Contrastive studies, as a method of linguistic analysis, have a long tradition dating back at least to the end of the nineteenth century, with three important landmarks: the 1920s and the 1930s in American structuralism, the Chomskyan revolution in the 1960s with the emergence of generative grammar, and the ‘post-revolutionary’ emphasis on theoretical contrastive projects which subsequently began to appear all over Europe. The latest landmark, dating back to the 1970s, is constituted by the transfer from an emphasis on grammatical competence to communicative competence as an element of sociocultural competence. Sociocultural competence incorporates grammatical as well as pragmatic issues (see Sajavaara 1981). Generally, contrastive enterprises in linguistics can be divided into two areas: theoretical and applied (Fisiak et al. 1978), or alternatively autonomous and generalised (Di Pietro 1971). Theoretical studies concern the ways in which a universal category is realised in two different languages or dialects. The category in question can belong to any level of linguistic analysis: phonology, morphology, syntax, semantics or pragmatics. For example, in phonology the functioning of a phonological rule can constitute a universal category investigated with equal interest and emphasis in two languages. Applied contrastive studies, often called contrastive analysis, concern the applications of the theoretical discipline for such purposes as language teaching, translation, or research on bilingualism and language acquisition. Nevertheless, as James (1990) points out, contrastive analysis lost some of its pedagogic impact in the 1970s due to the frequent changes of models of analysis in theoretical linguistics. ‘Contrastive analysis’ is often used as a term with a broader scope: it comprises both theoretical and applied research and for a justifiable reason: grammatical competence investigated by theoreticians goes hand in hand with communicative competence dealt with by practitioners, or even sociocultural competence which is of interest to both groups (cf. Sajavaara 1981; Lyons 1990). For these reasons we shall use the expression contrastive analysis (henceforth: CA) in the broad sense, incorporating theoretical research of contrastive phonology, morphology, syntax, semantics and pragmatics, as well as more speaker-oriented research in the domain of individual language use and the use of language in society. The differences between contrasted languages can be of three main types: structural (e.g. the presence or absence of subjectless constructions), categorial (e.g. a sentence may require an adjective in one language but an adverb in the other), or functional (e.g. a concept represented as a subject in one language may have to be represented as object in the other). In order to contrast languages methodically, a comprehensive ‘criterion of measurement’ is required. This criterion will differ from one level of analysis to another but it is essential for every level that the correct choice has been made. Depending on this criterion of measurement (or: ‘platform of reference’, Krzeszowski 1990), objects of study may appear similar or different and therefore they may seem either interesting or not worth juxtaposing. This platform of reference is commonly referred to as tertium comparationis (Krzeszowski 1990; Fisiak 1990) and refers to that which is common in the two compared objects and against which the differences can be specified. Since tertium comparationis is determined by the interest in the particular phenomenon, it may seem that comparing languages ‘globally’ is prevented. There have been various solutions proposed to avoid this unwanted conclusion. For instance, Kalisz (1981) suggested that the equivalence of two languages is reflected by the degree of similarity of the syntactic, semantic and pragmatic properties. This solution is not without fault however. The fact that some concepts are universal makes the above task easier. Wierzbicka (1992) says that the similarity of concepts is directly proportional to the degree to which languages are shaped by human nature. On the other hand, differences in culture will be reflected in the differences in concepts. The study of linguistic universals and ‘near-universals’, performed in the domains of lexicology, syntax, semantics and pragmatics, could constitute the first step in a contrastive enterprise by taking care of what is uniform and narrowing down the scope of CA to the units (words, constructions, speech acts) which are more ‘culture-bound’. The increasing sophistication of linguistic theories leads inevitably to specialisation in the field of CA. Chomsky’s syntactic theory, formal semantic models, or the recent developments in generative phonology may seem to be less adequate for the purposes of CA than the ‘simpler’ theories of the past. However, an explanatorily adequate theory is not overburdened with formalisms: all theoretization should serve the purpose of explaining practice (see Lyons 1990). CA is essentially an eclectic enterprise (using a positive connotation of ‘eclectic’). It benefits from increasing specialisation only to the extent to which the emerging theories follow the principle that no theoretization should be wasted. At present we cannot contrast languages without resorting to fragmentary analyses both in microlinguistics (essentially phonological, grammatical, and lexical CA) and macrolinguistics (aspects of semantics and pragmatics, as well as studies motivated by sociology and psychology). The mutual compatibility of the disciplines is maintained only if the practical implications of the sophisticated models are clearly delineated. Extending the analysis beyond the sentence as a unit added a new perspective to CA. Contrastive, and in particular cross-cultural, pragmatics (Oleksy 1989; Wierzbicka 1991; Blum-Kulka et al. 1989; Kasper & Blum-Kulka 1993; Verschueren & Bertuccelli-Papi 1987), discourse analysis (Oleksy 1984), and also text analysis and rhetoric (Hartmann 1980) highlight the importance of an adequate tertium comparationis. In conversation, speakers may utilise different speech acts to achieve a certain effect. In a cross-linguistic perspective, an utterance may constitute an invitation in one language, whereas its literal translation may render a question. Moreover, the requirements of the situation have to be considered: different speech acts may be required in different languages in analogous situations. Therefore, a specification of pragmatic equivalence of contrasted structures is required. The development of various approaches within pragmatics is also reflected in pragmatic CA. The conversation-analytic approach to ‘turn-taking’ will emphasise a different platform of reference than speech act theory, and the latter will possibly differ in this respect from Gricean conversational logic and its offspring, relevance theory. Text CA, with its chief application in translation, introduces an emphasis on context and text cohesion; lexicon and syntax do not suffice to produce an adequate translation. The study of deixis, anaphora, as well as ellipsis, substitution and conjunction (cf. Halliday & Hasan 1976) is required, as well as contextual and sociocultural background information. Equivalence in translation is achieved through intertextuality (de Beaugrande 1980): the relationship between the text and other texts arrived at through the experience of the text receivers, later classified on different levels of linguistic analysis. Intertextuality is thus further proof of the eclectic nature of CA and, possibly, of the advantages of this eclecticism over a ‘one-layer and one-theory’ perspective. Needless to say, such an integrated perspective finds multiple applications in language teaching and learning, translation, research on language acquisition, and many other areas. Semantic and pragmatic perspectives added new dimensions to CA by analysing language in discourse situations, in social and cultural settings. Both of these disciplines have a long contrastive tradition. The first deals with situations, including the background knowledge of the participants, the issues of reference, the relevance of time and place, etc. Pragmatics focuses on speech acts, recovering assumptions and communicative intentions, implicatures, presuppositions, and many other problems which can be identified and solved only on the macro level. A level of sociopragmatics has been postulated for capturing cross-cultural differences in the use of speech acts. Its subject of investigation is concerned with how pragmatic principles operate in various cultures, e.g. what constitutes a speech act of apology, complaint or request, and how they are distributed in discourse. There is also an important role for psycholinguistics: cognitive psychology may be seen as underlying the choice and construction of linguistic theories (cf. Sajavaara 1984). The more so that conversation is a complex process which involves more than mere encoding and decoding. Hence CA cannot be neatly divided into contrastive phonology, syntax, semantics, etc. Theoretical CA can be conceived of as a method of analysis of linguistic data either for its own sake or as a theoretical aid to practical difficulties involved in language learning and teaching, dictionary construction, research on language acquisition, or, what interests us here most of all, ‘mainstream’ linguistic research. To conclude, whereas the aim of autonomous CA would be to investigate how a universal category is realised in two languages, the aim of CA employed to aid a mainstream project is to provide evidence for a linguistic hypothesis.