![Sample our Humanities journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5939/TOC-Subject-Sample-2021-Humanities.jpg"})
ABSTRACT
Much philosophical attention has been devoted to the truth predicates of natural language and their logic. However, lexical truth predicates are neither necessary nor sufficient for a truth-attribution to occur, which warrants closer attention to the grammar of truth attribution. A unified analysis of five constructions is offered here, in two of which the lexical truth predicate occurs (It's true that John left and That John left is true), while in the three remaining, it does not (John left; It seems that John left; and It's that John left). This analysis is philosophically significant for four reasons. First, it explains why speakers of natural language find standard instances of Tarski-inspired equivalences (e.g. That John left is true iff John left) intuitively compelling. Second, it derives the widespread ‘deflationist’ intuition that truth has no substantive content. Third, insofar as the deflationist sees insights on truth as flowing from understanding our practice of truth attribution, it furthers the deflationist agenda through a new analysis of such attributions. Finally, it advances the philosophical project of the ‘naturalization’ of truth by reducing our understanding of truth to our competence in the grammar of truth, as an aspect of our biological endowment.
KEYWORDS:
Disclosure statement
No potential conflict of interest was reported by the authors.
Notes
1 The arguments presented here only hold if (1) is uttered with falling intonation typically associated with assertions. A change in intonation (e.g. a sentence-final rise) affects the assertoric force of the utterance. This will have to occur if a speaker seeks to avoid a commitment to the truth of what she says, just as a change in the grammar may correlate with a change in force as for example with conjectures (e.g. ‘if John left … ’, ‘Suppose John left’, etc.). Assertions that occur in play-acting do not cause problems for the present account: play-acting an assertion is no more to assert something, than Herodes’ palace on stage is Herodes’ palace.
2 Relatedly, it is doubtful whether, as one referee wonders, truth attribution could follow from a norm of assertion, like that one should assert p only if one knows p (since knows p entails that p is true). While norms can be flouted, the attribution of truth, if the grammatical configuration is right and normal intonation is used, cannot be. The response to an assertion: ‘You don't really know that, why do you assert it?’ is coherent. It objects to an assertion being made, but does not question that it was an assertion (and as such an assertion of truth). Lackey (Citation2007, 599) points to examples of ‘selfless assertion’, which does not express a speaker's belief and hence is independent of ‘knowledge’. This is not to say there is no interface between grammar and the norms of assertion, which in fact we consider crucial.
3 As Davidson (Citation2005) stresses, only whole sentences are assigned truth values. Clauses embedded in complex sentences can of course refer to a fact that is presupposed to obtain, but this is different from asserting the proposition expressed by the embedded clause. Sheehan and Hinzen (Citation2011) show that the grammar of clauses that are ‘factive’ is not a counterexample to the present claim, as both that grammar and the semantics of such clauses are different from those of non-factive embedded clauses. As also pointed out there, the explanation of factivity is not lexical-semantic either.
4 There are two ways to technically implement the claim that the predication of truth is grammatically configured just as there have been two ways to linguistically encode the predication relation between a subject and its predicate. It may be merely a function of a structural configuration (as for example in Williams' (Citation1980) analysis of small clauses) or it may be mediated by a dedicated head (lexical or abstract) whose function is to relate the subject in its specifier position to the predicate in its complement position (as for example in Stowell’s (Citation1981) or Den Dikken’s (Citation2006) analyses). In sections 5–6 we opt for the former approach, identifying this relation with a structural configuration: see also Hinzen (Citation2003). For abstract functional heads that have been posited in the grammar of truth, see e.g. Laka (Citation1994) or Höhle (Citation1992).
5 One cannot state a truth using a word only, including the word true. One can of course say the single word True. But this will only be felicitous if it is a reply to a truth attribution, which needs to take a sentential form. This account raises the important question of whether a truth attribution can be effected through sentence fragments, which as such fail to replicate the core elements of the grammatical configurations reviewed above and below. Hinzen (Citation2015) contains an argument against this option, suggesting that differences in sentential grammar always correlate with a difference in meaning.
6 Though there is an unsettled controversy whether clauses that appear to be subjects are indeed subjects. Emonds (Citation1972) takes such clauses to always be topics (cf. also Koster Citation1978; Stowell Citation1981; Safir Citation1985) whereas Rosenbaum (Citation1967); Emonds (Citation1972); Delahunty (Citation1983) and Davies and Dubinsky (Citation2009) argue that they are genuine subjects (see Lohndal Citation2013, for a recent overview).
7 In line with its status as a grammatical notion, Tense can remain expletive (i.e., temporally uninterpreted). This is the case in counterfactual conditional, for example, where past tense marking is compatible with an adverb of present time (If I had a car now) and hence is fake (Iatridou Citation2000).
8 This is of course not to say that a formal compositional semantics would not assign a semantic value to ‘true’, or to ‘is true’ as a constituent and a perfectly acceptable grammatical predicate of a grammatically referential sentential subject. But the content, in a referential sense, of a proposition does not change if the lexical truth predicate is added to it: the truth predicate applied to a proposition returns the same proposition. Moreover, as noted, ‘true’ appears to be the only predicate of this sort (as opposed to say probable, possible, … ), showing its special status in the design of language.
9 E.g., adjuncts are opaque for extraction, whereas extraction is possible out of arguments, and adjuncts are not canonical recipients of structural Case (see e.g. Chomsky Citation2004, 116–18).
