We examined the production and perception of voiced versus voiceless obstruents by thirty-three adolescent heritage speakers of Polish and Russian. First, a word list task was used to elicit the production of voiced and voiceless plosives in word-initial position. Voice onset time (VOT) values for both sets of stops were compared to the values reported for monolingual Polish and Russian speakers. To investigate the perception of phoneme contrasts in the heritage language, we used an auditory phoneme discrimination task that contained ten minimal pairs of real and nonce words with contrasting voiced and voiceless obstruents. The results showed an almost perfect perceptual discrimination and a separation of voiced and voiceless plosives in production; however, the VOT values differed significantly from those of monolinguals. Both groups showed convergence of VOT values towards the values reported for fortis and lenis stops in the majority language (i.e., German), leading to a non-native accent in the heritage language.

This study investigates the perceptual accuracy of eight English obstruents in the onset and coda position by Mandarin and Korean-speaking L2 learners and by a control group of native English speakers. According to the current theoretical models on second language speech learning, L1 Mandarin and Korean speakers are expected to differ in their perception of English obstruents due to the different correspondence between their respective L1 obstruents and those in English. On the other hand, theories based on intrinsic differences in the difficulty of different linguistic skills imply that some L2 sounds would be more difficult than others regardless of the L1 background. The results showed that all three groups were significantly more accurate in perceiving obstruents in the onset than in the coda position, voiceless than voiced targets, stops than fricatives, and labials than coronals. /θ/ and /ð/ were particularly poorly identified. The two learner groups were equally accurate in the onset position, but the Mandarin group outperformed the Korean group in the coda position. Regarding the specific obstruents, some patterns were predicted by mapping to the L1. Nonetheless, the general similarity between the two groups suggests a robust and pervasive language-independent tendency in speech perception.