LINGUIST List 15.2535

Sat Sep 11 2004

Review: Philo of Lang/Semantics: Recanati (2003)

Editor for this issue: Naomi Ogasawara <>

What follows is a review or discussion note contributed to our Book Discussion Forum. We expect discussions to be informal and interactive; and the author of the book discussed is cordially invited to join in.

If you are interested in leading a book discussion, look for books announced on LINGUIST as "available for review." Then contact Sheila Dooley Collberg at


  • Luca Storto, Literal Meaning

    Message 1: Literal Meaning

    Date: Sat, 11 Sep 2004 16:48:58 -0400 (EDT)
    From: Luca Storto <>
    Subject: Literal Meaning

    Author: Recanati, Fran�ois Title: Literal Meaning Subtitle: The Very Idea Year: 2003 Publisher: Cambridge University Press Announced at

    Gianluca Storto, Center for Language Sciences, University of Rochester


    Recanati's monograph's purpose is two-fold. On the one hand, the book aims at reviving the debate between two opposite philosophical views concerning the semantics/pragmatics of natural language - Literalism vs. Contextualism - and at providing an overview of the logical space between these two extremes. Recanati's contention is that, while Literalism is currently the dominant view, the central tenet of this view - the contrast between literal truth conditions and speaker's meaning - must be rejected. Thus, some version of Contextualism must be adopted. On the other hand, the book aims at presenting the author's particular theory of the semantics/pragmatics of natural language (a Contextualist one, of course), and at distinguishing it from alternatives that have been proposed in the literature.

    In the INTRODUCTION, Recanati traces the historical roots of the debate - the two opposite camps of Ideal Language Philosophy and Ordinary Language Philosophy - and introduces the fundamental tenet of the Literalist view: the basic unit of meaning in natural language is the literal content of a sentence, which is conceptually distinct from what the speaker conveys with an utterance of the sentence itself.

    CHAPTER 1 opens with the observation that the simple dichotomy between sentence and speaker meaning is insufficient and a basic ''triad'' of notions of meaning is required: on the one hand the context-independent meaning of a sentence type must be distinguished from ''what is said'', i.e. the context-dependent proposition expressed by the sentence in a given context; on the other hand, ''what is said'' must be distinguished from what is merely ''conveyed'' by an utterance of the sentence, which includes more than what is literally said (''what is implicated''). The essence of the debate between Literalism and Contextualism lies in how to map this basic triad onto the dichotomy of sentence and speaker meaning, and in particular in whether the crucial level of ''what is said'' has more in common with the sentence-type meaning or with ''what is implicated''. Literalism maintains that ''what is said'' is very close to the sentence-type meaning: taken together, the two constitute the literal meaning of a sentence used in context, a semantic object which contrasts with the pragmatically-determined speaker's meaning, i.e. what the speaker intends to convey with the utterance of the sentence. Contextualism maintains that ''what is said'' has much in common with ''what is implicated'': both are pragmatically determined, and as such they constitute the speaker's meaning, a pragmatic object that contrasts with the purely semantic sentence(-type) meaning.

    In the remainder of the chapter Recanati presents the reader with the basic assumption of Literalism that is the main target of criticism in the book and with his alternative Contextualist position. ''Minimalism'' maintains that ''what is said'' differs from the context-independent sentence meaning only in that it incorporates contextual elements that are required in order to ''complete'' the meaning of the sentence and make it propositional: only mandatory contextual processes like saturation contribute to ''what is said''. This implies that in many cases the Minimalist has to maintain that there is a divergence between the literal truth conditions of a sentence and the truth conditions that speakers would intuitively ascribe to a given utterance of the sentence. Recanati argues that this divergence is a problem: not all the material that the Minimalist would place outside of ''what is said'' displays the ''conscious availability'' property of prototypical implicated meanings. In many of these alleged implicatures speakers are aware only of the ''final result'', not of the inferential processes that lead to this result or of some level of ''literal meaning'' on the basis of which such inferences are drawn. The alternative approach proposed by Recanati takes conscious availability to be constitutive of ''what is said'': a theory of language understanding should capture the level at which ''normal interpreters'' have intuitions concerning the truth-conditional content of utterances. Within this alternative proposal, ''primary'' pragmatic processes - i.e. contextual processes that take place below the level of conscious availability (the ''sub- personal level'') - contribute to the determination of ''what is said'', irrespective of whether they are mandatory or optional. Thus, contrary to Minimalism, truth-conditional interpretation is assumed to be pragmatic to a large extent.

