Herman Cappelen

Fixing Language is a book about ways in which language (and other representational devices) can be defective and improved. In all parts of philosophy there are philosophers who criticize the concepts we have and propose ways to improve them. Once one notices this about philosophy, it’s easy to see that revisionist projects occur in a range of other intellectual disciplines and in ordinary life. That fact gives rise to a cluster of questions: How does the process of conceptual amelioration work? What are the limits of revision (how much revision is too much)? How does the process of revision fit into an overall theory of language and communication? This book is an effort to answer those questions. In so doing, it is also an attempt to draw attention to a tradition in twentieth- and twenty-first-century philosophy that isn’t sufficiently recognized as a unified tradition. There’s a straight intellectual line from Frege (e.g. of the Begriffsschrift) and Carnap to a cluster of contemporary work that isn’t typically seen as closely related: much work on gender and race, revisionism about truth, revisionists about moral language, and revisionists in metaphysics and philosophy of mind. These views all have common core commitments: revision is both possible and important. They also face common challenges: how is amelioration done, what assumptions need to be made, e.g., about the nature of concepts, and what are the limits of revision?

This chapter develops and defends the Master Argument for Conceptual Engineering: (1) If W is a word that has a meaning M, then there are many similar meanings, M1,M2,...,Mn, W could have. (2) We have no good reason to think that the meaning that W ended up with is the best meaning W could have: there will typically be indefinitely many alternative meanings that would be better meanings for W. (3) When we speak, think, and theorize it’s important to make sure our words have as good meanings as possible. (4) As a corollary: when doing philosophy, we should try to find good meanings for core philosophical terms and they will typically not be the meanings those words as a matter of fact have. (5) So no matter what topic a philosopher is concerned with, she should assess and ameliorate the meanings of central terms. I respond to seven objections to this argument.

This chapter both explores and critiques Paul Saka’s (1998) theory of mentioning and quoting, in which the phenomena are both distinguished and claimed to be multiply ambiguous. This chapter shows that both aspects of Paul Saka’s view is wrong. This has implications for Paul Saka’s criticism of alternative theories, especially Donald Davidson’s (1979) demonstrative theory. The argument proceeds by considering Paul Saka’s approach as semantic in nature (despite the author’s own confusion on this point) since it would not be a rival to the theories it opposes were it a ‘purely pragmatic’ account.

This chapter discusses the issue of quotation, which is revisited with a special focus on the application of the general theory to the case of mixed quotation. In light of Recanati’s paper ‘Open Quotation’, published in 2001, mixed quotation is explored as a means of capturing certain aspects of the interaction between semantic and non-semantic content. After an overview of both approaches, it is argued that a semantic account of mixed quotation is better than a non-semantic one. It is also argued that mixed quotation itself can be used as a test case for drawing the semantics-pragmatics distinction.

This chapter sets out to provide a unified semantic account of the phenomenon of quotation and its various incarnations. It provides a discussion of direct, indirect, pure, and mixed quotation. It starts by showing evidence for the interaction of the different types of quotation and identifies four constraints in accordance with these interactions that an adequate theory needs to satisfy. A joint demonstrative account of direct and pure quotation is combined with that of indirect quotation along the lines of Davidson (1968, 1979), in order to successfully satisfy the constraints of interaction and provide a common explanation of their opacity.

Traditional theories of communication in the philosophy of language and linguistics leave little room for the role of a creative interpreter. This chapter challenges the idea that the interpreter has no essential role to play in the correctness of the interpretation process. In opposition, it argues for a content co-creating role for the interpreter or content relativism. Content relativism is the view that not only are there interpretation-sensitive terms in natural language (terms whose correct interpretation varies across interpreters) but these terms are pervasive. Content relativism is shown to be fruitful in connection with other theories (such as Kaplan’s framework in ‘Demonstratives’ (1977)), speech act pluralism, utterance pluralism, contextual vagueness, and recent work on indexicals, conditionals, and epistemic modals. Finally, the view is defended against arguments to the effect that it has trouble accounting for assertion as it can be shown to enrich our understanding of the phenomenon.

