Pragmatics, Semantics and the Case of Scalar Implicatures

Kenneth A. Loparo

• Publisher: Palgrave Macmillan
• Year: 2014
• ISBN:
  9781137333285
• Topic: Languages & Linguistics
• Subtopic: Psycolinguistics
• Index: Languages & Linguistics

This book is an advanced debate on the nature of scalar implicatures, one of the most popular topics in philosophical linguistics in the last 20 years. Leading theorists in the field offer an up-to-date presentation of the subject in a way that will help readers to orient themselves in the vast literature on the topic.

1

Some Remarks on the Scalar Implicatures Debate

Salvatore Pistoia Reda

1.1 Introduction

The purpose of this book is to provide a discussion forum for participants in the recent scalar implicatures debate. The bulk of contributions are formal, except for two. Larry Horn’s chapter, which opens the book, provides an interesting and remarkably documented investigation on the phenomenon’s historic development, evoking proto-pragmatic accounts of De Morgan and Mill, and even extending discussion to Grice’s manner maxim. An experimental chapter is contributed by Alexandre Cremers and Emmanuel Chemla. In their study, they focus on the processing cost involved in the derivation of both direct (that is, from some to not all) and indirect (that is, from not all to some) scalar implicatures, and argue that the very same processing signature generalizes to all subclasses of the phenomenon. Based on their findings, they also provide a useful discussion on how to compare scalar implicatures with other kinds of inferences and on how to identify the subprocesses responsible for the processing cost of scalar implicatures. Within the formal contributions is a really useful survey, written by Michael Franke and Gerhard Jäger, of game theoretic models that captures, what they call, pragmatic back-and-forth reasoning about mutual beliefs and linguistic behavior. This contribution is a worthwhile read for those with no immediate interest in the scalar implicatures debate (for game theoretic discussion may provide useful insights to understanding the mechanisms behind a number of linguistic phenomena) as well as those researchers who are involved (for, from a pragmatic perspective, a rigorous theory of communicative rationality is needed to throw some light on the Gricean project on communication). The remaining contributions, written by Uli Sauerland, Roni Katzir, Benjamin Spector and Giorgio Magri, more directly interact with each other, or develop discussions found in previous work by the same authors. So, in the remaining sections of this introduction I provide some pretty general discussion on the topics that these contributors are tackling, so that the broad picture will not be missed by readers who are approaching literature on scalar implicatures for the first time, or those from a different background. For reasons of space, I will discuss just three important components ascribable to scalar implicatures: the exhaustivity operator, the generation of scalar alternatives and, what I’ll call the ‘avoid-contradiction’ procedure. I will briefly try to show how the authors involved in the debate develop only partially co-extensional theories of scalar implicatures starting from a common range of core facts. For presentation purposes, in discussing the third component I will make use of a classic tool in formal logic, that is the post-Aristotelian square of oppositions, while Spector and Magri, whose chapters are particularly related to the component, do not. However, I believe that doing so may help readers to get a clearer picture of the basic cases from which Spector’s and Magri’s subtle discussions on blindness and double strengthening originate.

1.2 The exhaustivity operator

Consider one of the most important components of scalar implicature, that is the so-called exhaustivity operator EXH (for various implementations see Groenendijk and Stokhof (1984); Krifka (1995); van Rooij and Schulz (2004); Chierchia (2006); Fox (2007)). The strengthened meaning of a scalar sentence is generally said to be obtained by virtue of the application of EXH to the uttered sentence. For instance, suppose that the scalar sentence contains the existential quantifier, the result of the application is that the literal meaning of the sentence gets conjoined with the negation of the excludable alternative sentence obtained by replacement of the existential with the universal quantifier.

Note that, while I’m assuming that set Excl(ϕ) of excludable alternatives is composed of one member only in this case, this need not be so; in fact, authors typically consider sentences obtained by replacement with medium quantifiers many and most as being potential alternatives of an existential sentence. As it is easy to see, the meaning of EXH as presented here appears to be similar, at least in certain respects, to the meaning classically assigned to focus particle ONLY (see the locus classicus Rooth (1992)). There is one important difference though, concerning the inferential status of the underlying modified sentence, often called the praeiacens or ‘prejacent’. For, while ϕ would follow from the formula ONLY(ϕ) with presuppositional status,¹ the very same sentence would be part of the asserted content of the formula EXH(ϕ).² One finds many passages, especially in grammarian writings, stressing this point:

While with [ONLY] the requirement that the prejacent be true is a presupposition, with [EXH] this requirement should be part of the assertive component. (Fox, 2007, p. 79)

