Children's acquisition of Mandarin tones in context

Mandarin is a tonal language with four lexical tones and two contextual tones, i.e., neutral tone and tone sandhi, exhibiting tonal variations across contexts. Previous studies found that Mandarin-learning children acquire lexical tones early (before 3yrs.), but the neutral tone and tone sandhi are later acquired (after 4;6). Children with hearing impairment/cochlear implants (CIs) have problems in acquiring lexical tones unless implanted early or have long CI experience, since CIs do not transmit pitch information effectively; it is unclear how those implanted early perform on contextual tones. The general aim of this thesis was therefore to better understand the acquisition of contextual tones by both typically developing children and those with hearing impairment using acoustic, rather than perceptual measures. This thesis consisted of seven studies. Firstly, we examined how contextua ltones are realized in children's language input, i.e., infant-directed speech (IDS) and clear speech, where slower, hyperarticulated speech (as directed to children or possibly hearing impaired populations) might destroy the context for appropriate realization of contextual tones, thus potentially explaining the later acquisition of these tonal processes (Studies 1 & 2). However, our findings showed that the key features of contextual tones are well realized in both registers, suggesting that later acquisition is not due to the input. Next, we examined children's contextual tone productions in novel (rather than known) items, exploring when their knowledge of contextual tones becomes productive (Studies 3, 4 & 5). The results showed that (1) normal hearing 3-year-olds have already acquired productive knowledge of contextual tones, correctly producing tonal variations across contexts, though adult-like acoustic implementation is not fully mastered until age 5; (2) children with CIs face challenges in producing correct contextual tones, but early implantation (before age 2) facilitates lexical and contextual tonal acquisition. Finally, we tested children's perception of lexical tones and tone sandhi in novel compounds using a mispronunciation/eyetracking task (Studies 6 & 7). The results showed that, (1) all normal hearing participants could detect lexical tone mispronunciations, but none were sensitive to the novel compound tone sandhi mispronunciations; and (2) even detecting lexical tone mispronunciations was a challenge for the children with CIs. Taken together, the findings of this thesis suggest that contextual tones are not later acquired by normal hearing children, and acquiring typical contextual tones is possible for children with hearing impairment as long as they receive CIs early.