Theories of assertion take many forms in the literature. In this chapter, assertion, as a proper subset of saying, is argued to be nothing more than a philosopher’s invention. The No-Assertion view, which claims that there is no theoretical use for a separate account of assertion, is defended against various theories which base assertions in terms of norms, effects, causes, or commitments on sayings. Although, the focus is on normative views of assertion such as Williamson’s, the model is general enough to provide arguments against other types of theories in favour of assertion. The No-Assertion view holds that there is no need for an additional or distinct category of sayings called ‘assertions’ as the act of saying, (which is similar to Austinian _locutionary acts_) combined with context-variable norms, causes, effects, and commitments is sufficient.

This chapter starts with a puzzle about how a semantic theory can both incorporate the high level of context sensitivity of ‘what is said’ (O1) and the ability speakers have to express the same claim/assertion/proposition across a wide range of contexts (O2). The problem is that these two observations seem to be at odds with one another, and any semantic theory that postulates O1 must do so in a way that respects O2. Various proposed solutions are considered including the denials of O2, shared content, and absolute truth. These solutions are all shown to be unacceptable. A positive solution—Pluralistic Minimalism—is proposed. This option requires us to reject the idea that the variability of ‘what is said’ is relevant to semantics and the view that indirect reports report on semantic contents.

This chapter sets out to defend semantic minimalism. Semantic minimalism holds _inter alia_ that there are relatively few context-sensitive expressions (essentially just the obvious cases involving indexicals, demonstratives, etc.) and that the context of utterance has little to no effect on semantic propositions or truth conditions. Against this thesis, opponents have launched two common objections, (1) it entails that comparative adjectives are context insensitive and (2) it renders the role of semantics in communication inexplicable. In response, it is argued that (1) can be accepted and explained through Speech Act Pluralism while (2) is more problematic for the opponents of semantic minimalism as their accounts of communicated content fail certain tests of psychological reality.

The thesis that the meaning of a sentence determines its truth conditions identifies two classes of sentences, i.e., those with truth conditions _tout court_ and those with semantic values relative to their context-sensitive linguistic items. There are two opposing views: moderate and radical pragmatics. In this chapter, it is argued that there is no principled distinction to be drawn between these two positions and furthermore that both positions are internally inconsistent. In addition, evidence for these positions relies on either an incompleteness charge or a context shifting claim, neither of which establish the thesis that meaning does not determine truth conditions.

This chapter deals with the phenomenon of context shifting arguments, i.e., arguments to the effect that unambiguous, non-vague, and non-elliptical utterances of a sentence can have in differing truth-conditions (and even truth-values). Three tests for genuine context sensitivity are proposed and defended, namely the _Inter-Contextual Disquotational test_, the _Inter-Contextual Disquotational Indirect Report test_, and the _Collective Descriptions test_. These tests are used to show that the majority of cases claimed to be context shifting, by contextualists and radical contextualists alike, are, in fact, not genuinely so. The tests are robust for uncontroversial cases such as indexicals, pronouns, and adverbs of time while the controversial cases involving ‘know’, ‘red’, and ‘weighs 80kg’ fail. A potential objection in terms of _Contextual Salience Absorption_ is entertained and shown to be more detrimental to the contextualist cause than helpful. Finally, two diagnoses are identified in terms of Kaplanian monsters and direct quotation.

An indexical is an expression whose meaning stays constant across contexts while its reference shifts. This chapter confronts the claim that many constructions of natural language contain ‘hidden’ indexical elements. It does so, in the first place, by proffering a _reductio_ argument against Stanley and Szabo’s ‘argument from binding’, which states that there is syntactic evidence to support the claim that hidden domain variables are required to account for domain restriction in quantifier and noun phrases in general. Lastly, it is argued that positing hidden indexical expressions incurs two unmet constraints, namely participation in anaphoric relationships and the generation of a priori truths.

