Acoustic analysis of EFL speech production corpora according to the Speech Learning Model

Abstract

It is widely accepted that L1 influences the acquisition of L2 phonology during Second Language Acquisition. Models such as the Speech Learning Model (SLM) have tried to explain this process by defining bi-directional interactions between the phonetic systems of both languages. Flege (2007) called them assimilation and dissimilation, by which non-native sounds are classified as either “new”, “identical” or “similar”. “New” would be likely not to be assimilated to any L1 category and, therefore, a new L2 category would be created. “Identical”, on the other hand, would be those sounds that the speaker assimilates, producing them in a native-like manner. Meanwhile, the ones classified as “similar” would be the most complicated sounds to both perceive and produce. The aim of this paper is to determine whether SLM can explain and predict FL (instead of L2) production mistakes by classifying them acoustically, i.e. using spectrograms. The data was provided by the supervisor of this paper, Dr. María Luisa García Lecumberri, and gathered at the University of Seville using the DiapixFL method as part of a larger corpus. This means that in order to prompt conversational speech between two subjects, a spot-the-difference task was used. In this case, native speakers of Spanish (more specifically, the Andalusian Spanish variety) were recorded. Afterwards, their production of English as a Foreign Language (EFL) was orthographically and phonetically transcribed and eventually classified by creating spectrograms. Finally, this classification of consonants and vowels was analysed in order to reach the following conclusion. The analysis of the results shows the limitations of SLM when dealing with phenomena such as hyper-correction and spelling pronunciation, as well as specific features of the variety of the study subjects. It would appear that, even though the acoustic analysis offered resourceful results, a more detailed research is preferable for upcoming occasions. Moreover, future collection of data is suggested, specifically in the Catalan and Galician regions. This addition could offer the chance to compare the production of English phonemes in native speakers of different varieties of Spanish.