2.5 Semantic Incompleteness
2.5.1 Covert Structure
One strategy, familiar from the literature on quantifier domain restriction, is to posit covert context-sensitive variables in the syntax of apparently semantically incom- plete sentences such that, relative to a context, they do indeed determine complete propositions. For example, Stanley (2000) claims that “the effects of context on the truth-conditional interpretation of an assertion are restricted to assigning the values to elements in the expression uttered” and that “[e]ach such element brings with it rules governing what context can and cannot assign to it” (p. 396).
Consider again the case of (17), which requires the specification of a task, or task- related item, that Mike is indicated to have finished in order to yield a corresponding
37 Unlike with ‘finished’, however, syntactically, ‘completed’ does not appear to allow for a verb
phrase.
38 Consider as further examples: ‘Katie used a rock’, ‘Jennifer and Jessica are ready’ or ‘I’m a lot like
proposition. On Stanley’s approach the idea would be, roughly, that the logical form of (17) includes an unpronounced element in its syntactic structure, which for our purposes we can represent as follows.
(17+) Mike has finished x
The variable in (17+) is taken to correspond to the task (or task-related item) that Mike has finished, and “the value of the relevant variable is supplied by context,” according to Stanley (p. 417). In this way, an apparently semantically incomplete sentence such as (17) would in fact yield a proposition relative to a formal context.
Such an account appears prima facie plausible in the case of genitive constructions
such as (11), where a rule might hold that context must assign a particular type of
relation holding between Mike and the motorcycle in question. Similar considerations
apply to possessive pronoun such as (12). It is also somewhat plausible in the case of (17), where a rule might hold that context can assign either a task or a task-related item (e.g., a drink) indicated to have been finished by Mike. To complete the account,
the formal context would need to be expanded further to include a relation parameter
and a task (-related item)parameter.
Nonetheless, it is implausible that the account works for all cases of semantic incompleteness. It is very unclear, for example, given the variety of completions with which it is compatible, what rule might govern the assignment of values to a covert variable in the logical form of (13), which, as (14-16) demonstrate, admits both prepositional and complementizer phrases as continuations. What feature of the context of use could plausibly be represented as a parameter in the formal context to account for (14-16)?
Perhaps (intended) goal is a plausible candidate in the case of ‘enough’. For example, if the (intended) goal in a given context is to get a pail of water, then, relative to that
context, (13) determines (14) as its propositional content. However, it’s not obvious whose (intended) goal would be relevant. Getting a pail of water is presumably not the speaker’s (intended) goal. What if filling a pail was not Jack and Jill’s objective in climbing? (13) could still be true in such a case. It’s not clear that seeing the valley behind the hill needs to be construed as a goal at all, though if it is, it’s presumably not the speaker’s objective, and it need not be Jack and Jill’s either. Likewise, though the objective might have been to become lightheaded, it’s not at all clear that that needs to be the (intended) goal of either the speaker or Jack and (/or) Jill. Who else’s goals could possibly be relevant in such cases? Does the context merely “provide” a
2.5 Semantic Incompleteness | 53
goal? The answers to these questions are in no way obvious on Stanley’s approach.39 We’re not given enough explanation.
Despite Stanley’s claims that sentences like (17) “have certain propertiesthat are best explained via the supposition that their true structures contain unpronounced
pronominal elements” (p. 410; emphasis added), the covert structure approach is neither well-motivated nor the best explanation of semantically incomplete sentences. Cappelen and Lepore (2004), for example, point out that Stanley’s argument would massively overgenerate the presence of covert variables (both within and across a range of constructions); lead to psychologically implausible conclusions about a range of affected expressions; predict the availability of anaphoric readings for certain expressions that do not appear to admit such readings; and predict a priori truths and falsehoods where there would appear to be none.40 Several others have raised further criticisms, which are well documented and need not be rehearsed here.41
What seems to emerge from these criticisms is that, in the case of a semantically incomplete sentence, where additional information is needed to yield a full-blooded, truth-evaluable proposition, nothing from the standpoint of compositional semantics
per se appears to motivate or demand positing covert indexical elements corresponding
to whatever additional information is needed to yield a complete proposition (which, as noted, can vary considerably depending on the case). Perhaps it is ultimately incoherent for an interlocutor toassert a semantically incomplete, or non-propositional,
entity in light of our best theory of linguistic communication.42 Nevertheless, having distinguished semantic value and propositional content, any such condition on assertion or communication need not carry over to the notion of semantic value.
Compositional semantics is restricted to assigning semantic values to the syntactic constituents of a given expression.43 Plausibly, then, any covert elements posited by the semanticist need to be licensed by syntax. In fact, Stanley takes the covert variable approach to be motivated primarily by syntactic rather than semantic considerations. Nevertheless, while such a strategy has the rare merit of bringing syntactic consid- erations to bear on semantic analyses, it has the unfortunate burden of justifying syntactic structure that is not itself motivated by current syntactic theory. Collins
39 Perhaps the answers are covert elements of the approach. However, It seems as though this tack
would create an unduly proliferation of parameters to deal with the range of cases of semantic incompleteness (e.g., ‘lost’, ‘ready’, etc.).
40 In particular, see ch. 6.
41 Cf. Bach (2000), Carston (2002), Recanati (2002), and Dever (2006) for further discussion. 42 Though perhaps not. There may be no important theory of assertion as such; cf. Cappelen (2011)
for discussion.
43 Unless it can be explained why interpretation is not required, as in the case of certain trace elements,
(2007) explains in detail why current theoretical accounts yield syntactic structures that are too sparse to admit the covert variables envisaged by Stanley, and thus effectively rule out the possibility of their presence. The upshot is that, “if recent generative theory is correct, Stanley’s account is incorrect, or at least lacks the support he claims for it, for there is too little syntax” (p. 826), and hence that Stanley’s “underlying assumptions about syntax are at odds with recent generative theory” (p. 829).
The irony of Stanley’s account is that he explicitly disavows revisionary treatments of syntactic structure on the part of semantic theorists and accepts that “[i]n semantic interpretation, one may never postulate hidden structure that is inconsistent with correct syntactic theory” (p. 397).44 And yet the covert structure approach seems to do just that. Without containing bona fide constituents that require the contextual assignment of values, however, it’s not at all obvious how the semantic values of sentences like (17) are meant to interact in any systematic way with context to yield a particular propositional content in virtue of their intrinsic semantic properties. If the purported “elements in the expression uttered” do not in fact exist there would appear to be no “rules governing what context can and cannot assign” to them. Hence, it is not all obvious that the covert structure approach can rescue the determination principle.