Abstract This study investigates how social expectations influence the perception of phonetic convergence in dialectal speech. Using an AXB discrimination task, we tested whether Seoul Korean listeners could detect convergence in North Korean speakers’ vowel productions, and whether this perception was modulated by explicit information about speaker identity. Four vowels (/ɛ/, /ɨ/, /o/, /ʌ/) were selected based on their dialectal salience. In Experiment 1, listeners judged similarity without knowing the speaker’s identity. In Experiment 2, listeners were told the speaker was North Korean. Results showed that convergence was generally perceptible, with /ʌ/ yielding the highest accuracy. However, when dialectal identity was made explicit, perceptual accuracy declined for /o/ and /ʌ/, suggesting that strong top-down expectations hindered categorization of acoustically ambiguous tokens. In contrast, accuracy for /ɛ/ and /ɨ/ marginally improved, consistent with their more flexible phonological status across dialects. These findings support socially enriched models of speech perception, showing that listeners use internalized, socially indexed phonological templates that shape convergence detection. Social cues enhance perception when they align with acoustic input, but disrupt it when expectations override the signal. This study demonstrates the category-specific and dialect-sensitive nature of expectation-driven perception in socially marked speech.
Jung et al. (Tue,) studied this question.