Models of linguistic variability

Models of linguistic variability
Patterns used in structural linguistics to describe a language and its various aspects (phonology, grammar, lexicon) in order to define more accurately linguistic concepts and their relationships. This helps to clarify the structures underlying the infinite variety of linguistic phenomena; sometimes the structures themselves are called models. Depending on their area of application, linguistic models are divided into phonological, morphological, syntactic, and semantic. In constructing models, the means and methods of mathematical linguistics are used. Any linguistic model establishes such things as the objects corresponding to the data of direct observation, including a large number of sounds, words, and sentences; objects constructed by the linguist (constructs) for descriptive purposes, consisting of sets of categories, markers, and elementary semantic structures whose size and scope have been rigorously limited from the outset.
If the initial material (“input”) of the investigation consists of sounds, words, and sentences, and the result (“output”) consists of categories and semantic structures, the model is called analytical. Such an analytical model is a model of the category of gender, which provides an unambiguous resolution of disputed questions. A word’s grammatical gender may be determined by the form of a word; for example, Russian words ending in -a are usually feminine, but the marker -a is not unambiguous, as shown by the word papa, “daddy,” “pope,” or by meaning (words designating feminine beings belong to the feminine gender, but this marker is likewise not unambiguous, as shown by German das Weib, “woman,” which belongs to the neuter gender).
In the model of gender, each word is taken to have its system of forms (for example, declension of stol “table”: stol, stola,stolu), and it is known which other word forms agree with the given word form (as in etot stol, “this table”: etot stol, etogostola). Two words, x (stol) and y (kakadu, “cockatoo”), belong to one gender if, for every form x1 of word x and every wordform z agreeing with x1, there will be found a form y1 of word y agreeing with z (etot kakadu, etogo kakadu), while the reverse will be true for every form y11 of word y. This model makes it possible not only to resolve disputed questions unambiguously but also to confront the category of gender with the category of part of speech (whereby gender is “inserted” into the part of speech); to establish which categories of other parts of speech are structured isomorphic ally (analogously) with the gender of the noun (for example, the category of verb agreement); and to compare the category of gender in Russian and other Indo-European languages with the category of grammatical class in, for example, the Bantu languages. Thus, analytical models are employed in language typology.
If the input material consists of categories and elementary semantic structures and the output consists of certain formal constructions, the model is called synthetic or generative (these models are also called generative grammars). A generative model contains a certain hypothesis about a language’s internal structure that is inaccessible to direct observation. The generative model is then tested by comparing a large number of objects deduced from the model with real linguistic facts .This permits the model to be classified and evaluated according to the degree of its correspondence to the facts of the language and according to the degree to which it reveals intuitively felt laws of the language (explanatory power). Since each model describes not a whole language but some one area or even a separate category of it, precise description of a language presupposes the simultaneous use of several models, relating both to the one area (for example, several complementary models of the categories of part of speech, case, and gender) and to different areas.