B. Problems concerning the composition of spans of utterance larger than the sentence

In recent years a new theory concerning the inner relations between context and form within the sentence has appeared. This theory, elabo­rated by S. Harris, N. Chomsky, M. Postal and others, is called Genera­tive Grammar. It maintains that grammar must not only describe the laws which regulate the functioning of linguistic units but must also be capable of generating new sentences.

"A grammar of this kind," writes John Lyons, "is 'predictive' in that it establishes as grammatical, not only 'actual' sentences, but also 'potential' sentences." l

The reference to Lyons's statement has direct bearing on the problems of stylistic syntax. The fact is, as will be seen later, that any one of the syntactical SDs is capable of generating an unlimited number of senten­ces within the given pattern. However, according to orthodox generative grammar, some of them are regarded as 'ill-formed' and even 'ungram-matical' inasmuch as they fail to meet the requirements of the basic (kernel) structures.

The theory further maintains that there are two kinds of structures — a deep structure and a surface structure. The latter are the actual senten­ces produced by the former, which is not presented in language units and therefore unobservable.

Mention of this theory is made here, firstly, because in modern sty-listics attempts are being made to build up a grammar which would generate deviant constructions and thus broaden the limits of the 'well-formed' sentences which are regarded as the only ones that are 'grammat­ical'. Another reason is that transformation, one of the basic methods employed in generative grammars, is very effectively used in stylistics when it is necessary to find the stylistic meaning of this or that sentence structure. A third reason is that generative grammars aim at reconstruct­ing the processes connected with the formation of sentences. This has direct bearing on the interpretation of syntactical SDs and particularly on their linguistic nature.

This theory enables the interpreter to look at a sentence from the point of view of what is 'behind' the sentence.

As J. P. Thorne states, "Generative grammar is important to stylist­ics because in addition to these'surface structure' facts, it is concerned

with the so-called 'deep structure' aspects of language, that is, those facts about linguistic structure which cannot be directly related to what can be observed. Most stylistic judgements relate to deep structure."1

It follows then that the so-called generative grammar is not so strik­ingly new. This is also noted by the well-known linguists John Lyons and D. Bolinger, 2 who state positively that there is nothing new in the theory of generative grammar.

Another development in linguistics also having direct bearing on the problems which concern us when dealing with syntactical SDs, is 'text-linguistics', as it is called. This development, which as yet has not been formed as a separate theory, aims at investigating the objective criteria concerning ways and means of constructing texts of different kinds and genres. 3

For this purpose it is first of all necessary to find the elements into which any text may fall. In other words, there must be certain constituent units of which any text is composed.

Phonemes, the smallest language units, function within morphemes and are dependent on them, morphemes function within words, words — within sentences, and sentences function in larger structural frames which we shall call "supra-phrasal units". Consequently, neither words nor separate sentences can be regarded as the basic constituents of a text. They are the basic units of lower levels of language-as-a-system, as is shown above.