    CHAPTER 2 details Recanati's view of how ''what is said'' is determined. According to Recanati, the literal meaning of the constituents of a sentence does not have a privileged status in the process of semantic composition: both literal meaning and meanings derived through the application of primary contextual processes - which include not only mandatory bottom-up processes like saturation, but even optional top- down processes like free enrichment, loosening, and semantic transfer - compete in parallel, and the most accessible interpretation (i.e. the interpretation that fits best in the broader context of the current discourse) is selected for semantic composition. Differently from Grice (1989), the derived interpretation is ''associatively'' derived from the ''literal'' interpretation, but it is not ''inferentially'' derived, since the literal interpretation of the whole sentence need not be calculated in order for its derived interpretation to be determined. Overall, the interpretive process is characterized as an interactive search for a coherent interpretation of the sentence with respect to the context of utterance, a process in which semantic frames/schemata play a crucial role in determining which among the possible (literal or derived) meanings of the constituents of a sentence are selected for semantic composition.

    In CHAPTER 3 Recanati points out an important difference between his proposal and the view of semantics/pragmatics advocated within Relevance Theory (Sperber and Wilson, 1986). While Recanati's theory draws a sharp distinction between non-inferential primary pragmatic processes and inference-based secondary pragmatic processes, Relevance Theory takes communication to be constitutively inferential: even the pragmatic processes involved in the derivation of the ''explicature'' of a sentence (the Relevance Theory name for ''what is said'') are based on inferences, albeit of a spontaneous, automatic and unconscious sort. Recanati argues that Relevance Theory's distinction between spontaneous inferences and explicit reasoning is correct, but that this does not correspond to the crucial distinction between unconscious (sub- personal) and conscious processes: even spontaneous non-explicit inferences can be consciously available to speakers in that not only the result of the inference, but also the inferential chain itself and its point of departure are available to the speaker. The result of such conscious processes, however, is not part of ''what is said''. In order to maintain the viability of his view, Recanati tackles some apparent counterexamples raised by proponents of Relevance Theory - the observation that the computation of ''what is said'' does not necessarily precede the derivation of implicatures (an inferential secondary pragmatic process), the phenomenon of mutual adjustment of explicature and implicature, and the case of ''implicated premisses'' - and further refines his notion of secondary pragmatic processes, by pointing out that the inferences on which these are based need not be conscious and explicit but can be tacit in a ''personal'' sense (compare: primary processes are based on inferences that are tacit in a ''sub-personal'' sense - the personal/sub-personal distinction is from Garcia-Carpintero 2001).

