Second language perception of accented speech

Abstract

The present study addresses a core issue in the study of speech perception, the question of how stable phonological representations are accessed from an inherently variable speech signal. In particular, the research investigates the perception of accented English speech by native and non-native listeners. It is known from previous research that foreign-accented speech is harder for native listeners to process than native-accented speech. The reason for this lies in not only qualities of the input (deviation from native production norms, for example) but also in qualities of the listener. Specifically, listeners' speech perception systems are tuned from an early age to pay attention to useful distinctions in the language environment but to attenuate differences which are not useful. This quality of the listeners' speech processing system suggests that in addition to being native speakers of a language or languages, we are also native listeners. However, what is a liability for native listeners (non-native input) may be a benefit for non-native listeners. When the foreign accent is derived from a single language shared between the speaker and the listener, application of native-language processing strategies to the accented input may result in more efficient processing of the input. The experiments in this dissertation address this possibility. In an experiment involving Dutch listeners processing Dutch-accented and American English-accented sentence materials, a reaction time advantage was observed for the mutually-accented materials. Experiments testing the main hypothesis with native Spanish-listening participants showed a different pattern of results. These participants, who had more experience with English overall that the Dutch participants, performed similarly to native-listening controls in displaying faster verification times for native accented materials than mutually-accented materials. These experiments lead to the conclusion that native-like listening, as assessed by the sentence verification paradigm employed in these experiments, can be achieved by non-native listeners. In particular, non-native listeners with little experience processing spoken English benefit from hearing input produced in a matching accent. Non-native listeners with sufficiently more experience processing spoken English, however, perform similar to native listeners, displaying an advantage for native accented input.