Evidence for the RT assumptions
2.3.2 The Problems of Connectivity and Non-Truth-Conditional Meaning: on implicatures and procedures
2.3.2.1 Non-truth-conditional meaning and speech acts
This latter problem is one Grice wrestled with. He introduced the term ‘conventional implicature’ to refer to the type of non-truth-conditional meaning encoded by e.g. but, so and moreover. According to Grice, in an utterance such as ‘He is an Englishman, he is, therefore, brave’,
“I have certainly committed myself, by virtue of the meaning of my words, to its being the case that his being brave is a consequence of (follows from) his being an Englishman. But while I have said that | he is an Englishman, and said that he is brave, I do not want to say that I have SAID (in the favored sense) that it follows from his being an Englishman that he is brave, though I have certainly indicated, and so implicated, that this is so. I do not want to say that my utterance of this sentence would be, STRICTLY SPEAKING, false should the consequence in question
fail to hold.” (Grice 1975: 44-45; small capitals in original)
As Bach (1999: 330; also Rosales Sequeiros 2012: 79) points out, therefore is not the best example, as it does seem to be truth-conditional – that is to say, the proposition would be false if the consequence should fail to hold. Let’s take another example:
(40) He penned that 80 years ago, but those comments could apply perfectly today. (from ESPNFC.com, ‘Do Arsenal have fewer leaders than their title rivals and does it matter?’, by Michael Cox. 1st of March 2016.61).
On Grice’s account, there would be some contrast between ‘being written 80 years ago’ and ‘being relevant today’, but this contrast would be implied rather than part of his ‘what is said’. The difference between this type of implicature and the more general conversational implicatures discussed above is that a conventional implicature depends on the presence of a specific linguistic expression (such as but), while conversational implicatures, according to Grice, arise from his maxims in some way (Neale 1992: 523- 524). Conventional implicatures are a semantic way of introducing implicatures – in other words, Grice assumes that semantics can contribute both to ‘what is said’ and ‘what is implied’. As such, his conception of semantics “cross-cuts the saying- implicating distinction” (Sperber & Wilson 2005/2012: 8).62 This was his way of salvaging
a truth-conditional conception of ‘what is said’: non-truth-conditional meaning can only contribute to ‘what is implied’ (Neale 1992: 521).
Grice assumes that (40) communicates two speech acts: one central (amounting to the proposition that ‘he’ (referring to a former manager of the football club Arsenal) penned a certain comment 80 years ago and that his comment could still apply today), and one ancillary (where the speaker is “indicating or suggesting” that there is a certain contrast between the two parts of the central speech act (Neale 1992: 523)). A conventional implicature, then, communicates a full proposition (see also Potts 2005: 213; Bach 1999: 332). This view is shared by Rieber (1997), whose conception of conventional implicatures is broadly Gricean, but “stronger”, as Bach (1999: 362) puts it. According to Rieber, conventional implicatures are ‘tacit performatives’ – that is, they
61 Accessed via http://www.espnfc.com/barclays-premier-league/23/blog/post/2818906/arsenal-criticism-
for-lack-of-leaders-is-missing-the-point, on the 2nd of March, 2016.
62 As Bach (1999: 329-330) notes, Grice’s idea of conventional implicature finds its roots in Frege’s work. Frege
(1892: 167) argues that although and but are non-truth-conditional and can hence be deleted without affecting truth conditions, but that “the light in which the clause is placed by the conjunction might then easily appear unsuitable, as if a song with a sad subject were to be sung in a lively fashion.” See also Horn (2013: 153), who states that Frege uses the verb andeuten to mean ‘implicate conventionally’ in a Gricean sense.
Note also that Grice’s conventional implicature is very different from Karttunen & Peters’ (1979) conception of the term – they argue that a conventional implicature is a presupposition (see also Potts 2005: 13; Bach 1999: 332). Potts’ (2005) notion of conventional implicature is again very different from both Grice’s and Karttunen & Peters’ conception. He argues that appositions, sentence adverbials (such as unfortunately) and expressives (such as friggin’ in ‘My friggin’ bike tire is flat again!’) are true conventional implicatures, and hence reanalyzes Grice’s conventional implicature as well (Potts 2005: 13).
encode a second-order speech act, but a strictly performative one. For example, but would encode something like the following in (69) (1997: 55):
(40’) He penned that 80 years and (I suggest that this contrasts) those comments could apply perfectly today.
Like Grice (and coherence theorists like Mann & Thompson – see §2.3.2.2 infra), Rieber assumes that conventional implicatures encode full propositions which say something about the relation between the propositions conjoined in (40’).
There are several problems with both Grice’s and Rieber’s account, outlined fully in Bach (1999) – I will discuss only the most significant ones here. The main problem with Grice’s account is that conventional implicatures are not part of ‘what is implied’. This might seem intuitively right (as Clark (2013: 315) points out, if something is expressed explicitly, how can it be part of what is implied?), but Bach also proposes an ‘IQ test’ – if a linguistic element is part of an indirect quotation (IQ) where the utterance which contains them is represented, it is obviously part of what was said (see also Blakemore 2002: 53):
(40’’) a. He penned that 80 years ago, but those comments could apply perfectly today.
b. Michael Cox said that Herbert Chapman penned that 80 years ago, but that those comments could apply perfectly today.