    CHAPTER 4 discusses a position that is in a way intermediate between a strict Minimalist approach and Recanati's own proposal. The ''Syncretic View'' holds that two legitimate notions of ''what is said'' can be distinguished, a purely semantic one and a pragmatic one, within a system that provides for four (rather than three) notions of meaning. According to Recanati, a position along these lines has been defended by various authors in the literature: the proposals by Salmon (1991), Bach (1994) and Soames (2002) are given as specific examples. Recanati argues that this position rests on a fundamental weakness: the very notion of ''what is said'' in a Minimalist sense cannot be maintained. Indeed, semantic interpretation - i.e. the deductive bottom-up computation of the meaning of a sentence based on the conventional meaning of its parts - does not deliver complete propositions. Semantic undeterminacy (beyond simple indexicality) is a constitutive feature of most natural-language expressions and constructions, thus it is either the case that the result of ''purely semantic'' interpretation is something gappy that must be complemented by pragmatic interpretation, or that semantic interpretation can be run only after pragmatic interpretation has pre-determined the values of semantically undetermined expressions. According to Recanati, ''minimal propositions'' has no real role to play in the analysis of language: they do not seem to have any psychological reality, and it is not even clear that they can be determined (by abstraction or otherwise) by inspecting the range of meanings expressed by a sentence in different contexts of utterance. The only purely semantic notion of ''what is said'' that seems viable is the ''reflexive proposition'' of Perry (2001), but this is virtually undistinguishable from the context-independent sentence-type meaning: the system collapses back into one that provides only for three notions of meaning.

    CHAPTER 5 addresses a potential issue with Recanati's own proposal. Can the intuitive folk-theoretical distinction between literal and non- literal uses of language be accounted for within a system in which ''what is said'' already incorporates the result of optional primary pragmatic processes? Recanati argues that non-literalness involves a perceived ''duality'' in the use of language: this can be either internal to the output of primary pragmatic processes (as in the case of figurative uses of language) or due to the perceived inferential link between ''what is said'' by a sentence and the somewhat different meaning derived in context through secondary pragmatic processes. These cases of perceived duality are ''exceptional'' situations even within a pragmatic approach to meaning: in normal cases the pragmatic processes that determine ''what is said'' do not generate such perceived duality. In effect, Recanati contends that the objection that a pragmatic stance to meaning obliterates the intuitive notion of non-literal uses of language rests on the mistaken Minimalist assumption that words in context usually express just what the rules and conventions of language mandate, i.e. something than can be determined in a purely bottom-up way.

    In CHAPTER 6 Recanati surveys the logical space of alternatives that could be pursued between Literalism and Contextualism. The extreme Literalist position holds that the truth conditions of a sentence are fixed by the rules of language independently of the speaker's meaning. Given the semantic undeterminacy that characterizes natural language, Recanati argues, this position is a definite non-starter. Weaker forms of Literalism obey Minimalism: we need to appeal to the speaker's meaning determining truth conditions only when the sentence itself demands it. The Syncretic View is a weak instance of this weaker Literalism: albeit optional pragmatic process affect ''what is said'', a narrower notion of semantic content - ''what is literally said'' - can be distinguished. A stronger instance of this weaker Literalism is the position that Recanati calls ''Indexicalism''. Indexicalism can be thought of as a research program: the attempt to re-analyse purported cases in which optional pragmatic processes contribute to truth conditions as instances of saturation, in which an open slot in the logical form of a sentence is contextually filled. Contextualist positions do not postulate a level of meaning which is unaffected by top-down factors. A weaker position - the position endorsed by Recanati, who calls it ''Quasi-Contextualism'' - considers such a level as theoretically useless and as lacking psychological reality, thus as dispensable. The stronger position - full-fledged Contextualism - denies that such a notion even makes sense: there can be no level of meaning that is both propositional and unaffected by top-down factors, since background assumptions (and thus pragmatic processes) are always necessary in order to determine truth conditions. Within full-fledged Contextualism the very distinction between mandatory and optional pragmatic processes becomes somewhat blurred.

