The term ‘typology’ has a number of different uses in linguistics. It is used to refer to the classification of structural types across languages (typological classification), to the study of linguistic patterns or generalisations that hold across languages (generalisation), and, finally, the term represents a theoretical and methodological approach that contrasts with other linguistic theories and which provides an explanation of grammatical phenomena on a broad empirical base (Croft 1990:1-2). All the three more specific definitions of ‘typology’ constitute the typological approach to the study of grammar, which the present paper endeavours to pursue. In the following chapters we will investigate the notion 'language typology' in more detail, dealing first with typological classification and then with generalisation.

3.1 Typological classification

In typological classification, languages are classified according to the features they have in common and which, in turn, distinguish them from other languages (Croft 1990:1).

The classification of languages originates from the nineteenth century, and was based on the morphological structure of words. The morphological typology of the nineteenth and early twentieth centuries, as Croft (1990:1) remarks, is an example of the "classical" use of the term ‘typlogical classification.’ According to the classical formulation of morphological typology (after August Schleicher) languages were divided into three types: isolating which did not use affixes at all; agglutinative that used affixes denoting single grammatical categories (e.g. number) and were "concatenated with relatively little phonological alteration" (Croft 1990:39), inflectional where affixes were often fused together with several grammatical categories (e.g. number, gender, case) into a single morpheme, and which "often underwent major phonological alternations when combined with roots" (Croft 1990:39).

In the 19^th century, the typological approach was strongly influenced by Darwin’s ideas according to which language types were viewed as representing different stages in linguistic evolution, and changes in a language were seen in terms of growth and decay (McMahon 1994:316). As Anttila (1972:312) notes, "[i]t was generally thought that isolating languages yield agglutinating ones and that these in turn yield flectional types."

At that time the typlogical classification of languages "recognised only a single parameter on which languages varied, the morphological structure of words," as pointed out by Croft (1990:39), and "it was a classification of languages as a whole, not parts of a language" (ibid.). More recent typological classification, in contrast, involves a particular construction rather than a language as a whole. For instance, "the nominal system of a language may be agglutinative while the verbal system is inflectional" (Croft 1990:42). In this respect, typology "owes to the structuralist (and generative) approach to linguistic analysis"(Croft 1990:39).

Nowadays, typological classifications tend to concentrate more on syntactic characteristics of a language than on morphology. For instance, Comrie (1981:49) writes that

morphological typology has a secure, but restricted, place in language typlogy, and it is to be hoped that general linguistic textbooks will not continue indefinitely to give the impression that this is the only, or most insightful, way of classifying languages typologically.

The fact that languages can be classified in some other way was illustrated by Joseph Greenberg (1990) who studied a variety of syntactic, semantic and morphological characteristics of languages and formed 45 universals. This research revealed, according to Hawkins (1983:3) that of all mathematically possible word order combinations across different phrasal categories, languages use suprisingly few (see Hawkins 1983). For example, with regard to ordering within major sentence constituents, Greenberg (1990:43) states that "[t]he vast majority of languages have several variant orders but a single dominant one." He (1990:43) brings out six possible orders of subject (S), verb (V) and object (O): SVO, SOV, VSO, VOS, OSV, and OVS, the last three of which are found to be very rare (Greenberg 1990:43).

Clearly, this kind of word order typology presupposes the viability of categories such as subject, verb, object, noun, and adjective as basic linguistic entities of all languages, as well as the viability of basic word ordering in natural language (Hawkins 1983: Ch.1.5). This, in its turn, leads us to the basic implication that typology has for contemporary linguistics: cross-linguistic comparison.

3.1.1 Cross-linguistic comparison

Any typological analysis requires cross-linguistic comparison of "the relationship between linguistic form and external function" as a descriptive prerequisite (Croft 1990:12). However, a number of methodological problems are related to that.

One of the important issues is that the identification of the fundamental grammatical categories such as noun, verb and adjective, subject and object, head and modifier, etc., appears to be controversial, because these grammatical categories display considerable variation in their structural expression across languages (Croft 1990:13). For example, "English nouns like food, faith, and love have to be rendered by verbal expressions in Mazatec (of Mexico); some Hopi verbs can be given by Kannada verbs, other correspond to Kannada adjectives" (Anttila 1972:316); names, for example, are verbs in such languages as Oneida, e.g. Kanastalukwa means ‘Shelled Corn’, Skanyataliyo ‘Handsome Lake’ (roughly ‘the water is again good for navigation’); layáthos has the meaning ‘he plants corn’ for the English noun ‘farmer’ and shakoye°nás ‘he arrests them’ for the English ‘sheriff’ (Anttila 1972:316). The grammatical relation of ‘subject,’ for instance, can be expressed structurally in several ways: "by case/adposition marking, by indexation or agreement, by word order, or by a combination of both of these" (Croft 1990:13). This, in its turn, requires cross-linguistic means to identify case/adposition, indexation/agreement and word order (Croft 1990:13). Although word order is regarded as the easiest to identify, the correct word-order statement requires the identification of the grammatical category of each unit (Croft 1990:13). For example, in order to assert that subjects in Yoruba can be identified according to their pre-verbal position, one has to identify verbs first, "not to mention the category ‘noun phrase’ or at least ‘noun’ which the subject is assumed to fall into (and not to mention a cross-linguistic means of individuating syntactic units)" (Croft 1990:13).

As for case/adposition and indexation/agreement, they also display remarkable variation across languages, and thus will not provide an unproblematic cross-linguistic definition, at least not by itself (Croft 1990:15).

Relying (only) on intuition would also present a problem, because "[o]ur intuitive notion of "subject" is based on English subjects (or "Standard Average European" subjects, to use Benjamin Whorf’s [1956:138] term), specifically, on the semantic relation between the event denoted by the verb and the participant denoted by the English subject" (Croft 1990:15). As it can be expected, the "English subject" may not correspond to the subjects of some "exotic" language, consider the following examples from Chechen-Inguish (CM is a class marker that agrees with a verbal argument) (Croft 1990:15-16):