France today {8} The speaker was saying who the exhibition has been visited by today
(9) Pegasus does not exist
( 1 0) ?It is nothing that is Pegasus ( 1 1) It is not Pegasus who exists Since (ro) is supposed to be an incorrect paraphrase of(9), and { I I) is a correct one, it may be concluded, according to Atlas, that in (9) Pegasus is not and what exists is in topic-position. Here, these positions are determined by the so-called 'cleft-test'. So it is Atlas's ultimate claim that existential propositions have 'all that exists' or 'the existing' in topic-position, and what is claimed to be there in a non-topical position. It is, I think, an appealing theory. But it has, I am afraid, some less convincing aspects, especially the way topicality is defined and topics are recovered. The criteria by which the decision on the separation of a
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
The NPs in (7) and {8} which occur in subject position in ( s ) and (6} respectively are supposed to have existing referents; ifthey do not, then their originals lack a cruth-value. This is, as a matter of fact, exactly the line of argument Atlas proposes: the cleft in (3) is a well-known way to place a constituent in focus-position and Atlas wants to claim that only definite descriptions in topic-position have existential presuppositions. It should be noted that this presupposes that focused constituents like clefts are never in topic position. I will comment on this in the next section. The whole idea is, as I already mentioned, not totally new; if we look at Sttawson's examples ( s ) and {6), in the light of the 'recipe' (4), we can see that his claim is approximately the same: taking for granted that in unmarked constrUc tions subjects are in topic-position, it is obvious that in ( s ) the King of France and in {6) the exhibition are topics. Sttawson's 'aboutness-test' is nothing but an embedded question-test for non-focusedness, to which we will tum. our attention in Section 3· To summarize, the point Atlas wants to make is the following: in (2) the King of France is topic, and hence his existence presupposed; in (3) he is in non-topic position and hence his existence is not presupposed, and so what has to be proved is that in ( r ) the King of France is not in topic-position either, by which the apparent contradiction would disappear. To provide proof for this last claim, Atlas presents, among others, the following examples:
Peter I. Blok
I5I
sentence in a topical and a non-topical part should be made, are exhaustively deale with by Atlas, and will be discussed in the next section. In Section 3 ic will be pointed out chat every test for aboumess in relation co semantic presupposition which makes use ofnon-focusedness, like the so called question-test, begs the question indeed. Finally, I will present alternative, pragmatic, concepts of topicality and presupposition and show chat it is because (I) can only be uttered in contexts which are in face about 'the King of France' chat the sentence cannot be accused of presupposition denying. TOPIC AND FOCUS
In chis section we will cake a closer look at the concepts 'aboumess', 'topic' and 'focus'. Although these concepts are used in many different ways in the literature, we will confine ourselves co the sense in which Atlas uses them. In Section 4 ofhis article, he states: In a normally stressed statement of Johnny deceived the girl', Johnny is a topic-designating Noun Phrase (NP), Johnny himself is the topic, and the comment about the topic is that he deceived the girl. In 'It was Johnny who deceived the girl', Johnny is not a topic-designating NP, so Johnny is not the topic, by which I mean, intuitively, that the statement is not about him.2 The simple subject predicate statement that corresponds to the cleft statement i� the contrastively stressed statement JOHNNY deceived the girl', which is not about Johnny. These data illustrate several generalizations about statements (not sentences, please note): (i) Topical NPs are unstressed; they receive primary nor contrastive stress. (ii) Leftmost NPs (in surface forms) are not necessarily topical NPs. (iii) NPs carry presuppositions of the existence of their designations only if they are topical NPs. (For example, the use ofJohnny in a statement of Johnny deceived the girl' carries a presupposition that its designation exists, but the use ofJohnny in a statement of 'It was Johnny who deceived the girl' does not.) (Atlas I 988: 383-4)
It should be noted that the reference to the concept 'comment' in the first sentence of the quotation above is slightly misleading: the sentence mentioned has nothing to do with topic-comment constructions or their interpretacion (c£ Li & Thompson I 976). The generalization (i) already states chis, because topics in topic-comment constructions like as-for sentences receive contrastive stress (c£jacobs I984; Chafe I976). Generalization (ii) tells us chat topic in Atlas's sense should not be confused with theme in the sense ofHalliday (I 967). There it is assumed chat the theme is always the initial constituent of the sentence (p. 2 I 2), and moreover, chat the sentence is 'about' this theme. Non-standard word-order does not change the posicion of the theme, but it is a way to change themes. Therefore, Atlas's concept of topicality has nothing to do with Halliday's concept of aboumess or theme.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
