‘Truth Predicates’ in Natural Language

Abstract

The aim of this paper is to take a closer look at the actual semantic behavior of what appear to be truth predicates in natural language and to re-assess the way they could motivate particular philosophical views. The paper will draw a distinction between two types of apparent truth predicates: type 1 truth predicates such as in English true and correct and type 2 truth predicates such as English is the case. It will establish the following points:

• 1. Type 1 truth predicates are true predicates, predicated of a representational objects of some sort, such as sentences, propositions, and entities of the sort of beliefs and assertions.

• 2. That-clauses with type 1 truth predicates do not act as referential terms, referring to propositions as truth bearers, but rather specify the content of contextually given attitudinal objects, such as ‘John’s belief that S’ or ‘Mary’s claim that S’.

• 3. Type 2 ‘truth predicates’ do not in fact act as truth predicates, but rather express the relation of truthmaking, relating a situation or ‘case’ to the content of a that-clause.

Appendix: Two Further ‘Truth Predicates’

In this appendix, I will briefly discuss two further apparent truth predicates in English, is a fact and is the truth, focusing on their rather special syntax and semantics.

• 4. Is a fact

Is a fact appears to act as a truth predicate below:

• (1) That the sun is shining is a fact.

We will see, though, that the semantics of the is a fact -construction is fundamentally different from that of type 1 or type 2 predicates: it involves neither attribution of truth to a representational object nor the expression of the truth-making relation, but rather a specification of the ‘content’ of a fact.

Let us look at the linguistic properties of the is a fact-construction. Like is true-sentences, is a fact-sentences allow for extraposition:

• (2) a. That S is a fact.

• b. It is a fact that S.

Moreover, like is true-sentences and unlike is the case-sentences, is a fact-sentences resist location modifiers and adverbs of quantification:

• (3) a.??? In our firm, it is never a fact that someone gets fired without explanation.

• b.??? It was twice a fact that someone was absent.

This means that that S in that S is a fact must be propositionally complete.

There are differences, though, between is true and is a fact. Unlike is true, is a fact allows only for simple quantifiers and pronouns in subject position and not for referential NPs:²⁸

• (4) a. * John’s belief is a fact.

• b. * That sentence is a fact.

• (5) a. Nothing is a fact.

• b. It is raining. That is a fact.

Unlike the case in is the case, a fact in is a fact is an ordinary indefinite NP, allowing for adjectival modifiers, relative clauses, and anaphora support:

• (6) a. That S is an interesting fact.

• b. That S is a fact that I had never noticed.

• c. That S is a fact. That fact is hardly known.

All this suggests that the is a fact-construction reflects the Identity Theory of truth: That S is a fact is true just in case that S picks out a fact. This would also account for the possibility of negation:

• (7) That S is certainly not a fact.

However, the identity-theoretic analysis is implausible. That S by itself cannot stand for a fact, not only because of the lack of referential independence of that-clauses discussed earlier. Let us look at the sentences below, which display the very same construction:

• (8) a. That S is a possibility.

• b. That S is a common belief.

If that S could by itself stand for a fact, then (8a) and (8b) could have a reading on which they are false just because S is true, since a fact is neither a possibility nor a belief. A fact is not a possibility since the possibility that S exists in circumstances in which S is not true. Moreover, a belief obviously is not a fact.

More plausibly, the that-clause in (1) occurs nonreferentially and serves to specify the ‘content’ of a fact. That is, (1) expresses a relation of specification that holds between the content of the that-clause in subject position and a fact (and not predication of the property of being a fact of the referent of the that-clause). This will be the very same semantic relation that obtains between fact and the that-clause in the fact that the sun is shining. The same holds of course for (8a) and (8b).²⁹

• 5. Is the truth

The truth predicate is the truth is a very puzzling one both from a semantic and a syntactic point of view:

• (9) That John is guilty is the truth.

Obviously, what the subject clause in (9) denotes cannot literally be ‘the truth’. It could not make up the one and only ‘truth’; there are lots of ‘truths’.

In what follows, I will identify a range of semantic and syntactic properties of the construction and point at the kind of syntactic and semantic analysis that it most plausibly has.

It is tempting to take the truth in this context to stand for the unique contextually relevant ‘truth’ (that is, true proposition) or to act as a predicatively used contextually restricted definite description. But a contextual restriction driving the interpretation of the truth in (9) is implausible. For (9) to be acceptable, no particular context is required that would restrict the denotation of the truth. The truth does not behave like predicatively used contextually restricted NPs as in the examples below:

• (10) a. This chair is the yellow chair.

• b. This piece of furniture is the yellow chair.

