The Role of Contextual Variation in Learning Cantonese Tones from Naturalistic Speech
Fengyue Lisa Zhao, Jennifer Kuo
A central question in speech acquisition is how infants learn contrastive sound categories from variable acoustic cues. The Distributional Learning Across Contexts (DLAC) proposal suggests that infants may identify contrastive sound dimensions by tracking distributions across contexts, with greater contextual variation signaling contrastive categories. Most research on acquisition of sound categories has focused on simple two-way vowel and consonant contrasts, leaving open how infants learn more complex or tonal contrasts. This study addresses these gaps by analyzing naturalistic speech to examine whether the ease of learning Cantonese’s six tonal contrasts is linked to their variation across different contexts. By comparing our findings with existing acquisition data, we show that this proposal could predict which contrasts are easier or harder to learn, suggesting that in absence of invariant acoustic cues, infants may rely on contextual variation to acquire complex sound categories.