    CHAPTER 7 is devoted to the rebuttal of what Recanati considers the strongest Literalist alternative to his Contextualist position: the Indexicalist position held by Stanley and colleagues (Stanley and Williamson, 1995; Stanley, 2000; Stanley and Szab�, 2000). In order to evaluate the two proposals it is crucial that some criterion is agreed upon in order to distinguish between the truth-conditional effects of saturation - the mandatory pragmatic process allowed for by Indexicalists - and other optional primary processes. Recanati compares his own Optionality Criterion and the alternative Binding Criterion proposed by Stanley: the two criteria converge in some cases but diverge in other cases, where the Optionality Criterion categorizes the pragmatic process at hand as optional while the Binding Criterion categorizes the same process as an instance of saturation. Recanati argues that the Binding Criterion is not reliable. He gives examples of utterances in which uncontroversially optional contextual ingredients of meaning can be intuitively ''bound'' by some operator, which according to the Binding Criterion should lead to the unpalatable conclusion that even those ingredients of meaning are linguistically articulated by a variable that is contextually saturated. The conclusion is that intuitive binding per se does not entail the presence of a free variable in the logical form of a sentence. Indeed, Recanati goes even further and claims that Indexicalist arguments against optional contextual ingredients of content rest on a basic fallacy: the assumption that the open variable which e.g. represents the location in the logical form of a quantificational sentence like ''Everywhere I go it rains'' is present even when the simple sentence ''It rains'' is uttered in isolation. Recanati shows that this implicit assumption is questionable, and - in particular - it is clearly wrong if e.g. locative modifiers are treated as constituents that contribute at the same time both a ''variadic function'' and the object that fills the new argument role introduced by the function itself.

    In CHAPTER 8 Recanati expounds further details of his own view of meaning in natural language and of the role played by context in determining meaning. Meaning in natural language has an irreducibly contextual nature because semantic evaluation requires not merely a content to evaluate, but also a ''circumstance'' against which to evaluate that content. This circumstance of evaluation is not part of the content articulated by a sentence, and must be provided by the context of utterance of the sentence itself. The basic unit of meaning in natural language is not the semantically incomplete content articulated by a sentence, but the semantically complete ''Austinian proposition'' consisting of this content and the circumstance against which this content is evaluated. Generalizing on modal and tense logic, Recanati sketches the basis of a Situation Theory in which (i) anything can count as a situation, provided that it makes sense given some sentence S to ask whether or not what S expresses is true in it, and (ii) sentences in general express only relativized propositions, absolute propositions and truth values being the domain of utterances (which provide a circumstance of evaluation). The chapter concludes with a comparison between Recanati's Situation-Theoretic view and the Syncretic View: while both postulate two distinct levels of content - the content articulated by the sentence and the richer content expressed by an utterance of the sentence - only Recanati's view allows for the articulated content to be a relativized proposition and/or to be affected by optional primary pragmatic processes.

    CHAPTER 9 is entirely concerned with the phenomenon of ''meaning modulation'', the process whereby the meaning of words are affected by the nature of the situation they are used to talk about. Recanati argues that (i) meaning modulation is truth-conditionally relevant, since words contribute their modulated senses to truth conditions, and not their pre-modulation meanings, and that (ii) modulation cannot be reduced to (pre-semantic) disambiguation and is essentially a top-down phenomenon, which is pragmatically rather than linguistically controlled. As such, the very existence of meaning modulation is by itself an argument against a Literalist approach to natural language meaning. Recanati reviews four Contextualist alternative approaches to meaning modulation, which progressively get further away from Literalism and move towards an extreme Contextualist conception of meaning in natural language. Of these four, only the ''Strong Optionality'' view takes meaning modulation to take place for contingent reasons, while the others take meaning modulation to be an ineliminable component of natural language semantics, either because it is required by the process of semantic composition that applies to word meanings (''Pragmatic Composition'' view), or because word meanings are not used at all in semantic composition, be it because they are not the right kind of object for it to apply to (''Wrong Format'' view), or because they do not really exist as such (''Meaning Eliminativism'' view). The second half of the chapter sketches in detail the basic ingredients and consequences of Meaning Eliminativism, and emphasizes a further dimension with respect to which context contributes to the determination of meaning: the central role of similarity (a context- dependent notion) in determining the extension of words. Recanati's contention is that Meaning Eliminativism, albeit probably too radical, is surprisingly very viable, and the alleged knock-down arguments that are usually raised against this view do not sound that devastating once they are stated explicitly.

