Date: Tue, 17 May 2005 14:35:45 +0200 From: Mikhail Kissine Subject: Semantics: A Reader
EDITORS: Davis, Steven; Gillon, Brendan S. TITLE: Semantics SUBTITLE: A Reader PUBLISHER: Oxford University Press YEAR: 2004
Mikhail Kissine, Laboratoire de Linguistique Textuelle et de Pragmatique Cognitive, Aspirant FNRS, Free University of Brussels
The book is a collection of papers in semantics, most of which are classical, previously published pieces of work.
Part I: Introduction, consists of chapters written by the editors. In the Preliminaries, the editors set the purposes of the volume: they provide an overview of theories that address meaning as a property of expressions, that have been applied to a given range of linguistic phenomena, and that are possible to evaluate. Such criteria of the choice of articles are thus not constrained by the ontological commitments of the theory, which, according to Davis and Gillon, should not take place prior to semantic theorising.
The second chapter, Linguistics and Psychology, offers a summary of the main theoretical positions on the nature linguistic ability, such as the opposition between behaviourism and mentalism, and the dispute between Chomskian partisans of an independent Language Acquisition Device and defenders of non-modular approaches. The authors define the notion of I- language (I-semantics), and discuss the way I-semantics (as the study of internalised semantic ability) can be applied to a whole community of speakers.
The third chapter is entitled Linguistics and Logic. It provides a gradual introduction to the main notions of formal semantics: propositional logic, predicate logic and quantificational logic. Each notion is first introduced with the help of simple formal examples and then applied to selected phenomena of English. The introduction to quantificational logic leads to the discussion of quantified noun phrases and different approaches to it: Evans' conception of pronouns as going proxy for definite descriptions, Discourse Representation Theory, and Dynamic Predicate Logic.
The fourth chapter, Theories of Reference and Theories of Meaning offers a classification of semantic theories. Davidson's satisfaction theory, which is given as an example of reference theories, is introduces and compared with model-theoretical approaches. Meaning theories are divided in two: atomic theories, that do not admit constituent internal to the lexemes, and molecular theories, that admit sub-lexical compositional structures. The molecular theories are sub-divided in analytic theories and encyclopaedic theories. The former are exemplified by Pustejovsky's approach, where analytical sentences are thought to give the meaning of lexical entries. Jackendoff view, which admits encyclopaedic information within concept constitution, illustrates the latter ones. Finally, the authors expose approaches that attempt to account for both the meaning and the reference: Frege's notions of sense and reference, and Kaplan's distinction between character and content.
The next chapter, Internalist and Externalist Semantic Theories, expands the former one. Internalist theories that assume that the lexical meaning pertains totally to the I-grammar of speakers are again sub-divided in atomic theories and molecular theories. The authors make a clear case against atomic theories, which amount to postulating unanalysable concepts, corresponding to lexical entries, within the language of thought. They discuss next the problem raised for molecular theories by entities that, form an ontological point of view, are not part of speakers grammars, and hence, cannot be part of the language of thought. They emphasise, however, that externalist theories are not incompatible with the existence of internal grammars, i.e. with a mentalist approach to semantics. Again, externalist theories are sub-divided into atomic, Davidson-like, theories, and molecular theories, à la Pustejovsky and Jackendoff. The authors discuss Frege's sense, Kaplan's character and model-theoretical semantics with relation to the question of externalism. This part is concluded by the suggestion, aimed at avoiding Putnam's (1975) claim that 'meanings are not in the head', to treat meaning as a set of rules defined with respect to an idealised speaker.
The final chapter, Semantics and Context, is devoted to the by-now classical problem of contextual contribution to the content of utterances. The authors argue against underlying place-holders, both syntactical (cf. for instance Stanley 2000) or semantical, as follows from Kaplan's notion of context. The end of the chapter contains an introduction to Relevance- theoretical notion of the explicature of an utterance (e.g. Sperber and Wilson 1995; Carston 1988, 2002). However, the authors point out that Relevance apparatus may explain how the context determine the semantic value in an epistemological sense of 'determine' only: since interpreters are sometimes mistaken, the semantic value provided by mechanisms of interpretation may not correspond to the actual semantic value of the utterance.
Finally, an appendix contains a very clear and condensed introduction to the notions of set theory necessary to the apparatus of formal semantics.
