Interpretive Semantics

When Noam Chomsky presented his revolutionary grammar in his 1957 Syntactic Structures, he did not pay particular attention to meaning or semantics. It wasn’t until 1963, that someone proposed a way to work with semantics as part of generative grammar. Jerrold J. Katz and Jerry Fodor, both philosophers at MIT, published an article in Language entitled The Structure of Semantic Theory. The semantic model outlined in this article was further developed in a book entitled An Integrated Theory of Linguistic Descriptions (1964) by Katz and Paul Postal, a linguist. The article in Language and the book by Katz and Postal were responsible for many of the revisions of the generative-transformational model which were proposed by Chomsky in his Aspects of the Theory of Syntax (1965).

Interpretive semantics, now also called the standard theory, relies on projection rules to give an interpretation to a sentence. According to this model, each lexical item has associated with a certain number of features. One of these features is the part of speech which is partially determined by the strict subcategorization constraints on a particular item. For example, any particular verb characteristically requires certain grammatical categories to precede it and follow it. Thus, an intransitive verb such as walk does not require anything to go with it except a subject, while a transitive verb such as hit requires both a subject and a direct object. A verb like dart must be followed by an adverb of motion. These are statements about the strict subcategorization of these three verbs. In sentences like John hit, or John saw of the strict subcategorization requirements of the verb are violated, in the first sentence by not having anything after the transitive verb, and in the second sentence by having a preposition after the transitive verb rather than the appropriate noun phrase.

Subcategorization rules are basically concerned with syntactic matters rather than semantic matters. But there are also certain semantic features associated with each lexical item. The semantic markers are those semantic features which have general importance in the grammar. A particular semantic marker will be a feature of not just a single word, but of many words. Semantic distinguishers have relevance for only a particular word. A distinguisher is the semantic feature which separates one lexical item from all others. For example, in the most common meaning of the word bachelor, the feature noun is the part of speech: the features human, male, and adult, are the semantic markers, and never-married is the distinguisher.

It is the function of the projection rules to scan the features, i. e., the part of speech, the semantic markers, and the distinguishers of each lexical item, and to combine these features with those of other lexical items with which the individual words form a constituent, and to determine which of the features are compatible with each other at a particular level. Thus, the projection rules would show that the word ball is ambiguous, meaning either a spherical object for use in a game, or a formal dance. But the projection rules would show that the expression soccer ball is unambiguous since the features of soccer are normally compatible only with the spherical-object meaning of ball, rather than the formal dance meaning. Projection rules can therefore be seen as rules of disambiguation. After the projection rules have scanned the entire sentence, including embedded sentences, if any, in much the same way as they scanned the features of the expression soccer ball, they would be able to specify if a sentence is deviant or has no composite meaning, i. e., is anomalous, or if it has one composite meaning, i. e., is unique, or if it has more than one composite meaning, i. e., is ambiguous. By making a distinction between the syntactic features and the semantic features, it is also possible to tell if an anomalous or ambiguous sentence is deviant because of syntax or semantics. For example, the sentence, He was seated by the President, is semantically ambiguous because the confusion is due to the double meaning of by. The sentence, I won’t speak at any more dull women’s meetings, is syntactically ambiguous because the confusion is due to the placement of the two modifiers dull and women in front of meetings. The syntax does not show which word dull is modifying.

The important differences between Chomsky 1957 and Chomsky 1965 result from the attempt to incorporate the semantic theory of Katz, Fodor, and Postal into the transformational model. One important difference is that the 1965 model makes frequent reference to the cover term, complex symbol or CS, which refers to all of the features of a particular lexical item. Chomsky did not actually write semantic rules into his revised grammar, but by using CS as his last notation on tree diagrams, he made it possible for semantic information to be plugged into his syntactic analysis. Another difference is that in the later model the tree diagram shows sentences to be embedded into other sentences in the following manner:

Before the 1965 Aspects, generative transformationalists made a distinction between kernel sentences, which are the most basic simple sentences, and derived sentences, which have been changed into such things as questions, imperatives, passives, and negatives. They also generated each sentence separately and then combined the sentences together by means of a generalized transformation such as the following:

This is the cat

The cat ate the rat } This is the cat that ate the rat

There were problems in applying projection rules in this earlier model because each sentence had to be scanned separately. This meant that the projection rules could not be used to determine the meaning of the super sentence that resulted when the two sentences were merged together in some way.

Further reading on these matters might be found in Roderick A. Jacobs and Peter S. Rosenbaum’s English Transformational Grammar (1968), Jerrold J. Katz’s Semantic Theory (1972), and Ray S. Jackendoff’s Semantic Interpretation in Generative Grammar (1972).