Based on data from a large scale vowel identification experiment, this paper will discuss empirically grounded ways to operationalize perceptual similarity in L2 by integrating contributions from SLA research, experimental phonetics and cognitive psychology. A method to visualize perceptual similarity of L2 categories by Multidimensional Scaling is presented.
Similarity is one of the central concepts in many models of L2 speech perception and acquisition (e.g. Best 1995; Flege 1987, 1995; Kuhl 1992, 1993), but few attempts have been made to operationalize it for speech perception. This contribution aims to critically discuss a view of perceptual similarity of L2 sounds in phonetic-acoustic terms. The claim that this is insufficient to understand perceived similarity in L2 will be illustrated by experimental data from a comparative study on L2 vowel perception (Author 2014) with 173 learners of German from 10 different L1-subsamples (+ a native control group) who performed a vowel identification task. We will discuss notions of (1) objective vs. subjective similarity, (2) inter- vs. intralingual similarity, and (3) empirically grounded ways to operationalize perceptual similarity within the TL in terms of (a) acoustic similarity and (b) a visualization of psychological similarity in a spatial MDS representation of the L2 vowel space (Shepard 1972, 1980; Johnson 2012). Alternatively to more traditional mono-directional conceptions of similarity between L1 and L2, we favour a cross-linguistic influence-approach focussing on biases associated with properties of stimuli (acoustic-phonetic) as well as responses (phonological). The cross-language comparison of the L1-subsamples shows that – rather than predicting perceptual similarity directly from phonetic properties – perceptual similaritysij between TL categories has to be modelled as the result of the complex interaction of phonetic proximitypij, stimuli biases bi and response biases bj (sij = pij * bi * bj). I will show that biases vary according to several factors such as attentional tuning to specific cues in the signal and the learner's L2 experiences and proficiency, but are also influenced by the experimental setting, the set of stimuli and response categories presented in the experiment and the learner's expectations and hypotheses of the target language vowel system.