Philosophy And Linguistics

Kumiko Murasugi, Robert Stainton

• Publisher: Taylor & Francis
• Year: 2019
• Print ISBN: 9780367282868
• eBook ISBN: 9781000302806
• Edition: 1
• Topic: Philosophy
• Subtopic: Philosophy History & Theory

In the 1960s and 1970s questions about the semantics of natural languages were of central concern to the vast majority of analytic philosophers. The work of Chomsky, Davidson, Grice, Donnellan, Kaplan, Kripke and Putnam was widely read by non-specialists. The three main branches of linguistics that are of special philosophical significance-syntax,

PART ONE
Philosophy, Syntax, and Semantics

1
Perceptual Reports Revisited

James Higginbotham

1. Introduction

In Higginbotham (1983), an article prompted by several important observations and a theoretical proposal due to Barwise (1981), I considered a number of issues in the syntax and semantics of perceptual reports with so-called "naked infinitive" (NI) complements in English, exemplified by (1):

• (1) John sees Mary enter.

Since publishing that article I have learned, partly on my own and partly owing to the kind offices of others, of work on the subject that overlapped with some of my views. In addition, there have been several further papers and published articles on the subject of perceptual reports, some of them directly critical of my proposal, or Barwise's, or both. In this article I will consider antecedents to Barwise's article and mine, and the subsequent critical literature. Naturally, my own reflections on the subject have continued over the years, and I will also explain here where these have led me.

After a brief general discussion in section 2 following, I will consider in turn a number of the syntactic and semantic points about NI complements that have came under recent discussion. Sections 3-10 are intended to serve as a critical guide to some of the major issues in an area where the last word has surely not been said.

2. General Discussion

Barwise (1981) pointed out that none of the standard semantic theories allowed embedded sentences to have the peculiar combination of semantic properties of NI complements. These properties might be summed up in the statement that (i) perceptual reports are both factive and almost exten- sional, in that they permit substitutivity of singular terms, and (with some exceptions to be discussed below) fail to exhibit the ambiguities of scope normally associated with clausal arguments, but (ii) conspicuously fail to permit substitution salva veritate of predicates, and, as a striking corollary, fail to permit the interchange of logical equivalents. How could a construction that is almost extensional fail to be fully extensional? And how could a non-extensional construction show the substitutivity properties and uniqueness of scope that NI complements show? If NI complements are clausal arguments, neither the Fregean perspective nor any of the variants of possible-worlds semantics seems equipped to answer these questions.

Barwsse was led to propose (2) as giving, under appropriate interpretation of its key terms, the truth conditions of (1):

• (2) [∃s] (John sees s & s supports the truth of 'Mary enter')

The thing John sees, if (1) is true, is a (visual) scene. Scenes may be represented as partial models, mentioning some but generally not all objects, and some of the properties they have and relations they stand in. What it is for a scene to support the truth of a sentence is a matter for definition, from which the semantic properties of NI complements should follow. But then the theory of scenes and situations, a piece of substantive metaphysics, must be combined with standard semantic methods if the properties of language are to be theoretically described.

In Higginbotham (1983), I responded to Barwise's points by sketching what I called the individual events account of NI complements. On this account, the assumption that NI complements are clausal arguments is abandoned. Instead, it is proposed that the NI constituent, the expression 'Mary enter' of (1), has the semantics of an indefinite description of events, whose quantifiable place, following Davidson, is supposed to be provided by a hidden open position within the predicate. I will follow my subsequent practice and call this the E-position of the predicate. On the individual events account, (1) can be rendered more explicitly as in (3):

• (3) [∃e: enter(Mary,e)] John sees e

Crucial to my account was the proposal that the clause 'Mary enter' is not a constituent of (3) at all. Nor is the subject-predicate combination 'Mary enter' an argument of anything. Hence, far from being mysterious from the point of view of the most important general accounts of non-extensional embedded contexts, NI complements are orthogonal to the questions these accounts raise, since they are not embedded at all.

I argued also that the properties that Barwise had discerned as peculiar to NI complements followed from the individual events account, and I gave several extensions that were said to favor it. These arguments will figure in the discussion in later sections of this paper. Besides these several points, however, there are also questions of philosophy and of general semantic orientation. Among recent critical discussions, ter Meulen and Bouma (1986), Neale (1988), and Barwise (1989) seem to me to have missed some of the philosophical points at issue, and in the case of the latter two to have misconstrued the conception of semantics with which I was working.

Neale (1988) makes rather heavy weather of the question of the relations of the English syntax to interpretation. In the interest of clarification, let me state the propositions (i)-(iii) to which that account is committed.