2
I 52 Focus and presupposition
{a) As forJohnny, he deceived the girl (a') As for the girl, Johnny deceived her (b) ?As for Johnny, HE deceived the girl (b) As for the girl, JOHNNY deceived her {c) ?As for Johnny, it was he who deceived the girl (c') As for the girl, it was Johnny who deceived her {Atlas 1988: 385) These examples have the following significance: In (a}Johnny is in a structural topic position {topic in relation to comment} and it is claimed that it is 'about' Johnny, and not 'about' the girl. In (a') it isjust the other way round. (b) and (c) sho\0 that elements which appeared in structural topics can hardly be focused afterwards. But this does not mean that structural topics are always the, and the only, topic (as opposed to focus) of the sentence! As Chafe (1 976} pointed out, topic-comment sentences are, in western languages like English and Dutch, mostly used to express contrastivity of some sort. If (a) is placed in a context like (a'): (a") As for Johnny, he deceived the girl, and as for Benny, he married her both conjuncts are, in my opinion. also 'about' the girl. Moreover, (b) and {c) get much better in a contrastive context: (c") (About Johnny and Benny, which of them deceived the girl and which of them married her?) As for Johnny, it was he who deceived the girl, and as for Benny, HE married her.5 Hence, the topic-comment division has nothing to do with the difference between focus and non-focus (called topic by us). It is a function of the topic comment construction. though, to make a statement explicitly 'about' the constituent in structural topic position. But it is certainly not the most
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
Since topical NPs are never stressed (generalization (i)), they cannot be in focus.3 But it seems, at least from the examples given by Atlas, that the converse holds, too: every NP not in a focus position is topical. This can be illustrated by the main argument Atlas gives himself to locate topics: 'If an expression N is a topical noun phrase in a statement, the statement is a linguistically acceptable answer to the question "What about N?" ' {Atlas 1988: 38 5-6}. But everything which is not explicitly in focus (is stressed, is an answer to a Wh-question. is in a cleft construction) obeys this rule. Therefore, I would conclude by saying that what Atlas calls topical NPs are NPs which are not in focus. We will call the part of the sentence which is not in focus topic. It has to be stressed again that this notion of topic has nothing to do with so called topic-comment constructions. However, many of the examples Atlas gives makes use of as-for constructions and left-dislocation. which are typical topic-comment constructions:
Peter I. Blok 1 5 3
important one (c£ Chafe I 976). In the next section we will have a closer look at the way in which the difference between focus and topic is made and, more importantly, is recovered. For if one knows the topic of a sentence, one knows its topical NPs, what it is about and then what is presupposed.
3 QUESTI O N S I N F O C U S
In order to give the topic-focus partition of a sentence, several tests can be applied. We came across some ofthem in the previous section, and Atlas applies them in the following way to existential sentences like (12):
(12) is supposed to be an answer to the questions (I 3 ) and (I4}; (I s) should be a correct and (I6} an incorrect paraphrase of (12). Most readers may feel a little uncomfortable judging these examples. Atlas gives in a note some other possibilities like 'as for existing, John does' and 'As for the existing, John is one', but I do not think these are improvements at all. Compared to (10), I do not see very much difference. As a matter of fact, the native speakers of American English who were confronted with these sentences either rejected (I s) and its paraphrases or preferred (I 6} as the topic-comment version of ( 12). So let us concentrate on the questions and initially on the question-test itsel£ Observe the simple sentence (I7}, (r7) JOHN comes (again, capitals indicate stress). It is hardly controversial to state thatJOHN is in focus-position and comes in topic. Moreover, most theories assume that the existential closure of the topic is a presupposition of a sentence like (r7); hence, (r7) states something like: 'Given that someone comes, it is John who comes.' The question to which (r7) is an answer can be (r8} but also (I9): ( r 8} Who comes? (I9) WHO comes? In Hajicova (r984) it is shown that questions like (r8) and (19) do not have the same presuppositions. This may be illustrated by the fact that (2o) is a correct answer to (r 8), but as an answer to (r9) it seems rather odd:
(20) Nobody comes
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(12) John exists (I 3 ) Who/what exists (I4} What about what exists ( I S} As for what exists,John does (I6} As for John, he exists
1 54
Focus and presupposition
(21}-(23) are some convincing examples given by Hajicova: (21) What did you buy for him for a Chrisonas PRESENT? (22) WHAT did you buy for him for a Chrisonas present? (23) You bought him something for a Chrisonas present