Part II: Background The first paper is Pelletier's "The Principle of Semantic Compositionality". The author provides a discussion of arguments against compositionality, and offers a method to get to a definition of semantic evaluation in a way that is non-compositional and compatible with facts about learnability and understandability of natural languages: such an evaluation should be recursive but non-inductive, i.e. it must be grounded by some basic operations. The paper is followed by originally published "Afterthoughts 2000" and "Afterthoughts 2004". The author claims that semantic atomism does not entail compositionality, and, consequently, that the possibility of non-compositional but atomistic semantic theorising constitutes a powerful argument against holism. He expresses also some doubts regarding holism in other fields. Finally, Pelletier makes a case for compositionality to be understood as a function of all syntactic components, and for empirical importance of compositionality.
The second paper is Gillon's original "Ambiguity, Indeterminacy, Deixis, and Vagueness", which draws clear-cut boundaries between these four concepts. The main part of the article shows that the most liable way to define ambiguity is the plurality of possible label bracketing, and the most reliable test to detect it is the existence of conflicting truth judgement, each of which is coupled with a different circumstance of evaluation. Ambiguity is then contrasted with indeterminacy (that admits, unlike ambiguity, endophoric expressions with a different sense than their antecedent), with deixis (that sets the context of evaluation), and with vagueness (that consists in conflicting truth-value judgement with respect to the same circumstance of evaluation).
Part III: Approaches contains a series of foundational papers that exemplify different streams in semantic theorising. This part begins with Lewis' "General Semantics", where is developed a semantics for natural languages based on categorial context-free grammars. While for primitive categories (e.g. NP) intension is a function from indices to extensions, for derived categories (adjectives, VPs, adverbs) intensions are complex, that is, corresponding to functions from intensions to intensions. Lewis defines meaning as "semantically interpreted phrase markers minus their terminal nodes" (p. 119).
Lewis's paper is followed by Davidson's very influential "Truth and Meaning". It contains a famous version of the 'Slingshot argument', according to which a proposition may correspond to whichever fact, and the formulation of a Tarskian holistic semantics for natural languages.
Kamp's paper "A Theory of Truth and Semantic Representation" is an early formulation of Discourse Representation Theory, further developed in Kamp and Reyle (1993). The principles of this non-compositional theory that attributes truth-conditions to intermediate representations (Discourse Representation Structures, which capture, in addition to truth-conditional information, availability for anaphoric relations) are discussed with the relation to quantificational noun phrases.
Groenendijk and Stokhof's "Dynamic Predicate Logic" represents an opposite solution to the same problem: truth is defined with respect to the assignment functions in a way to keep track of the context change, which allows quantifiers to get scope over conjunctions and sentence boundaries, and provides the right interpretations for "donkey sentences".
Barwise and Perry's "Situations and Attitudes" is an outline of Situation Semantics (see Barwise and Perry 1983) based on a discussion of attitude reports: meanings are taken to be relations between utterances and situations, i.e. between states of affairs consisting of entities, events, states, and relations between them (which implies a realistic stance towards relations).
Jackendoff's paper "What Is a Concept, That a Person May Grasp It" is a very clear and representative example of internalist generative theories that position themselves as incompatible with model-theoretic semantics. The organisation of concepts is thought of in terms of fundamental categories, realised in function-argument structure. This internal structure is used to account for the sameness of patterns across different semantic fields. The model also integrates outputs from other cognitive systems, such as vision.
The last paper of the second part is Fauconnier's "Mental Spaces, Language Modalities, Conceptual Integration". The paper offers a clear and summarised introduction to the theory of mental spaces (cf. for instance Fauconnier 1994), where grammatical information is seen as providing a skeleton for the construction of conceptual spaces, and the paths of access between them. It outlines a treatment of coreference, tenses and conceptual blending, and offers examples from American Sign Language.
Part IV: Topics The first two papers pertain to lexical semantics. Pustejovksy's "Generative Lexicon" as an analytical theory of lexical meanings that offers rules for the generation of new lexical meanings. Each lexical entry is composed by its quale structure, its argument structure and its inheritance (relatively to other concept) structure. Meaning is viewed as being flexible, and arising generatively by composition of lexemes.
