A hypothesis

So far, we have presented a set of syntactic-semantic correlations without considering how they might be accounted for. We shall continue by analyzing these facts and others to be pointed out in the course of the discussion, in terms of a tentative explanatory hypothesis, by which the semantic difference between the factive and non-factive complement paradigms can be related to their syntactic differences, and most of the syntactic characteristics of each paradigm can be explained. The hypothesis which we should like to introduce is that presupposition of complements is reflected in their syntactic deep structure. Specifically, we shall explore the possibility that factive and non-factive complements at a deeper level of representation differ as follows: 

If this interpretation is correct, then closest to the factive deep structure are sentences of the type

I regret the fact that John is ill.

The forms in the factive paradigm are derived by two optional transformations: formation of gerunds from that-clauses in position after nouns, and deletion of the head noun fact. (We do not pause to consider the general rules which take care of the detail involving that and of.) By gerund-formation alone we get

I regret the fact of John’s being ill.

Fact-deletion can apply to this derived structure, giving

I regret John’s being ill.

If fact-deletion applies directly to the basic form, then the simple that-clause is formed:

I regret that John is ill.

Although this last factive sentence has the same superficial form as the non-factive

I believe that John is ill

according to our analysis it differs radically from it in syntactic form, and the two sentences have different deep structures as diagrammed above. Simple that-clauses are ambiguous, and constitute the point of overlap (neutralization) of the factive and non-factive paradigms.

If factive clauses have the deep structures proposed by us, these various surface forms in which factive clauses can appear become very easy to derive. That is one piece of support for our hypothesis. The remaining evidence can be grouped under three general headings:

(1) syntactic insulation of factive clauses (section 4)

(2) indifferent and ambiguous predicates (section 5)

(3) pronominalization (section 6).