(24) John has given A BOOK to a girl (25) What did John give to a girl? I think that in the question (25) the existence of girls is presupposed and hence, in this context, (24) also presupposes this. To resume: the following can be observed: if one identifies the presupposi tions of a sentence as the existential closure ofits topic, with the question-test as a criterion for topicality, one inevitably begs the question because one has only access to the relevant question if one knows the presuppositions of the declarative. If one tries to avoid this problem by not identifying the presupposi-
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(23) is a presupposition of(22), but not of(21); (2 1) can be answered by 'nothing', which is not a normal answer to (22). We may conclude that if we want to say something about the presupposi tions of questions we have to know their topic-focus structure. How does one explore the topic-focus structure of questions? By means of their possible answers, of course! So ifwe want to know the presuppositions of a sentence like (17) using the question-test we have two possibilities: either that 'someone comes' is a presupposition, or that it is not. We will never know which one to choose. Therefore, I accuse the question-test for presuppositions of circularity. There seems to be a way out for Atlas, his argument being the following: I want to know the presuppositions of a declarative sentence-therefore I need to know the topic-focus structure-therefore I need to know to which question it is an answer-! have two possibilities with different presuppositions, but with one and the same topic-focus structure-! take the existential closure of this unique topic as the presupposition of my declarative sentence. To summarize: questions do not always have the existential closure of their topics as a presupposition but declaratives do. This argument, though, is immediately refuted by (2o), Nobody comes. If the argument had been correct, the sentence would have had the existential closure of the topic of its questions, 'who comes', as a presupposition. But obviously, it is not a presupposition here that 'to come' has any extension. Hence, the existential closure of the topic found with the question-test is not a presupposition here at all. It may be, then, that the argument only holds for definite descriptions: it is only true of definite descriptions that their existence is presupposed in declaratives ifthey occur in the topic of all questions to which the declarative is an answer. This, however, is obviously false, because it is true of almost every NP. Look at (24) and (25):
Peter I. Blok
I
5S
4
THE P RAGMAT I C S O F P RES UPPO S ITI O N
In the previous section it has been made clear that although the intuition that only sentences which are about a certain definite description have existential presuppositions with respect to it seems correct, this aboutness cannot be defined more precisely in terms of topicality without begging the question. This is because the notions of topic, as the counterpart of focus, and pre supposition are so closely connected that hardly anything about the one can be explained in tenns of the other without circularity. Therefore, everybody who entertains a semantic concept of presupposition, by which I mean pre supposition as precondition, in tenns of the meaning of some of the content of the sentence, on the possible models with respect to which the sentence can be evaluated, will have problems with sentences like (r8) and (2o) (and of course ( 1 )) . The alternative seeins to be that the existential closure of the topic is not presupposed automatically, but then there is no way to recover the presupposi tions either. In an extremely lucid article, Karttunen (1973) has already pointed out that a purely semantical approach to presuppositions has to fail. In the analysis of filter conditions on the projection of presuppositions in compound sentences, it turns out to be the case that the contextual environment is of conclusive importance; hence, the mutual knowledge of speaker and listener should be
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
tions of question and answer, one gets the truism that NPs must have some sort of referent to be able to say something about them, which does not hold for definite descriptions alone. The central problem here is, of course, that in the semantic concept of presupposition, like Atlas's, a notion of topicality is inherent already. This can be seen if one takes into account that the semantic presupposition concept is usually expressed in terms of the shared consequence of a proposi tion and its negation; and naturally, the scope of the negation is determined by the topic-focus division.6 Therefore, looking for the presuppositions and the topic-focus division is exacdy the same thing. This is made manifest in the paraphrases (7) and (8) of the Strawson examples ( s ) and (6). Whether something is presupposed or not, though, and to what degree, is a context-sensitive matter. This can be illustrated by the examples (2 1 ) and (22). The presupposition of (22) is a result of the context to which the sentence belongs (note that it is hardly ever the initial sentence of a discourse!). Hence, if we want to say something conclusive about this matter, we should not just flirt with some specially chosen small pieces of context, but have a rigorously pragmatic view of it This holds afortiori for the topic-focus distinction.