Gillon's "Towards a common semantics for English count and mass noun" offers, as its title indicates, a unified account for count and mass nouns. Counts nouns are associated with the feature +CT, that allows a free selection of grammatical number + or -PL, while mass noun have the feature -CT, and are associated either +PL or -PL. The feature -CT associates with the name the greatest aggregate of which the name or the NP is true, while +CT associates a set of all minimal aggregations of which the name or the NP is true.
Burge's "Reference and proper names" defends a view according to which proper names are not abbreviations of predicates of the kind of "X is the name of _" or individual constants. Instead, proper names are analysed as playing the role of demonstratives and of such predicates at the same time.
The next two articles are devoted to the problem of the semantic status of pronouns. Evan's paper "Pronouns" argues against the view that the referents of bound pronouns are determined by extra-linguistic factors. Instead, using a Fregean notion of satisfaction (according to which object x satisfies a predicate A, iff A(b) is true, such that b is an expression referring to x,), Evans argues that pronouns are co-referential with their antecedents. The pronouns that have a quantifier phrase as antecedent (E- type pronouns, in Evan's terminology) and are not bound by the quantifier of that phrase are analysed as referring to those objects only that verify the antecedent.
The same topic is addressed within the Government and Binding framework by Higginbotham's "Pronouns and bound variables". The core idea of the article is that pronouns are variables coindexed with its antecedents: this is achieved thanks to a reindexing rule that deletes the bound pronoun referential index and thus leaves only the anaphoric index available.
The next paper is the foundational Barwise and Cooper's "Generalized quantifiers and Natural Language". The main insight of generalised quantification, as applied to natural languages, is to treat quantifiers as constituted by the combination of a determiner and a set expression, and to take them as denoting families of sets. Beyond the surface level, every natural language contains an NP constituent that express generalised quantification over the domain of discourse. Among the most important achievements of Barwise and Cooper are probably the distinctions between strong and weak quantifiers, and monotone decreasing and monotone decreasing quantifiers. More generally, the paper illustrates the possibility to study of linguistic competence with the help of logical tools (like the notions of validity and inference).
The next topic is the semantics of tense. An excerpt from Reichenbach's Elements of Symbolic Logic, "The Tense of Verbs", contains the seminal formalisation of English tense system with the help of the threefold distinction between the time of the event, the time of reference, and the time of speech.
Barbara Partee, in "Some Structural Analogies between Tenses and Pronouns in English", makes a case for introducing time variables in the logical form of sentences, and not treating tenses as time operators. She notices that present tense may be used only deictically, like "I", while the past tense has also a referential reading, like "they", and may be used as anaphorically.
Kamp's paper "Two theories about adjectives" starts by exposing the analysis of adjectives as being functions from meanings of NP's to meanings of NP's, and rejecting the view according to which some adjectives are functions from properties to properties. In order to solve the problem posed by comparatives to the former theory, an alternative theory of vagueness is defended, where some sentences do not pertain to the union of the set of all true sentences and all false sentences. This account, which maintains that adjectives are one-place predicates, predicts that in absence of contextual criteria of comparison two sentences containing a given adjective cannot be judged false or true.
In "Prepositions and points of view", Cresswell provides a lambda- categorial formalisation aimed at capturing the perspectival information contained in prepositions, such as "across", without treating the point of view as a contextual index. The basic machinery for doing so is associating a time interval with each possible world and introducing a path function p over moments m, such that (p,a,w)(m) is the space occupied by a at m in w.
Three papers address the semantics of adverbs. Bellart's "On semantics and distributional properties of sentential adverbs" questions the two-partite division of adverbs into predicative adverbs (e.g. "slowly") and sentential adverbs (e.g. "fortunately", "possibly"). In addition to semantic criteria, distributional properties, which take into account the position of the adverb and the semantic category of its admissible arguments, provide a more fine-grained classification among sentential adverbs.
The distinction between predicative adverbs, formalised as functions from singulary proposition functions to singulary proposition functions, and sentential adverbs, formalised as functions from propositions to propositions, originates from Thomason and Stalnaker's paper, "A semantic theory of adverbs", which is reprinted in this volume after Bellart's paper. Four sufficient conditions to be a sentential adverb are put forward: (a) inducing opaque contexts, (b) giving rise to scope ambiguities, (c) including a sentence modifier within its scope (d) being replaceable by "It is ADVERB true that".
