Cochlear implants users report great difficulty in pitch perception. This problem is acute among implantees in tone language environments, and especially in children. This thesis addresses the development of perceptual training methods for children in tone language environments with cochlear implants via 5 experiments, each providing insight into the best methods for solving this problem. Experiment 1 was a comprehensive study of perceptual cues for Mandarin tones by normal-hearing adult native and non-native (Australian English) listeners. Participants discriminated between minimal pairs of Mandarin tones in 5 conditions: Auditory-Only, Auditory-Visual, Simulated Cochlear Implant (CI) Audio-Only, Simulated CI Auditory-Visual, and Visual-Only (silent). While discrimination was much weaker in CI audio than in undegraded audio, there was considerable variance among tone pairs. In particular, tone pairs disambiguated by strong non-f0 cues were more easily discriminated than those without. It was also found that visual speech information augmented tone perception in both native and non-native listeners, particularly in CI audio. In the Visual-only condition, both native and non-native listeners discriminated tones above chance, and, strikingly, Australian English speakers outperformed Mandarin speakers. This experiment suggests a possible role for visual speech information in successful training of lexical tone perception, and also provides valuable information regarding the relative perceptibility of different Mandarin tone contrasts for both native and naive speakers, for CI users, and with and without visual speech information. Experiment 2 attempted to train a group of normal-hearing tone-naive 3-year-old children in the perception of Mandarin tones, using a Go/No-Go training paradigm, in which participants listen to repetitions of one tone, and respond with a button press when there is a change to another tone. Correct responses are rewarded with a short animation. The experiment met with limited success - the training session was too long, too difficult, and across too broad a range of stimuli. Very few participants completed the entire session, and those that did showed limited perceptual improvement. Experiments 3 and 4 examined several facets of perceptual training. In Experiment 3, a group of normal-hearing tone-naive adult listeners were trained to discriminate a Mandarin tone contrast pair in either undegraded or CI audio, on auditory-only stimuli with exaggerated or diminished tone features. Participants trained adaptively (beginning with easier-to-discriminate exaggerated stimuli, and progressing to more difficult diminished stimuli) showed steady improvement across the course of the training session, whereas participants who were trained on a random schedule of stimulus difficulty showed apparent perceptual deterioration across the course of training. However, difficulties in interpreting data from the Pre- and Post-tests (owing to large differences in difficulty) meant that it was unclear what perceptual gains had actually been made by participants. Experiment 4 used a similar method to Experiment 3, but with an improved Pre- and Post-test, and feedback throughout the training session. In training, participants showed similar patterns of perceptual performance to Experiment 3 - however, the Post-test revealed that participants who had been trained on a random schedule showed much more perceptual improvement after training than those trained adaptively, who showed little to no improvement. This unexpected result appears to indicate that while increased task difficulty and variability may appear to offer little benefit in training, they in fact lead to the strongest perceptual learning outcomes. This conclusion is consistent with findings in the skill and visual perceptual learning literatures, but until now has not been considered in speech perception training research. In addition, data from Experiments 3 and 4 indicate that increased phonetic variability is an important factor in successful perceptual learning, and a comparison between Experiment 3 and 4 points to a clear benefit of incorporating feedback in training. Experiment 5 applied Experiment 1, 3 and 4 findings and built on the shortcomings of Experiment 2 in a pilot study on a small group of tone-naive 4-year-old children. Participants were trained on one Mandarin tone contrast in CI audio in a 20-minute session. Training used the same Go/No-Go paradigm as Experiment 2, using auditory-visual stimuli throughout, with either exaggerated or natural tonal features, and importantly incorporated relief blocks of easier (normal audio) stimuli, interspersed between training blocks. In contrast with Experiment 2, most participants completed the entire training session. Participants trained on unexaggerated stimuli showed improvement in tone perception in CI audio, whereas participants trained on exaggerated stimuli did not. In summary, it was found that there appear to be non-F0 cues available for lexical tone that may be available to CI users, including amplitude contours and both temporal and spatial visual cues, with a plausible physiological basis (head and laryngeal movements). These cues appear to be available cross-linguistically, suggesting that they may also be readily used by inexperienced listeners. They may also be underused by native Mandarin speakers. Further, it appears that effective methods for perceptual training may involve the use of schedules that appear to be more difficult in training than easier, but less effective, schedules. The use of phonetic variability and of feedback may be crucial in this regard, and should be kept in mind when constructing training programs. Importantly, this work provides further evidence that training schedules cannot be assessed outside of the context of a Pre- and a delayed Post-Test. Finally, a Go/No-Go method for training perception, combined with the use of relief blocks, appears to be effective at keeping children involved and engaged in the procedure, although its actual effectiveness as a training schedule has not been tested here. The approach taken to this research problem here allowed testing of general principles using large, homogeneous populations. Several important findings emerged from these experiments, with respect to auditory and visual cues for lexical tone, patterns of tone perception in CI users, perceptual training, and appropriate methods for training children. The findings have both general interest and immediate applicability to the problem at hand. Based on these results, a blueprint for a successful training program for CI-implanted children in tone language environments is proposed, incorporating visual speech (both in concert with auditory stimuli and in silence), stimuli with exaggerated and diminished tonal features, feedback, high phonetic variability, and highly variable task difficulty, conducted within a paradigm appropriate for use with children.
| Date of Award | 2010 |
|---|
| Original language | English |
|---|