1 56 Focus and presupposition
taken into account for any successful treatment of the matter. A more pragmatic view is necessary, for which there are a number of quite similar proposals, e.g. Jackendoff: 'presupposition is the information in the sentence that is assumed by the speaker to be shared by him and the hearer', or Sgall: 'Knowledge, or other items, stored in the memory of the speaker, and supposed by him or her to be also present in the hearer's memory'.7 One should be extremely careful when trying to reinterpret the notion of topic in connection with a pragmaticized notion of presupposition, because if one identifies the one with the other again, one inevitably identifies the topic focus partition with the old-new distinction, which is rather dubious, I think. Examples (26)-{27), which are from Reinhan (198 1), may illustrate this point:
HimseDfin (27) is obviously in focus position, but its reference is not new in the context of (26). As Danes (1 986) states, it is not the topic which is 'old information' and the focus which is 'new', but the connection between the two is new. So, in the sentence :John swims', in whatever intonation, :JOHN swims', :John SWIMS', or neutral, the new informative aspect is not :John', or 'to swim' respectively, but the fact that the predicate holds for its argument. So, if the topic-focus partition is �ot the distinction between what is semantically presupposed and what is not, nor the distinction between 'old' and 'new' information in a ·more pragmatic sense, what kind of function does it have? I think it should be stated in terms of search-strategies rather than in terms of informational content. The databases of my conversational panner and me are conceivably rather big; if I state something, there are two possibilities: either I assume that my partner does not have any opinion on the matter, or I assume that he or she has one opposed to mine. In the latter case, I will help my panner to find this opinion opposed to my utterance in his or her database by means of intonation, sentential construction, or in general by means of the topic-focus articulation. This view implies that one can only talk meaningfully about a topic-focus distinction within a context In the past, this has only been stated implicitly; something like the question-test is, of course, doing nothing but creating a suitable context. But what assures us that every sentence can occur in the context of a wh-question? Let us have another look at the problematic sentence ( 1) 'The present king of France does not exist'. One utters a sentence like that only if one assumes that one's partner in discourse has something to do with the present king of France, his concept or imaginary being. Hence, there is always some connection between a context, shared by you and your partner, and the sentence, which says something like: skip everything on the king of France, i.e. every file you
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
(26) Whom does John love? (27) John loves HIMSELF
Peter I. Blok 1 5 7
have on him. These are the files with his name (description) on it, where he is in
topic-position.
��
a
b Whoj<:omes,
Nobodyr comes, �come? a
b
. . . come . . . WHOr comes, Nobody !! In {28), there is no existential presupposition of the topic; the answer 'Nobody comes' is quite acceptable and, moreover, if someone doubts its relevance, for
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
My opinion on the relation between the topic-focus distinction and pre suppositions is the following: in a dialogue, which I consider game theoretically, the participants make a move by turns. The moves are bound by rules, depending, among other things, on the kind of move at hand. Hintikka (1986) differentiates e.g. assertoric moves, interrogative moves, deductive moves and definitory moves. Presuppositions are dependent on the type of move: questions do have presuppositions, according to Hintikka, which they share with their answers, but assertoric moves do not have to have presupposi tions, but they may, etc. My general concern is the relation between the moves in an information-seeking or -exchanging dialogue, the ways they depend on each other. These context-dependent game-rules, one could say, are the relevance conditions on mc.7es. Intuitively, it must be clear that this has something to do with topic-focus structure: about a certain topic information is being exchanged, questions are being asked, etc. This idea is not new: Danes (1974) has already tried to describe the progression of discourse in these terms. Let us assume, for the sake of the argument, that we only have the following progression rule: a move is relevant if the information about a certain topic increases, or questions for information are asked about it, a criterion proposed by Klein & von Stechow (1982). Since we are assuming a cooperative attitude, one has to interpret a sentence uttered by one's parmer in such a way that it is relevant, i.e. form the topic-focus division in such a way that the move is a relevant one, if possible. We will tum our attention to this in the next section again. We add to this general rule that prohibits one from questioning one's own knowl_edge. Look at the schemata (28) and (29):
1 5 8 Focus and presupposition
5
RULE S A S REPRE S E N TAT I O N S
We will now take a closer look at the schemata in order to get a formal definition of the difference between presupposed existence and relevance. In (28), rhe move in which b doubts the relevance of the answer 'Nobody comes', a
doubt which is of course initiated by the misconceived idea that a predicate cannot be relevant if it has no extension, is not a normal one in information exchange dialogues. Of course, it is possible to perform such a move to recover the stage of the game to which rhe participant's last move refers, but in rhat case one performs it, so to speak. with the intention of losing. This point illustrates why it is important to distinguish between several kinds of games:
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