Lewis' paper, "Adverbs of quantification", deals with adverbs like "always, sometimes, never, usually, frequently, rarely etc". Instead of taking them as quantifying over times or events, Lewis analyses these adverbs as quantifiers over cases (a case being an acceptable assignment of values to the free variables of the modified sentence): such adverbs bind unselectively all the free variables of the sentence in which they occur. The paper also contains a discussion of devices used to restrict the admissible assignments (e.g. "if"-clauses).
Approaches to connectives are illustrated by papers by Posner, Gazdar and Jennings. Posner's "Semantics and pragmatics of sentence connectives in natural language" defends a Gricean approach, which preserves a truth- conditional semantics for connectives and provides additional meanings by contextual enrichments driven by pragmatic considerations.
Gazdar's "A cross-categorial semantics for coordination" analyses non- sentential coordination as the union of properties denoted by NPs or VPs.
An original paper by Jennings, "The meaning of connectives" is a plea against truth-conditional semantic theory. The first part of the paper attempts to show that standard truth-conditional analysis of the connective "or", even augmented with Gricean mechanisms, cannot account for all the uses of it in English. More particularly, it is pointed out that "or" may have disjunctive as well as conjunctive uses. Then it is claimed that no satisfactory truth-conditional account may be provided for "but", since it can also have both disjunctive and conjunctive uses. Jennings outlines next a biological theory of meaning, whose purpose is to apply methods from population biology to the syntax of sentences. The origin of connectors is traced back to spatial or temporal meanings in order to individuate contemporary meanings. The meaning of an n-tuple predicate is taken to apply to a set of n-tuples of which it has been claimed that this predicate is true. Since, across time this set is uniformly increased, the meaning of the predicate becomes less and less specified. The next step is allotropy, where a sentence may yield two different syntactic representations with the same satisfaction-conditions. When syntactic differences do yield a difference in satisfaction- conditions, lexical items may receive a new meaning, in order to preserve the sameness of satisfaction conditions. These processes are exemplified with different connectors, the process of shortening of scope being discussed in more detail.
Non-declarative sentences are discussed in the next two papers. The core idea of Higginbotham's "Interrogatives" is that interrogative sentences express abstract questions, i.e. mutually exclusive and non-empty partitions of the possible states of the worlds. It is to abstract questions that reference is made in indirect questions.
Vanderveken's original article, "Success, satisfaction, and truth in the logic of speech acts and formal semantics" offers an up-to-date summary of the illocutionary logic (Searle and Vanderveken 1985; Vanderveken 1990, 1991). In this approach, primary units of meanings are illocutionary acts, whose force is determined by the conditions of success, and whose satisfaction is determined by the correspondence between the propositional content and the world (i.e. the truth of the propositional content) and the direction of this correspondence. The set of all possible illocutionary forces is defined recursively with the help of the six components of the illocutionary force: the illocutionary point, the mode of achievement, the conditions on the propositional content, the preparatory conditions, the sincerity condition and the degree of strength. The truth-conditions of the propositional content are formulated in way that allows the understanding senses of propositions without knowing their denotation: the truth-conditions of a proposition in a given circumstance are all those truth-conditions associated with its atomic propositions that make the proposition true with respect to that circumstance. New semantic relations are defined: illocutionary entailment (the speakers cannot perform an illocutionary act P without eo ipso performing an illocutionary act Q), truth-conditional entailment (P cannot be satisfied unless Q is satisfied), illocutionary entailment of satisfaction (the success of P entails the satisfaction of Q), and truth- conditional entailment of success (the satisfaction of P entails the success of Q).
The last paper of this part is Davidson seminal "The logical form of action sentences", which introduces events in the ontology of semantic theory, in order to allow entailments like the one from "John butters a toast in the bathroom at midnight" to "John butters a toast". Action names are thus taken to be alternative descriptions of events, i.e. predicates that take events as one of their arguments.
Part IV: Context Dependency The first paper of this part, devoted to contextual influences on semantics, is Kaplan's "Demonstratives". It contains an explicit and detailed statement of Kaplan's distinction between "the character of an expression", i.e. a function from contexts to contents, and "the content of an expression", i.e. a function from circumstances of evaluation to individuals. Demonstratives and indexical are taken to be directly referential, since they have a character that determine always the same function from context to content, and hence is not part of the content of the proposition expressed by sentences in which they occur. A further crucial point is that contingency and necessity are notions that are applicable to content only, while logical truth applies exclusively to the character.
