Cross-language phonetic similarity of vowels

Current theories of second language (L2) speech perception/production have been developed to explain the relative difficulties L2 learners have in learning specific non-native phonetic segments/contrasts. In each theory, the concept of cross-language (L1/L2) phonetic similarity is invoked to account for the fact that some non-native contrasts are more difficult to perceptually differentiate (and to produce authentically) than others, both initially and after years of experience with the L2. In this paper, two kinds of empirical techniques used to establish L1/L2 similarity of vowel systems are described and critiqued: 1) quantitative analyses of L1/L2 acoustic similarity and 2) perceptual assimilation paradigms that directly ask listeners to judge the similarity of L1 and L2 vowels. Methodological issues addressed include: a) the development of appropriate production/listening protocols, b) assessment of the relative contribution of spectral (quality) and temporal (quantity) parameters in specifying crosslanguage phonetic similarity of vowels, c) the design of perceptual assimilation tasks, and d) the statistical treatment of similarity judgments. Finally, the conditions under which results from acoustic comparisons and perceptual similarity tasks either agree or conflict will be discussed in terms of their implications about the nature of the underlying representations of L1 and L2 phonetic categories.