Unlike (9), (10a) and (10b) do require a particular previous discourse context that was about a unique yellow chair.

In fact, is the truth belongs to a different construction than that of a subject-predicate sentence, as well as that of that S is a fact. The truth has neither the status of a predicatively used NP nor of a referential or quantificational NP (whatever the view of definite NPs may be).

Three properties distinguish is the truth from ordinary predicates. First, is the truth allows subject-predicate inversion, as seen in (11), unlike ordinary predicates, such as type 1 and type 2 truth predicates, as in (12a) and (12b):

• (11) a. That John will not return is the truth.

• b. The truth is that John will not return.

• (12) a. * True/ Correct/ Right is that John will not return.

• b. * The case is that John will not return.

Second, unlike predicates taking clausal subjects, is the truth does not allow for extraposition:³⁰

• (13) * It is the truth that John will not return.

Third, is the truth requires a definite determiner in the truth, unlike ordinary predicates such as is a chair:

• (14) a. * A truth is that he will not return.

• b. * That John will not return is a truth.

Given these three properties, we can conclude that is the truth is not a syntactic predicate taking clausal subjects. Moreover, the truth in that construction does not act as an ordinary definite NP used predicatively. Otherwise the restriction to definiteness would be unexpected.

The construction more plausibly is a type of specificational sentence (Higgins 1979). Specificational sentences come in two sorts: with a free relative clause in subject position, as in (15a), and with a definite NP in subject position, as in (16a) (Higgins 1979):

• (15) a. What John did was kiss Mary.

• b. Kiss Mary is what John did.

• (16) a. The best player is John.

• b. John is the best player.

(15b) and (16b) illustrate, inversion is possible in both cases.

Semantically, specificational sentences have been analysed in one of two ways:

[1] as expressing a question-answer relationship, with the subject acting as a concealed question and the postcopula expression partially specifying an answer (the Question-Answer Analysis) and [2] as expressing an identity among possibly higher-level semantic values (the Identity Analysis).³¹ It turns out that neither analysis can be right for specificational sentences with that-clauses in general. To see this, let us look at some more familiar kinds of specificational sentences involving that-clauses:

• (17) a. John’s claim is that it is raining.

• b. That it is raining is John’s claim.

• (18) a. The idea is that there will be a party.

• b. That there will be a party is the idea.

It is not entirely obvious how the Question-Answer Analysis applies to specificational sentences with that-clauses. On that analysis, the subject of (18a) might stand for a question of the sort ‘what idea is there?’ or perhaps of the sort ‘what is the idea?’. The postcopula NP presumably will partially specify an answer of the sort ‘the idea that there will be a party’.

This kind of analysis is not generally applicable, however. In particular, it is not applicable to sentences with the truth as subject. On that analysis, the postcopula that-clause in (9) would partially specify an answer of the sort ‘the truth that John is guilty’. But the truth that John is guilty is ungrammatical: truth does not accept that-clauses. The difficulty arises with other specificational sentences as well. Higgins (1979) already observed that in specificational sentences, the subject and a postcopula that-clause need not be able to form an NP. Thus, (19a) is a specificational sentence as well, but (19b) is ungrammatical:

• (19) a. The proof that John is guilty is that his fingerprints are on the knife.

• b. * the proof that John is guilty that his fingerprints are on the knife

The Identity Analysis does not straightforwardly apply to specificational sentences with that-clauses either. What a that-clause generally is taken to stand for is not a claim, an idea, or a proof. A that-clause specifies the content of such entities, but is not identical to them. The relation expressed by a specificational sentence with a that-clause could only be that of content specification, not that of identity.

A further question that the is the truth-construction raises is, why is the subject a definite NP when there need not be a unique entity it stands for? The Question-Answer Analysis would say that an NP of the sort the fact that S or the claim that S is obligatorily definite, since indefinites such as a fact that S or a claim that S are unacceptable. But we have seen that that analysis was not generally applicable to specificational sentences with that-clauses. There is a more plausible way of explaining the obligatory definiteness. It is a general requirement that specificational subjects always be definite, illustrated below (Heycock and Kroch 1999):³²

• (20) a.??? A good player is John.

• b.??? A problem is that it is raining.

Given that (9) is in fact the inverted structure, this means that the obligatory definiteness of the truth in is the truth would be an instance of a more general condition on the subject of specificational sentences. Of course, the definiteness condition on specificational subjects needs to be explained itself (perhaps by associating the subject of a specificational sentence with a particular semantic role, for example by attributing it an anaphoric status relating to what is at least implicitly under discussion).

Let us then summarize this rather inconclusive discussion by stating that is the truth is a pseudo-truth predicate involving a complex syntactic structure whose semantics is far from well understood.