In "Truth and Demonstratives", Weinstein attempts to formulate Tarskian truth-conditions for non-eternal sentences, i.e. sentences that contain demonstratives and indexicals. The new T-sentence contains, on the left- hand the truth-predicate that has for argument a meta-language variable corresponding to the utterance of s (s being the structural description of the sentence uttered), such that all the referents of demonstrative pronouns of s correspond to variables of the meta-language, and, on the right-hand, the translation in the meta-language, obtained by replacing the nth occurrence of demonstrative pronouns by the nth variable of the object-language.
Lewis' "Score-keeping in a language game", exposes the by-now famous idea of accommodation -- the psychological mechanism that corrects the interpretation of the conversational events in a way to fit the rules on which the kinematics of the conversation depends. The illustrations used are presupposition (the audience tends to assume the truth of the presupposed proposition), permission utterances (that shift the boundaries of what is permissible), definite descriptions (the audience tends to assume that there is a salient referent in the domain of discourse corresponding to it), the shift of the spatial point of reference by the verbs 'going' and 'coming', the vagueness in the applicability of some predicates (here the accommodation increases the standards of the applicability), the domain of application of modals, the truth of performative utterances and planning.
Carston's original paper "Explicature and semantics" is a Relevance- theoretical survey of the solutions provided to the problem of the existence of phonologically unrealised components of meaning. More particularly, she argues against the 'Indexicalist' view that postulates hidden variables within the logical form, sometimes bound and sometimes free: among the counter-arguments Carston puts forward are the unnecessary proliferation of hidden variables, the difficulty to set the boundary between cases where the variables are free and those where they are bound, and the difficulty to find hidden variables for some examples. She provides a Relevance-theoretical solution, in which the logical form of the sentence, -- output of the linguistic decoding module--, provides a mere blueprint to the interpretation module (which is part of a more general mind-reading module). Since each utterance comes with the presumption of optimal Relevance, the audience fills in indexicals (saturation), expands the logical form (free enrichment), and, adapts some concepts, all that done in way to derive the maximal cognitive benefit and minimise the processing effort.
Stainton's paper, "Quantifier phrases, meaningfulness 'in isolation', and ellipsis" is a good illustration of the problem of sub-sentential utterances. He argues that quantifiers phrase do not contribute to the sentence meaning without being part of it: they correspond to generalised quantifiers, i.e. to functions from sets to propositions. When they occur in isolation (e.g. 'Six apples' uttered in a grocery) the second set is provided pragmatically, through a Relevance-driven process. Such quantifier phrases in isolation are not to be assimilated to non-natural meaning, since their interpretation depends on their constituents, combined in a compositional way. Neither do they exhibit the behaviour peculiar to ellipsis, that is, of to phonologically shortened sequences that correspond to the same underlying syntactic structure as non- shortened counterparts.
The volume brings together very much of foundational literature of contemporary semantic theory, and manages to provide a broad sample of existing approaches and problems. The organisation is very neat and handy. In this respect, the book may be of great use to academics and advanced students in semantics. To be sure, an unfamiliar reader must be prevented that many articles do not reflect anymore the present state of the art. For instance, Discourse Representation Theory (DRT) has evolved in many aspects since Kamp's paper, originally published in 1981 (for instance, in the treatment of generalised quantification with the help of duplex conditions). Likewise, event semantics has known great changes since Davidson's original article.
Related to this point is the question whether it is possible to use the book as a textbook for students in semantics. It seems that this is one of the points of the extensive introduction: the topics addressed are systematically related to some of the articles. However, some parts seem too terse for a beginner, and too tedious for the scholar. More particularly, even if the second chapter "Linguistics and Logic" provides a very clear introduction to propositional and predicate calculus, the discussion on DRT is very dense and short, which makes me doubt that it could be of a great use to the unfamiliar reader, or bring new insights for a familiar one. Perhaps, the same remarks could apply to some parts of the chapter "Semantic and Context". It is regretful that the authors devote much of space to Relevance theory, already exposed in great detail Carston's paper, without mentioning alternative approaches like Recanati's (e.g. Recanati 1989, 2004) or default-interpretations ones (e.g. Levinson 2000).
