Cross-category phonological effects on ERP amplitude demonstrate context-specific processing during reading aloud

Abstract

The phonetic realisation of speech sounds depends on their context: e.g., /t/ is aspirated in ‘top’, but unaspirated in ‘stop’. Similarly, Beijing Mandarin Tone 3 (T3) usually has a low contour, but preceding another T3 syllable has a rising contour (sandhi T3). Importantly, Tone 2 (T2) also has a similar rising contour. Yet how such phonetic variation is processed online during reading and speech production is not well understood. ERP amplitude was measured as native Mandarin speakers read aloud words preceded by a 48ms masked prime. Prime and target always differed in tone category. Critical targets were T2-initial words. Primes were either sandhi T3 (contour-match) or low T3 (mismatch). Generalised additive mixed models (GAMs) revealed a complex interaction between prime type, prime frequency, and target frequency emerging around 100ms and 300–400ms following target presentation. In the contour-match condition, when prime and target frequency were both high or both low, there was increased negativity, suggesting competition between prime and target. In the mismatch condition, there was relatively little effect of item frequencies. This difference in the pattern of effects between the contour-match and mismatch primes provides evidence for top-down effects of context on phonological processing of masked primes during reading aloud.