Deep-structure acceptability

Facts about lexical items that relate to the formal properties of sentences can be separated into two sets: requirements on the deep-structure and requirements on the surface structure. The former determine the acceptability of a given word in deep-structures of certain types; the latter specify those grammatical modifications of sentences whose operation is determined by lexical information. The surface conditions are provided in the grammar in the form of the rules which convert deep structures into surface structures (transformational rules), and possibly, in some cases, by the elaboration of special constraints on surface structure.

I shall take the position that content words may all be inserted as predicates, and that their realization as nouns, verbs or adjectives is a matter of the application of rules. Therefore we need not consider part-of-speech classification among the types of information relevant to the lexical insertion into deep structures. What Chomsky has referred to as ‘ strict subcategorization’¹ corresponds to what I have treated here in terms of the number of arguments a predicate takes and their case structure. What Chomsky has referred to as ‘ selection'² is described here with the concept presupposition and is taken as being more relevant to semantic interpretation than to lexical insertion.

The deep- structure requirements that are of chief interest for this discussion, then, are those of the type Perlmutter has been referring to as ‘deep structure constraints’ or ‘input conditions’.³

Examples, due to Perlmutter, are the requirement for budge that it occur in a negative sentence, as shown in grammaticality judgments on (79) and (80):

{79) *I budged

(80) I didn’t budge

the requirement for lurk (discussed earlier) that its Agent be non-coreferential to the Agent of the ‘ next higher ’ locutionary verb; or for try that its Agent be coreferential to the (eventually deleted) subject of the ‘next lower’ sentence, as suggested by (81) and (82):

(81) *I tried [for you] to find it.

(82) I tried [for me] to find it.

I have included deep-structure constraints in this survey of types of lexical information, but I have nothing new to say about them. I would like to suggest, however, that it may not be necessary to require the extent of detail which Perlmutter envisions or the transformational apparatus which that sometimes entails. Where Perlmutter requires that the Agent of try match the Agent of the embedded sentence, it may only be necessary to require that the coreferential noun-phrase in the embedded sentence be the one which is destined to be the subject of that sentence. And where Perlmutter requires sentence (83) to be derived transformationally from the structure underlying (84):

(83) He tried to be misunderstood

(84) He tried to get to be misunderstood

this may not be necessary if try is merely described as a verb which expresses, of its Agent subject, the intension and attempt to bring about the situation identified by the embedded sentence. This may be necessary in order to account for the way in which we understand sentence (85):

(85) He tried to seem cheerful

a sentence which cannot be straightforwardly paraphrased in such a way as to reveal an underlying agentive notion in the embedded sentence.

¹ Noam Chomsky, Aspects of the Theory of Syntax (1965), M.I.T. Press, esp. pp. 95-100.

² Ibid. pp. 148 ff.

³ My knowledge of Perlmutter’s work on deep-structure and surface-structure constraints comes from Perlmutter’s presentations at the January 1968 San Diego Syntax conference and from references in papers and presentations by J. R. Ross.