The development of an empirically based model of communicative competence is discussed in terms of conceptual, statistical, and methodological considerations. A distinction is made between descriptive and working models of communicative competence. Working models attempt to show how components of communicative competence are interrelated psychologically to form a set of independent factors and to determine whether aspects of language are psychologically distinct and can be learned separately. In reviewing the research on this question, several methodological and statistical issues must be addressed. Among these issues are interpretation of the Pearson product moment correlation coefficient, subjects' language background, patterns of language proficiency, and criterion versus norm referenced language measures. Examples are provided of the ways in which these issues can influence the interpretation of patterns of test results. Research on the components of communicative competence are reviewed and the findings reinterpreted in light of the methodological issues. (Author/RW)