10 The former reading is eased with stress on ‘appears’ and forced if ‘only’ is added to ‘seems’. For an example, suppose e.g. that John is a house guest and we open the door to our apartment. Nobody there. ‘He seems to have left’. This would usually be understood in line with the reading (10). But now suppose we also know that John is a fugitive who is desperate to hide and has been caught hiding before. Again, we open the door to the apartment. ‘He seems to have left’, we say, stressing ‘seems’. This is reading (9).
11 Note that it is not the case that presuppositions require the proposition to be in the common ground. They allow for accommodation. In the sentence: I am sorry that I am late. I had to take my daughter to the doctor, to the speaker presupposes both that he is indeed late and that he has a daughter, but neither of these facts need to be known by the addressee.
12 (11) is not fully grammatical for some speakers of English, though it is for most we have consulted (and it improves with ‘just’ inserted before ‘that’). It is also fine in German (Es ist (nur), dass sie geweint hat) or Italian (C’è que Gianni e partito). For some, (12) with ‘the truth’ is out. This correlates with the fact that the proposition embedded in (11) is factive (presupposed as true and referenced as a fact), rather than asserted as a truth. We expect this, given the general fact that no proposition, as long as it is embedded at all, is asserted as true (assertion being a matrix phenomenon). The closest we get to truth-theoretic force in an embedded context is factivity (referencing a fact taken to exist) or else an approximation of truth ascription given the right structural configuration (see also further below).
13 See Betti (Citation2015) for an excellent exposition of other problems with Fine's claim.
14 The view that it is ‘covertly’ a noun phrase is critically discussed and rejected in Sheehan and Hinzen (Citation2011).
15 That the mayor is the predicate of the SC and as such non-referential is confirmed by the answer to the question: ‘Who's the mayor?’ The natural answer is: ‘it's John’, not ‘he's John’ (Mikkelsen Citation2005). Similarly, when trying to determine everyone's role in the city, one can ask: ‘What is John?’ to receive the answer ‘John is the mayor.’ In this case, what serves as the question word for a predicate. Relatedly, in languages like Catalan, the mayor would be cliticized by the neuter clitic ho, which does not pick out referential expressions.
16 That, of course, is equally true for proper names, all of which can (but need not) function as predicates: e.g., ‘The track needs more Usain Bolts’ (International Herald Tribune). See further Hinzen (Citation2016).
17 See Haegeman and Ürögdi (Citation2010); Sheehan and Hinzen Citation2011. None of that prevents either CPs or DPs from functioning referentially, in some syntactic contexts, as the same authors also document, unless referentiality, a grammatical concept, is confused with nominality (which does not entail referentiality as just noted).
18 The question still remains why the grammar would choose the pronominal element it as the predicate, rather than, say, the dummy verb do which may be used as a dummy auxiliary (John left and Mary did too). It could be that the reason is that as a verb, do is intrinsically associated with an event argument, and hence cannot remain dummy when inserted in a lexical predicate position. No such event argument is associated with the pronominal it.
19 In the case of (8), this circumstance would only be consistent with a focal stress on seems (It (only) seems that John left), which immediately shifts the reading to the other one distinguished above (seems as appears/looks like).
20 To avoid major misunderstandings, note that we are crucially not saying that lexical proper names necessarily refer directly to an individual in the speech context. Of course, they don't, in general, but they do so in a certain grammatical configuration, namely one that supports a vocative interpretation, which is the case in (51) but not in (52). Specifically, in (52) John functions as an argument, and argumenthood is in complementary distribution with a vocative interpretation. This supports the view that reference is grammatically configured and hence is compatible with the claim that truth, a special kind of reference, is grammatically configured. (51) is only felicitous when John is present in the speech context and called upon. (52) is only felicitous when he is not.
21 For some necessary qualifications, see Schwager Citation2006, who shows that in some contexts embedded imperatives are possible. However, they are usually more constrained than embedded interrogatives or declaratives, which are already constrained. According to Kaufmann (Citation2014), they appear in reported speech contexts and come with syntactic and semantic restrictions. Thus, imperatives can only embed when particular conditions are met. What derives these conditions, and how they relate to the present proposal is beyond the scope of this paper.
22 Assuming that interrogatives are derived by transforming (i.e., un-doing) the very configuration in question: John is here → Is [John __ here].
23 Of course, there is an embedded Tense (specified on left). However, at the matrix level, Tense is evaluated as indexical in relation to the point of speech (the Now of the speaker), and embedded Tenses are evaluated differently, namely at least partially in relation to the temporal reference point established by the matrix Tense. In particular, (6) has a reading where what Bill said is John left, hence where the leaving took place prior to the saying (Giorgi Citation2010).
24 Even if this was true of natural language and ordinary talk, one referee points out that there is a flurry of recent work in metaphysics devoted to ‘grounding’ facts (Schaffer Citation2009; Rosen Citation2010; Fine Citation2012). While this point deserves a wider discussion than is possible in the confines of this paper, the metaphysical issue arises from the desire to ‘ground’ some (not so fundamental) facts in other (more fundamental) ones. This issue of connecting ‘levels’ does not entail the existence of a more fundamental formal-ontological category like the putative ‘pact’. At whatever level of facts reality's rock-bottom is reached, we cannot dig deeper.
25 This also explains why it is much more felicitous to answer (i) with (ii), than to answer (iii) with (iv):
You studied in Leeds.
True.
It is true that you studied in Leeds.
*True.
26 Again this appears to be orthogonal to the metaphysician's desire to ground (some) facts in others. See footnote 24.