Helene Springer, Doctoral Student in General Linguistics, Lund University

Accurately perceiving speech sounds in a second language (L2) is one of the greatest challenges in L2 acquisition. A central assumption in all influential models of L2 speech perception (e.g., SLM, PAM, L2LP) is that perceptual similarity drives difficulty: L2 contrasts whose vowels are phonetically close to L1 categories and acoustically similar to one another are expected to be the most challenging. Previous work has focused on learners acquiring contrasts that are absent in the L1 (few-to-many), a task considered to be more difficult than acquiring fewer phonemic distinctions in the L2 (many-to-few). However, empirical baselines of perceptual similarity across these scenarios remain limited.

To address this gap, we examined perceptual similarity in two vowel systems. The nine long vowels of Swedish and the eight long vowels of German were presented to German L1 listeners naïve to Swedish (n=20) and Swedish L1 listeners naïve to German (n=20), respectively. In a free classification task, participants sorted the vowels based on perceived similarity with and without access to visual cues from the speaker’s mouth movements. Both languages contained a comparable number of perceptually similar contrasts, and the overall similarity patterns were strikingly similar across groups. These findings challenge the assumed difference in difficulty between few‑to‑many and many‑to‑few scenarios and instead suggest that both listener groups face similar perceptual challenges when acquiring the other language. Moreover, access to the speaker’s face had no impact on classification outcomes, suggesting that naïve listeners may lack the experience required to benefit from visual speech cues. 

In ongoing work, we test whether the perceptual similarity patterns can predict L2 learners’ ability to discriminate Swedish and German vowels in an ABX task. Preliminary results will be discussed in relation to models of L2 speech perception and audiovisual processing.