Initially, the semantics of quantifiers seems to suggest a context-insensitive account of their interpretation unlike, say, indexical or demonstrative noun phrases. However, most philosophers think that incomplete quantifiers and the need for domain restriction shows that all quantifiers are, in fact, context sensitive. This chapter carefully establishes a set of intuitions prevalent in the literature in favour of a context-sensitive analysis of the semantics of quantifiers. These intuitions are shown to be misguided through a Kripke test, a technique which establishes the consistency of a set of assumed counter-examples and a given phenomenon by showing that in a language in which this phenomenon is stipulated, these counter-examples persist. This task is achieved by using the connection (or rather lack thereof) between reports of what is said/asserted/claimed and semantic content to show that the context-sensitive intuitions are the result of the _non-interpretive_, _non-semantic_ effects of true indirect reports.

In this chapter, the dialectic of the book is set out. In 1995, the authors began a collaboration on two series of papers: one series of papers began with their 1997 paper ‘On an Alleged Connection between the Theory of Meaning and Indirect Speech’. The other series started with their 1997 paper ‘Varieties of Quotation’. The first series led ultimately to their 2005 book _Insensitive Semantics_ and the second to their 2007 book _Language Turned on Itself_. This volume collects many of those joint efforts combined with some of their single-authored papers along the same lines. In this introduction they briefly locate these papers in their historical context, provide some reflections on their legacy, and say a bit about how they see the connection between the papers and the two books.

This chapter discusses reasons why the practice of quotation is of central significance to the philosophy of language and to philosophy more generally. It argues that quotation is an important but neglected philosophical topic. The chapter sketches seven topics that can be illuminated significantly by understanding quotation, and together constitute a strong _prima facie case_ for its philosophical significance. These are: understanding metalinguistic discourse, opacity, the language-world connection, what is said, compositionality, the semantics-pragmatics divide; and indexicality, context sensitivity, and monsters.

This chapter examines the Proper Name and Definite Description Theories. According to Quine and Tarski, quotations lack semantic structure (their position is often misleadingly called the “Proper Name Theory”.) According to (other passages by) Quine and Tarski, and also Geach, quotations are structured definite descriptions. The Definite Description Theory fails for reasons very similar to those that undermine the Proper Name Theory. It is argued that quotation appears to provide a counter-example to Neale's NC theory: Every meaningful noun phrase (NP) in natural language is either a semantically unstructured, referring expression (singular term) or else a semantically structured, restricted quantifier. Quotation expressions don't fit easily into either of the two categories Neale claims that all noun phrases belong to (and quotation expressions are, at least sometimes, noun phrases). The versions of the Proper Name and Description Theories discussed in this chapter were at the centre of the semantics for quotation until Davidson's paper “Quotation”. Since then there have been few, if any, attempts to resurrect these theories.

This chapter focuses on Donald Davidson's paper “Quotation”, which is the most discussed and influential paper on quotation. Davidson's view is alternatively called “the Demonstrative Theory of Quotation” or “the Paratactic Theory of Quotation”. The strengths and weaknesses of the Demonstrative Theory are discussed. It is argued that seen in its historical context, Davidson's proposal is extraordinarily interesting and highly original. It took into account data that no other theory at the time could accommodate, and took the device of quotation more seriously than any other theory at the time did. However, it is now time to leave that theory firmly behind. The subject has moved to a point where we are much clearer on the totality of data and the relevant facts about syntax and semantics than Davidson was. As a result, the Demonstrative Theory is no longer a useful starting point.

Some theorists argue that the data surrounding the practice of mixed quotation establish that it is not a real form of quotation (and in so doing they present a particular interpretation of (D7) and (D8)). In particular, these theorists argue that quotation marks in mixed quotes can be dropped without a loss of meaning. This chapter examines these data and their various interpretations. It is argued that “impure quotes” do not pose a challenge to a theory of quotation, though they do present a challenge for a theory of the _saying relation_.