(i) The syntactic structure at the linguistic level LF of (1) is (4):

• (4) [_s[Mary enter]_i [_s John sees t_i]]

(ii) The matrix '[_s John sees t_i]' of (4) is an open sentence in one free variable t_i and its truth conditions are as in (5):

• (5) If a is an assignment of values to free variables (traces), then a satisfies '[_sJohn sees t_i]' iff [∃e] [sees(John,a(t_i),e]

(iii) The prefix '[Mary enter]_i' of (4) is interpreted as an existential quantifier binding t_i, whose restriction is to assignments to t_i satisfying 'Mary enter'; these are assignments a such that

enter(Mary,a(t_i))

From (i)-(iii) taken together it follows that (1) is true iff (6):

• (6) [∃e': enter(Mary,e')] [∃e] sees(John,e',e)

These points may clear up any obscurity that was responsible for Neale's assertion that the semantic import of my suggestion was vague, and still more his statement that "if event quantification is general [as in (6) above] then the syntactic, logical, and interpretive character of NI clauses is quite superfluous" (1988: 317). Event quantification is general, in Davidson's sense: each simple sentence contains an implicit existential quantification over events. But what was said to be peculiar to NI complements was not that they contained quantification over events, but that they themselves were existential quantifications. The objection confuses the two.

Barwise's brief comments (1989: 3-4) construe my view correctly as intending semantics proper, and not just conditions on syntactic structures. However, Barwise contrasts the "Davidsonian program of truth-conditional semantics, with a heavy reliance on logical form" with his work "within the model-theoretic tradition, which is quite different." Distinctive of the latter tradition, according to Barwise, is the view that "valid entailments are valid not in virtue of form, but in virtue of content." In Barwise's opinion, "Higginbotham and I start from such different starting points that it really is hard to make contact in a useful way."

Barwise's statement is complex, and subject to interpretation. But there is one remark at least that seems to be a misconstrual. This is the remark that investigations into logical form are committed to the thesis that what Barwise calls "valid entailments" are valid in virtue of "form" rather than "content." There is a tradition of calling implications "formally valid" when their correctness follows from logic alone; but that should not be taken to imply that it is anything other than the content of the logical words that makes them valid. Anyway, the individual events account of the semantic properties of sentences with NI complements does not rely upon a form-content distinction.

I close this section with a brief discussion of an early published treatment of NI complements that sticks within the possible worlds framework, the article Niiniluoto (1982) (this article was not known to me at the time I wrote). Citing Barwise, Niiniluoto acknowledges that the NI complements are not covered by an account in terms of sentence-embedding with a possible-worlds semantics, and he proposes one that invokes what he calls individual events and also retains the idea that all clausal complements to perception verbs are clausal arguments of those verbs. In the case of (1), Niiniluoto would give (7):

• (7) (∃e) e consists of Mary's entering & John sees [(∃e') e' = e]¹

The idea is that, although there is always a clausal complement to verbs of perception, the complement is not the NI complement itself, but the supplied constituent 'something is identical to the event e'.

An apparent difficulty with Niiniluoto's view is that, just because it interprets the NI complements on a par with complements of other types, and takes both, following Hintikka, on the model of the propositional attitudes, it may not allow the genuinely non-epistemic use of the NI complements. Perhaps, whenever an agent sees an event e she sees also that there is an event identical to e. At least this might be so if events have to register in thought to be truly seen (by the person, not her eyes), and if they actually have to be singled out in thought. If so, what applies to events would apply to other objects. It is at least plausible that if I see a box b, then I see that something is identical to b. To the extent that such implications hold, Niiniluoto would have located in (7) a necessary condition for the truth of a perceptual report with an NI complement. But it seems very doubtful that (7) is also a sufficient condition for the truth of such a report, at least as intuitively understood. First, it seems to fail to sufficiency for the case of seeing ordinary objects. For example, if I look at the reverse of a coin x, it seems that I see that something is identical to the obverse y of x; but I do not see y. For the case of perceptual reports with NI complements, similar examples seem to me compelling: if I see by looking around the empty room that John left, then I see that there is something identical with his leaving; but I do not see him leave.

But we cannot in view of these examples conclude that Niimluoto's interpretation is not faithful to the semantics of NI complements. For there are at least two degrees of freedom in the theory. In the case of the coin, it might be responded that I do not have a visual impression of y, and so cannot be said to see that there is something identical to it. The requirement that I have a visual impression might be built into the definition of truth. In the case of seeing that John left, there might also be a scope distinction, between seeing that there is something identical to John's leaving, and his leaving being a thing x such that one sees that there is something identical to x. However, it does appear that, if the individual events account of NI complements can be sustained, then Niiniluoto's further apparatus is redundant.

3. The Role of E

The individual events account of perceptual reports makes essential use of Davidson's view that ordinary predicates, or some of them anyway, contain an extra argument-place for events. This hypothesis of an E-position has played a role in issues of the semantics of modification and adverbial quantification, and other places. The hypothesis fits into a rather general framework for looking at the problems of linguistic semantics. The most significant feature of that framework is the thesis that the only ways of combining constituents, apart from that of predicate and individual argument, are given by the quantifiers and truth functions. Arguments of higher type then go by the board, and combinations that have been taken on the model of functions of higher type call for other devices to play the role of semantic glue. The major such device is the extra argument place provided by the E-position. In such a manner does ontology— the individual events over which the E-p...

Table of contents

• Cover
• Half Title
• Title
• Copyright
• Contents
• List of Contributors
• Acknowledgments
• Introduction: From Ordinary Use and Regimentation to Systematic Theory
• Part 1 Philosophy, Syntax, and Semantics
• Part 2: Philosophy, Semantics, and Pragmatics
• Part 3: Linguistics and Philosophy of Science
• Part 4: A Case Study in Philosophy and Linguistics: Mixed Quotation