    In the CONCLUSION Recanati summarizes his argument and addresses one residual issue (I refer the reader to the last section of the book for the latter). The conclusion (re-)states explicitly the core message that the monograph wants to convey: (i) Contextualism is not dead, since most arguments that have been made against it and in favor of Literalism/Minimalism rely on unstated premises that essentially amount to assuming Literalism, and (ii) once Contextualism is considered among the possible alternatives, it is Literalism that seems to face the most serious problems, since it is committed to a substantial and highly controversial conception of word and sentence meaning.


    Recanati's monograph constitutes an important contribution to the debate on the nature of meaning in natural language. The monograph is indeed quite successful in reviving a debate which is often taken to be settled in favor of Literalism in the linguistic semantic literature, and in characterizing in some detail the author's Contextualist view of meaning. However it is my opinion that, albeit quite successful, the monograph is not completely successful, and that its main drawbacks are due to its ambitious ''hybrid'' nature.

    As pointed out at the beginning, the monograph aims both at presenting and defending the author's position and at providing an overview of the logical space spanned by the debate between Literalism and Contextualism. Now, while the monograph goes a substantial way towards both aims, it can be argued that either aim could have been achieved in a more successful way had it been pursued as a separate endeavor. And this is not just because the two different aims could require two rather different ways of organizing the material to be presented (while the book is overall well-written, it might be hard at times for the reader to weave the contribution of a chapter into the ''big picture'' depicted by the monograph). Indeed, while the two aims definitely converge with respect to the discussion of Literalist approaches, they might diverge with respect to the Contextualist side, where a detailed overview of the logical landscape of alternatives can in principle distract from the precise characterization of the specific alternative that the author subscribes to, and vice versa.

    At various places the monograph seems to try to strike a compromise between these two contrasting requirements, with the result that while the reader is left with the clear feeling that Literalism in general - and Indexicalism in particular - cannot be maintained, she is not left with a well-defined alternative in hand, but just with the (partly uncharted) continuum of space that lies ''beyond'' the rejection of Minimalism, a portion of logical space that is possibly daunting, but is at the same time promising for future research.

    Admittedly, however, this might have been the author's main aim in writing this very inspiring monograph, which is bound to have a prominent place on the linguist/philosopher's bookshelf, both for the results that it achieves and for the issues that it raises (and in part frames) for future research.


    Bach, Kent 1994. Conversational implicature. Mind and Language, 9:124- 162.

    Garcia-Carpintero, Manuel 2001. Gricean rational reconstructions and the semantics/pragmatics distinction. Synth�se, 128:93-131.

    Grice, Paul 1989. Studies in the Way of Words. Cambridge, Mass.: Harvard University Press.

    Perry, John 2001. Reference and Reflexivity. Stanford, Cal.: CSLI Publications.

    Salmon, Nathan 1991. The pragmatic fallacy. Philosophical Studies, 63:83-97.

    Soames, Scott 2002. Beyond Rigidity: The Unfinished Semantic Agenda of 'Naming and Necessity'. Oxford: Oxford University Press.

    Sperber, Dan and Deirdre Wilson 1986. Relevance: Communication and Cognition. Cambridge, Mass.: Blackwell.

    Stanley, Jason 2000. Context and logical form. Linguistics and Philosophy, 23:391-434.

    Stanley, Jason and Zoltan Szab� 2000. On quantifier domain restriction. Mind and Language, 15:219-261.

    Stanley, Jason and Tim Williamson 1995. Quantifiers and context- dependence. Analysis, 55:291-295.


    Gianluca Storto is a NIH post-doctoral fellow at the Center for Language Sciences at the University of Rochester. His current research focuses on probing the semantics/pragmatics interface using psycholinguistic experimental methodologies (eye-tracking in particular). More generally, he is interested in all aspects of meaning and interpretation in natural language. His dissertation (UCLA, 2003) investigated the semantics of possessive constructions.