[EXH], so defined, is not exactly equivalent to [ONLY], for [ONLY] is usually assumed to trigger various presuppositions which [EXH], according to the above entry, does not. (Chierchia et al., 2012, p. 2304)

Now, be that as it may, what really should concern us here is what the exhaustivity operator stands for in the above representation. On the one hand, proponents of a neo-Gricean theory contend that the exhaustivity operator ought to be interpreted as a shortened form for an exhaustivity reasoning. This is sometimes called a Gricean reasoning, and its earliest formulation dates back at least to Levinson (1983). Note that common discussion in the neo-Gricean tradition, perhaps inappropriately, assumes it to be a hearer’s reasoning, while from Grice’s letters one infers that it should be (at least) available to the speaker. Thus, reflecting on the status of the outcome of such reasoning, Horn (2009, p. 224) correctly claims that “[. . . ] by definition an implicature is an aspect of speaker’s meaning, not hearer’s interpretation”.³ Building on the intuitions of Groenendijk and Stokhof (1984), van Rooij and Schulz (2004) provided what seems to be the clearest recent instance of a neo-Gricean interpretation of the exhaustivity operator (but see also discussion in Krifka (1995, §, 2.2), as they are claiming in the following relevant passage:

the generalization of Groenendijk and Stokhofs (1984) approach we are going to develop can be interpreted as formalizing some of the maxims of conversation. Thereby it links Groenendijk and Stokhofs (1984) exhaustivity operator to Grices theory of conversational implicatures. (van Rooij and Schulz, 2004, p. 493)

One crucial feature of a Gricean reasoning is that it is activated by an act of saying something. But an act of saying something cannot be performed by means of sub-constituents of a sentence (so-called ‘unasserted’ clause). Then, if one is willing to maintain that the exhaustivity operator is to be interpreted as being equivalent to the Gricean reasoning, one cannot allow applications of the operator within the scope of other logical operators in a sentence. That means, one has to limit allowable applications at root levels, or globally. This argument was used by Cohen (1971), in his early attempt to criticize Grice’s Conversationalist Hypothesis (see also Cohen (1977)), and by Anscombre and Ducrot (1983) who, in the same years, resisted “[. . . ] the straightforward application of Gricean ideas to scalar phenomena”, Récanati (2003, p. 302).⁴

On the other hand, the interpretation given by grammarians is one according to which the exhaustivity operator is divorced from the reasoning, and thus freed from the associated constraints: “exhaustification would be more than just a way of expressing Gricean reasoning compactly. It would become a grammatical device”, Chierchia et al. (2012, p. 2304). More precisely, the exhaustivity operator is now a hidden counterpart of focus particle ONLY, though syntactically (in some versions, mandatorily so) realized. As a consequence of this interpretation, the exhaustivity operator can also be applied at internal levels. In other words, applications are also allowed in which the exhaustivity operator appears within the scope of arbitrarily many other logical operators, while having immediate (i.e., without intervening operators) scope over the propositional constituent containing the scalar term.

Now, as was first observed in Sauerland (2012, § 4), internal applications of the exhaustivity operator could be accounted for by theories of lexical ambiguities. Thus, via Levinson (2000), one might argue that, contrary to classical assumptions, neo-Griceans can allow both global and internal applications of the operator (but see Pistoia Reda (2014, chapter II) for foundational doubts on the theoretical justification of the resulting theory). Suppose that option is taken, then one needs a further testing ground to discriminate between predictions made by grammarians and predictions made by neo-Griceans. To this effect, in his contribution to this volume Sauerland is taking into consideration what he calls ‘intermediate implicatures’, i.e., scalar implicatures generated by virtue of intermediate applications of the exhaustivity operator. These differ from internal applications in one single respect, that now at least one logical operator appears within the scope of the exhaustivity operator. As it is clear, while grammarians can allow for intermediate applications, there seems to be no viable extension of a neo-Gricean theory that would obtain the same result.

1.3 The generation of scalar alternatives

Consider a second component of scalar implicatures, namely the generation of scalar alternatives included in set Alt(ϕ). Note that set Excl(ϕ) of excludable alternatives is just a subset of Alt(ϕ) (here I’m adopting standard terminology following, e.g., Magri (2009)). Scalar alternatives are obtained by replacing relevant items with corresponding expressions as part of suitable sources of substitution. Tradition stemming from Horn (1972) and Gazdar (1979) assumes that such sources of substitutions are to be regarded as lexical entities, so-called lexical or Hornian scales, that is ordered sets of semantically comparable lexical items (see also Levinson (2000, § 2.2) for discussion).⁵ Note that, as demonstrated by Sauerland (2004), lexical scales are really partial orders...