Linguistic data as complex items

The article discusses the problematic relationship between linguistic data and theories. First it presents data as characterized by different forms of internal and external complexity. Then it shows special kinds of data, such as empty categories, as an example of ‘theory-laden’ data, and clitics as examples of interface or cross-level items. Finally it discusses data according to the twentieth century distinction between competence and performance. Throughout the article interest for the richness and diversity of linguistic data is considered methodologically relevant for linguistics both in comparative and in one-language studies of data. It is the author’s belief that, to take into account the complexity of linguistic data, theoretical linguistics should rely less on overly-rigid architectures in order to allow diversity as much as regular behavior of data to emerge. Finally it is suggested that one way to make theories more flexible is offered by optimality theory. Only by adopting this line of research can theoretical approaches allow both universal and language-specific aspects of language to coexist and be accounted for.