which I use the double question marks, this party will lose immediately, for he introduced the topic himself In (29), a started to admit that there are things coming and therefore it is impossible to answer the question, which is stated in a way that shows that the questioner is aware of this presupposition by using stress, with 'Nobody'; it contradicts a's database. There are a number of points I wish to make with these simple examples; the first is that the relevance of the second move in (28) does not provide or presuppose existence; generally, existence and relevance are not directly connected. On the other hand, in (29), the question, with stress on WHO, indicates that there is an existential presupposition of the topic, dependent on the context, which, then, has to be existential itself because of what we saw in (28). This implies that the existential impact in (29) is a special case of relevance. I think this can be explained quite easily: stress is used to indicate to which part of the context the lingusitic content of a speech-act refers (c£ Carlson 1984); a question like the one at hand asks for information about such a topic and there fore, the move being interrogative, submits itself to the information already given that 'to come' has extension. This explanation predicts that in contexts where the initial phrase of (29) topicalizes 'to come', but does not state that it has actual extension, the answer 'Nobody' is not as awkward as in the cases mentioned before. Take for example 'I sold all the tickets, but as for coming, I don't know'. Then, 'WHO comes?' can be answered by 'Nobody (I fear)'. This may illustrate that the relationship between topicality and existential impact is quite coincidental; also, in Hajicova's examples, the only thing we may conclude is that the stressed variants are not initial sentences of a discourse, with all its consequences, of course. The relation is no more than that generally things which are talked about do have extension.
Peter I. Blok
1 59
A Dialogue-rules ( 1) a move is not allowed if it would lead to an immediate loss in a discussion with the same history as the dialogue (2) the players make moves by turns (3) the players should be cooperative, i.e. interpret a sentence of their partners with switched roles
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
information-seeking or -exchanging games should not be mixed up with discussions and vice versa. It is undesirable to mix up different goals at one and the same level. In this example, the goal, losing, can be reached in many ways, e.g. by beating up one's partner or by contradicting onesel£ This is, of course, not what we intend by defining the rules of game stage recovery. What is possible, on the other hand, is the ability to define the rules of an information-seeking game in terms of a formal discussion. One could say that the answer. 'Nobody comes' to the question in (28) is possible because one would win immediately if attacked on its relevance in discussion. Discussions perform meta-rules for information-seeking dialogues, and are rule-governed themselves.8 So, if a party is obstructing the rules of the object game, this can be pointed out one level higher. The whole notion of relevance itself can be stated, I think, in terms of discussion-games. According to this view, the rule that one cannot 'question one's own data base' can be stated as follows: first, there is a rule of topic introduction; it may be assumed that this move is explicitly performed in one way or another. In most West European languages it suffices to make a statement with the intended topic in subject position, giving no particular stress whatsoever. Now one can say that it is possible to introduce a topic and state something about it at object level because one will not lose a discussion game about it at the meta level. Then, analogous to the rule that one cannot make a statement and attack it in the same discussion, one cannot introduce a topic and subsequently attack it on grounds of irrelevance. If one does so, the game will be lost immediately by a 'you said it yourself!' move of the opponent. Hence, a move on object level will be relevant with respect to topicality if the opponent at meta level loses by questioning it. To summarize, let me give an overview of the representation I have been proposing: an information-seeking discourse is at least an eight-tuple consist ing of two participants, say a and b, their knowledgebases, say K(a) and K{b), their respective stacks of moves D(a) and D(b), a set of rules R, and an initial thesis i. If we assume that the representations for both parties are the same, we can abbreviate this to parties, knowledgebases, dialogue-scheme, rules and initial theses. The rules are defined in terms of formal discussions and we encountered the following:
160 Focus and pre3Upposirion
{4) topics are represented in K as file-names, focuses as their content every utterance is interpreted as a topic-focus structure which corresponds
( s)
to the file-structure of the knowledgebases K. That is, a party who commits
himselfto a certain topic must have a file in his K with this topic as its name and the focus as its content. A parry who lacks a file with the topic as name assumes his partner to have one while interpreting the sentence, as a consequence of rule
(6)
3·
the goal of the game is to make the knowledgebases of the parties match with respect to the topics talked about
I hope the picrure is now clear. There is no direct connection assumed between topicality and existential impact, therefore the attack on relevance in (28) is lost and, hence, the answer 'Nobody' is possible. In (29), on the other hand, things are a little different. The attack on the answer does not focus on its relevance, but on the fact that it is contradictory to earlier utterances made by the same party. There should be, of course, a rule which prevents one from contradicting oneself without first withdrawing a
former presumption. This is expressed by dialogue-rule I which says that a move may not lose immediately in a discussion. Here again we see the connection between the two levels: the general rule for information exchange is
expressed in terms ofwinning and losing a debate. Note that this does not imply that only non-self-contradictory databases can participate: the only thing is that they should not show it to their partner! (28) and (29) tum out to be schemata for certain dialogue-rules, not for dialogues. The first is a scheme for topical relevance, the second for consistency. Ifin (29) the initial sentence is 'somebody comes', it is inconsistent to answer the question with 'Nobody'; this has nothing to do with relevance. The only
connection is that the question itself is relevant with respect to the initial sentence, a relevance referred to by the speaker by using stress.
