Native and Non-native Speech Perception

INTRODUCTION A central goal of research on speech perception has been to uncover the auditory dimensions that listeners use to derive linguistically meaningful utterances from the speech signal. In order for listeners to successfully decode an utterance, they must be able to reliably distinguish the speech sounds of a particular language. The speech signal presents listeners with a multitude of acoustic information, along different auditory dimensions (e.g., formants or voice onset time), that lies within the limits of human hearing. Perceiving speech is not simply a task of attending to these auditory dimensions equally; of critical importance is how listeners integrate the multiple dimensions to successfully map them onto particular speech sound categories. This crucial task is exemplified by the difficulties non-native listeners may face in discriminating non-native speech sound contrasts that make use of auditory dimensions in different ways from those in their native languages. Consider the well-known example of the discrimination of the English /l-r/ contrast (in words such as ‘lead’ and ‘read’) by Japanese learners of English. While both native English listeners and Japanese learners attend to the same auditory dimensions, e.g., the second and third formants (F2 and F3) of /l/ and /r/, they do so in different ways. Iverson et al. [1] show that Japanese listeners are most sensitive to variation in F2, but not to variation in F3 which is a more reliable cue for successfully separating the two sounds. English listeners, on the other hand, are most sensitive to variation in F3 and consequently exhibit far greater discrimination accuracy. The relative preference for certain cues over others, referred to as cue weighting, differs between native and non-native listeners. The sensitivity to these cues develops early in life and is related to the acoustic dimensions found in infants’ ambient language. Unsurprisingly, Japanese does not have an alveolar sound contrast that is differentiated by F3 as in English. Though notoriously difficult, Japanese learners of English can begin to change their relative use of auditory dimensions, or attend to new dimensions, to improve their discrimination accuracy [2]. This review first provides a brief examination of how speech perception develops early in life and offers a theoretical model to account for this. The implications of this early experience are then reviewed in three related areas of non-native speech perception, namely cross-dialect, cross-language and secondlanguage speech perception.