Reconfiguration of the perceptual space during second language speech learning

Abstract

In second language (L2) speech acquisition, it has been observed that identification and discrimination of L2 phonemes, as well as the categorical perception of certain phoneme pairs, are significantly different between L2 learners and native speakers. These differences could be due to the influence of the native language (L1) of the L2 learners. Many previous models addressing these questions were based on a single phoneme or a pair of phonemes. For instance, the Perceptual Assimilation Model (PAM) focused only on a pair of phonemes, and ignored the acoustic dimensions. The Speech Learning Model (SLM) and the Native Language Magnet model (NLM) focus more on the acoustic dimensions, but restricted to one or two phonemes. Nonetheless, the NLM revolutionarily introduced the multiple acoustic dimensions between the two phonemes studied, and revealed the distortion of the perceptual space, compared with the equally distributed acoustic space. The multiple acoustic dimensions perspective was later extended in the Attention-to-Dimension (A2D) model, which considered the whole acoustic dimension beyond the two phonemes. From this perspective, the phonemes are interrelated within as well as between L1 and L2. Therefore, the perceptual space shaped by the long-term language experience should be studied in a more systematic way. It is important to investigate not only the detailed acoustic features of the phonemes in each language, but also the acoustic dimensions that distinguish those phonemes, as well as the acoustic differences between the phonemes in L1 and L2, especially those sharing identical IPA symbols in the two languages. During L2 speech learning, the perceptual space of L1 speakers undergoes the following changes or reconfigurations: (1) The sensitivity or attention to the acoustic dimensions used to distinguish certain phonemes is enhanced; (2) The sensitivity to acoustic changes around the phoneme prototype is decreased; (3) Along with (2), the sensitivity around the phoneme boundaries, especially those having rapid formant transitions, is increased; (4) The multiple acoustic dimensions gain different weights for perceiving certain phonemes, which may cause discrepancies in hearing the L2 phonemes, especially those having different acoustic parameters but sharing the same IPA symbols as the L1 phonemes; Noting these, we design some new experiments that incorporate more phonemes in L1 and L2 to reveal not only how language experience helps shape the perceptual weighting of the acoustic dimensions, the magnet effect and the boundary effect around and between phonemes, but also how certain acoustic cue differences between the correspondent phonemes in L1 and L2 affect the outcome of perception. The preliminary results reveal a global effect of the acoustic dimension weighting, and some specific perceptual changes around the L1 phonemes, both of which affect the perception of the L2 phonemes. This study proposes a novel theoretical framework and a new research approach to investigate L2 speech learning.