It is well known that the study of the sentence and its types and especially the study of the relations between different parts of the sentence has had a long history. Rhetoric was mainly engaged in the observation of the juxtaposition of the members of the sentence and in finding ways and means of building larger and more elaborate spans of utterance, as for example, the period or periodical sentence.
Modern grammars have greatly extended the scope of structural analysis and have taken under observation the peculiarities of the relations between the members of the sentence, which somehow has overshadowed problems connected with structural and semantic patterns of larger syntactical units. It would not be an exaggeration to state that the study of units of speech larger than the sentence is still being neglected by many linguists. Some of them even consider such units to be extralinguistic, thus excluding them entirely from the domain of Linguistics.
Stylistics takes as the object of its analysis the expressive means and stylistic devices of the language, which are based on some significant structural point in an utterance, whether it consists of one sentence or a string of sentences. In grammar certain types of utterances have already been patterned, thus for example, we have all kinds of simple, compound or complex sentences, even a paragraph long, that may be regarded as neutral or non-stylistic patterns. At the same time, the peculiarities of the structural design of utterances, which bear some particular emotional colouring, that is, which are stylistic and therefore non-neutral, may also be patterned and presented as a special system, which we shall call "stylistic patterns". Stylistic patterns should not be regarded as violations of the literary norms of standard English. On the contrary, these patterns help us to establish the norm of syntactical usage, inasmuch as their study reveals the invariant of the form together with the variants and what is more, reveals the borders beyond which the variants must not be extended.
Stylistic syntactical patterns may be viewed as variants of the general syntactical models of the language and are the more obvious and conspicuous if presented, not as isolated elements or accidental usages, but as groups easily observable and lending themselves to generalization.
Prof. G. Vinokur maintains that in syntax it is no new material that is coined, but new relations, because the syntactical aspect of speech is nothing more than a definite combination of grammatical forms, and in this sense the actual words used are essentially immaterial. Therefore syntactical relations, particularly in poetic language, are that aspect of speech in which everything presents itself as actualization of the potential and not merely the repetition of the ready-made. By "the potential" G. Vinokur apparently means variations of syntactical patterns. It follows therefore, that in order to establish the permissible fluctuations of the syntactical norm, it is necessary to ascertain what is meant by the syntactical norm itself. We have already pointed out what the word norm means as a generic term. In English syntax the concept of norm is rather loose. In fact any change in the relative positions of the members of the sentence may beregarded as a variant of the received standard, provided that the relation between them will not hinder the understanding of the utterance. But here we are faced with the indisputable interdependence between form and content; in other words, between the syntactical design of the utterance and its concrete lexical materialization. Syntactical relations can be studied in isolation from semantic content. In this case they are viewed as constituents of the whole and assume their independent grammatical meaning. This is most apparent in forms embodying nonsense lexical units, as in Lewis Carroll's famous lines, so often quoted by linguists.
"Twas brilling, and the slithy toves
Did gyre and gimbol in the wabe:
All mimsy were the borogroves,
And the mome raths outrabe."
The structural elements of these lines stand out conspicuously and make sense even though they are materialized by nonsense elements. Moreover they impose on the morphemes they are attached to a definite grammatical meaning, making it possible to class the units. So it is due to these elements that we can state what the nonsense words are supposed to mean. Thus we know that the sequence of the forms forcibly suggests that after twas we should have an adjective; the y in slithy makes the word an adjective; gyre after the emphatic did can only be a verb. We know that this is a poem because it has rhythm (iambic tetrameter) and rliyme (abab in 'toves - borogroves;' 'wabe - outgrabe'). A closer examination of the structural elements will show that they outnumber the semantic units: nineteen structural elements and eleven, which are meant to be semantic. The following inferences may be drawn from this fact: 1) it is the structural element of the utterance that predetemines the possible semantic aspect; 2) the structural elements have their own independent meaning which may be called structural or, more widely, grammatical; 3) the structural meaning may affect the lexical, giving contextual meaning to some of the lexical units.
 (I.R.Galperin. Stylstics. Moscow., 1971, pp. 190-193).