Gillon's, Vanderveken's and Carston's original articles provide clear and good overviews in their respective domains. More controversial to my mind is Jennings paper. More particularly, I would like to point out a potential counter-argument to the first part of the paper, which contains a case against truth-conditional accounts of English connective 'or'. Jennings argues that, since in sentences like (1) 'or' has a conjunctive meaning, it cannot be analysed along the truth tables of the logical disjunction. (1) You may do this or you may do that.
This latter fact still does not show that a truth-theoretical account of disjunction is impossible. For instance, some authors (Zimmermann 2000; Geurts forthcoming) analyse disjunction as a conjunction of modalities, which gives a unified account for (1) and (2): (2) He will come or he will not come.
Hence, even if the origin of 'or' is traceable back to several distinct sources, it is still not shown that a unified formal account for the contemporary uses of 'or' is impossible.
As for the truth-conditional analysis of 'but', which Jennings claims to be impossible, it is a widely accepted fact in neo-Gricean theories that such elements do not pertain to the logico-conceptual structure of the utterance, but rather play a procedural role by restricting the possible contexts of interpretation (cf. Blakemore 1987). This is quite in line with Jennings claim that 'but'-clauses discriminate some operations on inferential effect produced by utterances.
To conclude, I believe that the volume will be of great use for anyone working in semantics. I recommend also its use for teaching purposes, for it contains the most important illustrations of virtually all the topics one may want to address in a course in semantics.
Barwise, J. and J. Perry (1983) Situations and attitudes. Cambridge, Mass.: MIT Press.
Blakemore, D. (1987) Semantic constraints on Relevance. Oxford: Blackwell.
Carston, R. (1988) "Implicature, explicature and truth-theoretic semantics." In R. M. Kempson (ed.) Mental Representations: the interface between language and reality. Cambridge: Cambridge University Press, pp. 155-81.
Carston, R. (2002) Thoughts and Utterances. The pragmatics of explicit communication. Oxford: Blackwell.
Fauconnier, G. (1994) Mental spaces: aspects of meaning construction in natural language. Cambridge: Cambridge University Press.
Geurts, B. (forthcoming) "Entertaining alternatives: disjunctions as modals."
Kamp, H. and U. Reyle (1993) From discourse to logic : introduction to modeltheoretic semantics of natural language, formal logic and discourse representation theory. Dordrecht: Kluwer.
Levinson, S. C. (2000) Presumptive meanings : the theory of generalized conversational implicature. Cambridge, Mass.: MIT Press.
Putnam, H. (1975) "The meaning of 'meaning'." In K. Gunderson (ed.) Language, Mind and Knowledge. Minneapolis: University of Minnesota Press, pp. 131-93.
Recanati, F. (1989) "The pragmatics of what is said." Mind and Language, 4, pp. 295-329.
Recanati, F. (2004) Literal meaning. Cambridge: Cambridge University Press.
Searle, J. R. and D. Vanderveken (1985) Foundations of Illocutionary Logic. Cambridge: Cambridge University Press.
Sperber, D. and D. Wilson (1995) Relevance: Communication and Cognition. Oxford: Blackwell.
Stanley, J. (2000) "Context and logical form." Linguistics and Philosophy, 23(4), pp. 391-434.
Vanderveken, D. (1990) Meaning and Speech Acts: Volume I: Principles of language use. Cambridge: Cambridge University Press.
Vanderveken, D. (1991) Meaning and Speech Acts: Volume II: Formal semantics of success and satisfaction. Cambridge: Cambridge University Press.
Zimmermann, E. (2000) "Free choice disjunction and epistemic possibility." Natural Language Semantics, 8, pp. 255 - 90.
ABOUT THE REVIEWER:
ABOUT THE REVIEWER
Mikhail Kissine is PhD student at the Laboratoire de Linguistique Textuelle et de Pragmatique Cognitive, Université Libre de Bruxelles with the financial support of the Fond National de la Recherche Scientifique. His thesis investigates the cognitive factors that make possible the interpretation of speech acts. It is an attempt to combine recent findings in cognitive linguistics with more traditional approaches to pragmatics and semantics, and to contribute to current debates on the semantics/pragmatics interface.