These rules are only a rough sketch of what I have in mind, but the general idea should be clear. Of course, the restrictions on relevance are far too strong on the progression side and too weak on the informational side. There are many
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
B Discussion-rules ( I ) all rules known from the literature on two role zero sum discussions, so called 'formal dialectics'. (2) by stating i, a player introduces a topic ( 3) both players can question a topic, i.e. ask where a topic comes from (I use the Spanish question marks for this kind of move) (4) a player who fails to give an answer to the question defined in discussion rule 3 loses immediately ( s) a player loses immediately ifhe questions a topic which he has used himself
Peter I. Blok
161
more possibilities than saying something about a topic encountered before and, on the other hand, superfluousness is not prevented. The rule on cooperativeness needs special attention. Even if we assume that there are no problems in interpreting utterances at all, e.g. by identifying the semantics of the language with itself, one can always obstruct the dialogue by misinterpreting the topic-focus partition. For example, the following dialogue: (Jo)
b
a Who comes
John comes
In (Jo), a obstructs by interpreting b's answer as saying something about john. The cooperativeness rule prevents this: because
a has to interpret the sentence
uttered by b as if he said it himself, he will have to assign it the correct division because otherwise he would lose the discussion about its relevance. There is still another point I want to make before turning to the king of France again. It is obviously not true that all sentences can be initial. For instance, a heavily stressed sentence always refers to a previous one.9 There are
sentences, on the other hand, which are typical 'first sentences'. A question, for instance, is such a sentence. It sets a topic and asks for information about it. One could say that initial sentences have the presuppositon of relevance. This
point is not always appreciated. Hintikka, e.g. in his I 98 I paper on information seeking dialogues, persists that questions have existential presuppostions, which is obviously not true. He gives as an analysis for the question 'Who lives
in that house?' 'Bring it about that I know for some x that x lives in that house'; the desideratum is stated in (3 I): (J I ) (Ex)K1(x lives in that house) and he explains this in the following manner: 'What else could it possibly mean to know who lives in that house than to know of someone that he or she lives there?'.10 This implies that if I know that nobody lives in that house, I do not know who lives there, which seems quite absurd. If, on the other hand, it is already given that the house is inhabited, the wh term gets stress and the analysis seems correct. But then it is not an initial sentence anymore. Set-theoretically speaking, one could say that in the former case one asks for a set and in the latter for its members, its non-emptiness being given. This explains why in the former case the empty set is possible as an answer but not in the latter. . The important difference we have encountered between relevance and consistency is very well illustrated by the existential statement about the king of
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
John?