The thesis of Quotation Context Sensitivity (QCS) states that: let S be a sentence with a quotation Q. Two utterances, u and u', of S can express different propositions because Q in u and in u' quotes different items. This chapter refutes any semantic interpretation of QCS and shows that even a pragmatic construal of the data encounters problems. The first section presents four arguments against a semantic construal of QCS. These arguments are not intended to deny data that support QCS, but only to deny that they should be explained by a semantic theory. The next section presents some pragmatic options for explaining these data in support of QCS, and then argues that some of these variability data resist even pragmatic explanation. That result leaves an explanatory gap that is addressed in Chapter 12.

The chapter introduces an obvious (though surprisingly overlooked) distinction between signs and expressions and invoke this distinction and the Minimal Theory of Chapter 11 to underwrite an account of quotation expression individuation. It then exploits this account of quotation expression individuation in order to explain (away) the recalcitrant variability data from Chapter 7. The chapter explains why so many authors (and speakers) are misled into thinking a single quotation expression can be used to pick out distinct quotable items on different occasions.

The following disquotational schema for quotation was noted in Chapter 3: (QS) “e” quotes “e” (where “e” is replaceable by _any_ quotable item). The Minimal Theory of Quotation (MT) is the view that QS is _the_ semantic rule for quotation. MT presumes a principle of Containment — Containment: For any quotable item e, if a quotation expression Q quotes e, then e is contained in Q. This chapter develops the semantic ramifications of MT in some detail; in particular, it aims to explain how MT is compatible with the practices of pure, direct, and mixed quotation.

This chapter presents the central data, (D1)-(D12), on the theory of quotation. It argues that an adequate theory of quotation should say _something_ about how to accommodate the data surrounding (D1)-(D12). The following are discussed: D1 - opacity; D2 - quantifying in; D3 - infinitude; D4 - extant lexicon; D5 - the proximity constraint and the disquotational scheme; D6 - syntactic chameleonism; D7 - simultaneous use and mention in mixed quotation; D8 - indexicals inside mixed quotation; D9 - context sensitivity in mixed quotation; D10 - iterability in quotation; D11 - quotation without quotation marks; ad D12 - “impure” direct quotes.

A central thesis in the philosophy of language, at least since Frege and Russell, is that the meaning of (declarative) sentences determines the conditions under which they are true. Of course, declarative sentences in English divide into two classes — those that have truth conditions _tout court_, and those that have truth conditions only relative to certain contextual parameters being fixed. With respect to the latter class, identifiable linguistic features determine which contextual parameters go with which parts of the sentence in question. A number of authors have challenged, in varying degrees of scope, the thesis that sentence meaning determines truth conditions. Let's call the extreme critical position ‘radical pragmatics’ (RP). According to RP, sentences lack truth conditions. Sentences have meaning, according to RP, but only utterances have truth conditions. Sentences do not have truth conditions _in vacuo_, but only in the context of an utterance, only against a background of ‘normal circumstances’. Philosophers who explicitly subscribe to an all out RP are John Searle and Charles Travis. Both these philosophers allude to Wittgenstein and Austin as their chief influences. According to a more moderate pragmatics (MP), the truth conditions of many assertions go beyond anything that a semantics legitimately assigns to the sentences uttered. Many philosophers and linguists (Bach, Sperber/Wilson, Carston, Recanati, Perry, Crimmins, Stich) endorse MP to some degree. If either RP or MP is correct, then semantics is not about truth conditions. The polemical strategy of this chapter is to argue MP is an unstable position. To wit, there is no principled line to draw between MP and its fanatical cousin RP. So if you accept the arguments or intuitions behind MP, you inevitably will wind up endorsing RP. Secondly, it argues that neither RP nor MP has valid arguments against the view that sentence meaning determines truth conditions. And, lastly, it is argued that RP (and so MP) is wrong. Showing that RP and MP, and their collective arguments against truth conditional semantics are flawed is not a defence of truth conditional semantics directly, but it does fend off what has become a rather irksome opponent.