The nature of phonological representations in a second language

This study explores the nature of phonological representations in a second language (L2). In particular, it explores whether abstract and exemplar phonological representations are available to adult learners of an L2. To do this, the study looks at the ability learners have to perceive and store fine phonetic detail in an L2, and their ability to generalise perceptual learning. Particular attention is paid to the effect of speaker variation in the perception of phonological categories that are difficult to learn in an L2. The population studied is Spanish native speakers who are learning English as a foreign language. To investigate L2 learners’ ability to store acoustically rich representations in their L2, three experiments were conducted. The results from these experiments indicate that L2 learners can store phonetic detail which facilitates lexical access of words previously experienced in the same voice. However, this ability does not guarantee good discrimination of L2 phonological contrasts. On the contrary, acoustic variation due to different speakers can make the perception of non-native contrasts more difficult. Despite this difficulty, this study found that L2 learners can generalise to other voices and lexical items, which suggests that L2 learners can abstract the knowledge gained from exposure to a specific set of exemplars. This study contributes to the ongoing discussion about the nature of phonological representations. The difficulty of learning new phonological categories in an L2 has been previously presented as evidence of the abstract nature of phonological representations. Nevertheless, the results of this study show that L2 learners can preserve some of the acoustic characteristics of words they have experienced, which indicates that, for L2 learners, phonological representations cannot be purely abstract. Considering that abstract representations are also available to L2 learners, the results of the thesis support the development of hybrid models of L2 phonology learning