162
Focus and presupposition
France. Let us have a look at the kinds of contexts in which the sentence 'The King of France does not exist' occurs. First, there is the simple situation where someone asks (32): {32) Does the King of France exist? which, in my opinion, means: do you have a file with the name 'the king of France'? If the answer is 'No, the King of France does not exist', the King of
France is topic (we are dealing with the name of the file) but the questioner has
no attack whatsoever on the answer. It is relevant, and not contradictory with anything said before, it being the first thing b says. In a context like (2), the
Now, he just reports that he does not have such a file, and advises a to skip his. Again, he cannot be accused of inconsistency or irrelevance. What, then, is the role of existential presupposition of definite descriptions in the representation presented here? It is no more than the observation that one cannot make a statement about something one has no access to. Since there are two ways in which one has the right to say something about a certain topic, in an initial sentence and as a response to an utterance of one's partner, there are also two ways to have access to a topic. In the former case one has to have a file at one's disposal which has the intended topic as a name. In the latter, one can either react factually, and give information about one's own file with that name, or formally, for instance, by denying the existence of the file, as in the example at hand. The possibility of a move like this depends on the contextual character of it: it is relevant because it is a reaction to a certain topic proposed by one's partner, and it is consistent because one has not yet stated that one has a file with the topic as a name. Therefore, the sentence 'The present king of France does not exist' is possible and meaningful, but it cannot be the initial sentence of a discourse. The way the relation between dialogues and discussions is structured allows us to make the following observation: in everyday discourse, the levels are frequently 'mixed up'. There is nothing wrong with that, one can talk
about
the dialogue if problems arise and rerum to it when they have been solved. And
()
that is exactly what happens if 1 is uttered in the context of (2). There it is a
discussion move which attacks the correctness of (2). The conclusion that this prevents the sentence to be an initial one holds afortiori. The whole solution to the problem seems so trivial that one might think
there has to be something wrong with it. This feeling may disappear, though, if one takes into account that asking for the presuppositions of an utterance is no more than asking for the possible contexts in which the sentence can occur. I
think that this is the pragmatic interpretation ofthe Strawsonian solution in the first place: in contexts when there is no king of France, the question whether he
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
move is approximately the same as in the first case; if b does find a file in his database with the name 'the King of France', he will write the baldness down.
Peter I. Blok 1 6 3
is bald or not does not arise. The same holds for his non-existence: sentences about the existence or non-existence of something only occur in contexts where somebody has made it known that they hold the opposite opinion, for instance, by referring to the object of the controversy as a topic.
Acknowledgements I would like to thank Jay Atlas, Ton Carbon,Jaap Hoepelman, Frans Zwarts and specially Sjaak de Mey for their stimulating conversations and useful suggestions. I thank Ale de Boer and Rudi Schnitzer for their help to make my contribution ready for publication in English. I take full responsibility for all remaining mistakes and errors, with respect to both content and style of this paper.
PO Box 7 1 6 9700 AS Groningen
Netherlands
N O TES 1 Hintikka (1986). Atlas quotes this obser vation at the beginning of his paper.
cleft version (c). It would be interesting to know whether this is a coincidence or not.
This view of Atlas, though very common,
Jacobs (1984) gives similar examples like (c..) in German.
2
is not held by everyone; e.g. Somicola (1988: 3 5 6) states, referring to ( 1) The one who goes to Rome is me
It is me who/that/0 goes to Rome 'they (It-clefts and Wh-clefts) do not
6 Note that it does not make sense to differentiate between different types of negation within such a theory for the same reasons: one would have to know
share values on the aboumess parameters. ( r ) is "about" "the one who goes to Rome", whereas (2) is "about"
the presuppositions to be able to decide between the two negation types and one would have to be sure which negation is
me'. This may illustrate the fuzziness of the matter once more. One of the few properties that we can be sure so-called focus-constituents have in common is that they can receive, stress. 4 In personal communication, Jay Atlas
the one at hand to get the right pre suppositions. This problem is known as Hom's dilemma. 7 Citations from Gaular (I 979).
8 In the representation as presented here, necessary frame rules and stop rules are
confirmed that (b) and (c) are, in his dialect, not totally ungrammatical but only hard to interpret. My intuitions
omitted. The essence, the backbone is an idea of Jaap Hoepelrnan (personal com munication) to model information
concerning the Dutch counterparts are the same for these examples if considered in isolation. 5 Gundel ( 1 977) rejects sentences of the form (b), but she does not mention the
exchange between two parties as a discus sion between them on the one hand and Nature on the other. An optimal strategy
( 2)
So for example,
to defend a thesis together will consist of a maximal use of the databases with respect
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
PETER I. BLOK Dept. of Computational Linguistics University of Groningen
164
Focus and presupposition to it. And that is what we are aiming at.
supposition is dependent on the rype of
The discussion rules berween Narure and
move: sentences with stressed constiru ents are used correctively or as answers to
the rwo databases will constirute the dialogue rules in a way analogous to what
is indicated here. This idea will be elabor ated in Blok (forthcoming).
9
C£
( 1 984).