But, then, can “occur straightforwardly in indirect quotation” and hence passes Bach’s IQ test (Bach 1999: 339). Even on Grice’s narrow conception of ‘what is said’ (i.e., a conception which is not fully pragmatically enriched; see §2.3.1.1), conventional implicatures are part of ‘what is said’ and not ‘what is implied’.
Rieber’s analysis of conventional implicatures also entails that they are non- cancelable. Yet one of the crucial aspects of implicatures is that they are cancelable. Take (40) again:
(40’’’) He penned that 80 years ago, and those comments could apply perfectly today – although I do not want to suggest that there is a contrast between being old and being relevant today.
(40) might have given rise to the implicature that there is a contrast between being old and being relevant today – even without the conventional implicature but, as (40’’’) demonstrates. By adding the ‘although’-clause, the speaker cancels that implicature. On Rieber’s account, however, this would lead to a logical contradiction, as Bach (1999: 364) notes:
(40’’’’) #He penned that 80 years ago and (I suggest that this contrasts) those comments could apply perfectly today – although I do not want to suggest that there is a contrast between being old and being relevant today.
In addition, it is not clear how the performative verb ‘suggest’ should be understood. Rieber (1997: 54) says that it means “tentatively put forward a proposition”, but there need not be anything tentative about suggesting a contrast with but or, for instance, a consequence with so. As such, “there is no unique performative with which” different conventional implicatures are “synonymous” (Bach 1999: 363). Finally, on Rieber’s account in (40’), conventional implicatures also do not pass Bach’s IQ test. As such, conventional implicatures, on Rieber’s account, are also part of (the truth-evaluable) ‘what is said’ (1999: 364).
Bach (1999: 340-341) notes that not all conventional implicatures pass the IQ test (moreover and in other words, for instance, flunk it). His response is to argue that these are not conventional implicatures, but ‘utterance modifiers’ which “are verhicles for the performance of second-order speech acts”. This analysis is compatible with Grice’s analysis, as Bach himself indicates, except that, on Bach’s account, their contribution is ‘explicitly indicated’ (and not implicated). As such, Bach places his analysis squarely in the speech-act tradition which Grice is a part of as well (Blakemore 2002: 39). Bach’s analysis of but is, in fact, quite Gricean on the whole: he proposes that but works as an operator on the conjoined propositions. As such, he assumes that someone who hears (40) will recover a proposition “with something like the concept CONTRAST as a constituent” (Blakemore 2002: 55). In other words, Grice, Rieber and Bach all assume that but leads the hearer to recover a full proposition.
There are several problems with Bach’s analysis. One of the most obvious ones is that his IQ test doesn’t work – Hall (2004: 218) points out that, while utterance-internal but indeed passes the IQ test, utterance-initial but flunks it (see also Iten 2005: 59); Carston (2002a: 176-177) argues that some of Bach’s ‘utterance modifiers’ do pass the IQ test. As such, the IQ test is not a productive or convincing method for determining whether a given item is part of what is said explicitly. Moreover, Iten (2005: 91-93) points out that Bach’s analysis of utterances as being able to express more than one proposition leads to problems with truth-conditionality and explicit content. Cohen’s (1971) famous scope test, according to which an element is part of the explicit content and truth-conditional if it falls under the scope of a logical operator (Carston 2003: 311-312) can only take one proposition into account. On Bach’s analysis, however, both expressions such as but and so on the one hand and ‘utterance modifiers’ on the other are truth-conditional and contribute to explicit content (1999: 328) – they point to a second-order proposition which is less important than the primary one, but which can be true or false in its own right (1999: 346-347). If there is a single proposition, it can be placed under the scope of a logical operator (if or because, for instance) – everything which falls under the scope
test is part of the proposition, while everything which falls outside of it is part of what is implied and non-truth-conditional (Iten 2005: 91; see also (20’) in §1.1.3.2):
(41) a. Peter went out although it was raining.
b. Because Peter went out although it was raining, he got wet.
Although does not fall under the scope of the logical operator because, and is hence non-truth-conditional:
“The assumption that Peter’s going out was unexpected, given the rain or that Peter’s going out and it raining are somehow incompatible is not understood to be part of the cause of Peter’s getting wet or conditions under which Peter will have got wet” (Iten 2005: 89)
But the notion of a multipropositional utterance breaks the scope test: (42) *Because Peter went out, it was raining, he got wet. (from Iten 2005: 91)
This is ungrammatical – there is no way to tell whether the propositions fall under the scope of the logical operator. In other words, the test ‘if item X falls under the scope of a logical operator, it is part of the explicit proposition’ does not work if there is more than one proposition – as Bach assumes is the case when, for example, but is added to a single-proposition utterance. However, as Blakemore (2002: 77) points out, this conundrum is only relevant if we assume that truth-conditionality and explicitly communicated assumptions are inextricably linked. If truth-conditionality were an “essential part” of distinguishing explicitly from implicitly communicated assumptions (as Bach does), a broken scope test would be hard to overcome. On a RT view, on the other hand, this problem does not arise, because explicit content is distinguished from implicit content not in that it is truth-conditional but in that it is derived ‘locally’, as a development of the logical form (in contrast with implicatures, which are derived ‘globally’, taking the explicature as one of the premises in the inferential process) (ibid.). This RT view is what I’ll take a look at in the next section.