We have brought up several distinct types of English clausal constructions, and have been lead to the descriptive generalization in (14),repeated here as (33): (33) Reduced clauses cannot occur in non-complement positions. The generalization in (33) refers to two theory-internal notions, reduced clauses and non-complement positions. Both notions are concerned with the composition of syntactic structures to be defined by X-bar theory. Without these theoretical notions, it would be difficult to describe in a general form the fact that certain types of complement clauses-namely, null-that clauses, if-clauses, Acc-ing gerund, ECM complement clauses, and Raising complement clauses-cannot occur in particular syntactic positions. Instead, one would have to describe this fact for each clause type, in such a way that null-that clauses cannot occur in such and such positions, and if-clauses cannot occur in such and such positions, and Acc-ing gerund cannot occur in such and such positions, and so on, although the positions in which they cannot occur are totally the same. Given the terminology of X-bar theory, however, it has turned out that these types of complement clauses are all reduced clauses, and the positions where they cannot occur are all non-complement positions. Then, the generalization has obtained that reduced clauses cannot occur in non-complement positions. It is a theoretical issue, and differs depending upon theories, how to explain why such a descriptive generalization holds at all. Hopefully, the demonstration here provides a piece of evidence showing that a theory or a particular theoretical nation plays an important role in the description of linguistic facts. Moreover, I have made a crucial prediction on the basis of the well-accepted theoretical assumption the ECM complement clauses and Raising complement clauses are reduced clauses; namely, the prediction that these types of clauses cannot occur in non-complement position. The prediction based upon the theoretical assumption is actually borne out, as illustrated earlier. The illustration of the prediction, I hope, shows that a theory or a particular theoretical assumption, coupled with another theoretical assumption, allows us to make some interesting predictions. Predictions serve to widen a range of linguistic facts to be described. A theory plays a crucial part in finding out interesting facts as well as in describing them in some general forms. Finally, let me state a few words as to the recent generative theory in connection with linguistic description. The recent generative theory is getting more and more abstract. I think it is moving toward a good direction as cognitive science. It will contribute, among others, to the inquiry into what is knowledge that is very specific to language faculty, and into how it interacts with other cognitive faculties. However, I am suspicious about how much the abstract generative theory will contribute to the description of linguistic facts in a particular language. While generative theory is claimed to aim both for descriptive adequacy and for explanatory adequacy, the recent generative theory is likely to put much more weight on explanatory adequacy. In my view, a less abstract theory is enough, or even more useful, for the purpose of linguistic description. Of course, how abstract theory one should adopt as a framework differs depending upon what aspect of language one attempts to describe. What I would like to emphasize here is that linguistic theory does not conflicts with linguistic description, and a linguistic theory with an appropriate degree of abstractness serves as a tool for finding out new interesting facts, as well as for describing them in some general, elegant forms.