Carlson
This
is
another
instantiation of Hintikka's idea that pre-
questions, and these functions constirute the presuppositional form of the locu tions.
10 Hintikka ( 1981: 2 1 8 ).
R E F E RE N CE S
( 1 988),
Atlas, J. D.
'What are negative
Philosophy, n
Linguistics and
Adas,J. D. & S. C. Levinson ( 1981 ), 'It-clefts, informativeness, and logical form: radical
pragmatics', in Cole (ed.) ( 1981 ).
Becker, W. & W. K. Essler (eds) ( 198 1 ),
Kon
zepte der Dialektik, Klostermann, Frankfurt a. M.
Blok, P. I. (forthcoming), 'Focus and dialogue' (working tide), doctoral dissertation, Uni versiry of Groningen. Burton-Roberts,
N.
( 1989).
'On
Horn's
dilemma: presupposition and negation',
journal oJLinguistics, 25.
Carlson, L. (1 984) , 'Focus and dialogue games' in Vaina and Hintikka (eds) ( 1984).
Chafe, W. L.
( 1976), 'Givenness, contrastive
ness, definiteness, subjects, topics and
( 1 976). ( 198 1 ), Radical Pragmatics,
point of view', in Li (ed.) Cole, P. (ed.)
Academic Press, New York. Danes, F.
(1 974),
'Functional sentence per
spective and the organisation of the text',
Papers on Functional Sentence Perspective, Prague.
Danes, F. ( 1986), 'The "Question test" re examined', in Mey (ed.)
Fodor, J. D.
( 1979),
(1 986).
'In defence of the truth
value gap', in Oh & Dinneen (eds) ( 1 979).
Gazdar, G. ( 1979), Pragmatics, Academic Press,
New York.
Grandy, R. E. & R. Warner (eds)
( 1986),
Philosophical Grounds ofRationality, Oxford
Universiry Press, Oxford. Gundel, ]. K.
(1 9771I974), 'The role of topic
and comment in linguistic theory', Indiana
Universiry Linguistics Club, Blooming ton.
Hajicova, E. ( 1 98 4), 'Topic and focus', in Sgall (ed.) ( 1 984) .
Halliday, M A. D. ( 1 967), 'Notes on transitiv
iry and theme in English II', Journal of Lingusitics, 3 · Hinrikka,J. ( 1 98 1 ), 'The logic of information seeking dialogues: a model', in Becker &
Essler (eds) ( 1981 ). Hinrikka,J. (1 986), 'Logic of conversation as a logic of dialogue', in Grandy & Warner (eds)
( 1 986).
Jacobs,]. ( 1 984), :Funktionale Stazperspektive und
Illukationssemantik',
Berichte, 91.
Kartrunen, L.
( 1 973),
Linguistische
'Presuppositions of
compound sentences',
Linguistic Inquiry, 4·
Klein. W. & A. von Stechow ( 1982), 'Intona tion und Bedeutung von Fokus',
forschungsbereich Constance. Li, C. N. (ed.)
99.
Sonder
Univer5iry
of
( 1 976), Subject and Topic,
Academic Press, New York. Li, C. N.
&
S. A. Thompson
( 1 976), 'Subject
and topic: a new rypology of language', in
Li (ed.) (1976). Mey, J. L. (ed.) (1986), Language and Discourse: Test and Protest, Benjamins, Amsterdam. Oh. Ch.-K. & D. A. Dinneen (eds) (1979), Syntax and Semantics 1 1 , Academic Press, New York.
Pumam, H.
( 1958),
concept "About"',
Reinhart, T.
'Formalisation of the
Philosophy oJScience, 25.
(1981� 'Pragmatics and linguis Philo
tics: an analysis of language topics',
sophica , 27.
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
existence statements about?',
Peter I. Blok 1 6 5 Sgall, P. (ed.) (1984), Contributions to Functional
SyntiiX, Semantics and Language Comprehen sion , Benjamins, Amsterdam Academia,
Prague. Sornicola, R (1988), 'It-clefts and wh-clefrs: two awkward sentence types', Linguistics, 24.
Strawson, P. F. (1964), 'Identifying reference
and truth-values', Theoria , 30. Vaina, L & J. Hintikka (eds) (1984),
Constraints on Communication ,
Cognitivt Reidel,
Dordrecht.
journal of
Downloaded from jos.oxfordjournals.org by guest on January 1, 2011
