Résumé
In this study we will examine how the phonological properties of a spoken language interact with its specific phonetic details in determining the perception of non-native sounds. We tackle this issue by investigating: (i) how the English consonants /b, v, w, ð/ are perceptually similar to the consonants of the native phonological system, by two groups of listeners, Italians and Danes, whose native phonetic-phonological systems differ significantly, both from each other and in relation to English, (ii) how these assimilations correlate to the ability of the two groups of listeners to perceive the corresponding contrast of English /b/-/v/, /w/-/v/, and /ð/-/v/. In particular, we will evaluate the adequacy of some theoretical models of empirical perception, based on the results of a test of categorization and a test of discrimination. We will look at three models of perception in adults: a model of the perception of non-native sounds in functional monolingual adults (Perceptual Assimilation Model - PAM) and two models of perception in adult learners of a second language (Speech Learning Model - SLM and PAM-L2). In addition, for an additional ‘control’ of the predictions generated by the models under consideration, we will compare the results with the predictions of a recent model of the perception of native sounds in children (Articulatory Organ Hypothesis - AOH), which we applied here, with adaptations, to the perception of the adult. Our results confirm core predictions of PAM, SLM and PAM-L2. The confirmation is only partial: none of these models manages to give an account of how one of the English contrasts is perceived by the Italian subjects and by the Danes. We will discuss the implications of these findings for existing hypotheses on non-native perception and indicate the direction in which we believe it is appropriate to extend the models considered.