UC Berkeley

Information from auditory feedback plays an important role in speech motor control. This has been shown in particular in experiments that make real-time changes to speakers’ own productions, which are fed back to them while speaking (e.g. Houde and Jordan 1998). In response, speakers change their articulations so that their own altered feedback sounds more like the target word. While such experiments have succeeded in showing the importance of the role of auditory feedback, they have focused on the acoustic output of the speech system, the fact that speakers did change their production in response to altered auditory feedback. The experiments presented here focus on how speakers change their motor plans using the novel addition of ultrasound imaging. Well-documented cases of articulatory variation in American English provided the test case to study this question. The articulatory strategies which speakers use to compensate for the altered feedback are evidence of how much knowledge they have about how their articulations map onto acoustics. These experiments consider how individual vocal tract anatomy might affect this knowledge.

Chapter 2 uses an articulatory synthesizer with three different palates to show that there is a more sensitive map between articulation and acoustics for flatter than more domed palates. This provided the motivation for Chapter 3, which tested whether these differences influenced variability in the production of normal, unperturbed speech. This ultrasound experiment found that people with flatter palates reduce their articulatory variability in comparison with people who have more domed palates. Reducing articulatory variability effectively constrains acoustic variability in /r/ production but not /s/ production.

Chapter 3 found significant differences in articulatory variability that correlated strongly with palate shape. Chapter 4 presents an experiment that uses ultrasound to capture the strategies that speakers use during adaptation to altered auditory feedback to test whether these production differences determined by palate shape carry over into how speakers make corrections to speech production on line. In one block, speakers said the target word “head,” and their F1 was gradually raised. In a second block, speakers said the target word “hood,” and their F2 was gradually raised. The ultrasound revealed how speakers search their articulatory-acoustic space and settle on a compensation strategy. A global effect of palate shape was not seen in compensation; degree of compensation did not correlate with palate shape. If palate shape does inform articulatory habits or the detail in a speaker’s knowledge of how their articulation maps onto acoustics, this knowledge does not influence speakers’ compensation degree. Despite a lack of overall trend in amount of compensation, palate shape did correlate with secondary characteristics in compensation.

Finally, Chapter 5 describes a preliminary study where speakers said the target word “heard,” and their F3 was gradually raised. The modeling in Chapter 2 predicted that a generally effective way to lower F3 is to raise and retract the tongue to minimize constriction size, a strategy that every speaker used.

The results from this dissertation suggest that while palate shape influences articulatory and acoustic variability in normal, unaltered speech, this behavior is not predictive of how much speakers will compensate when presented with altered feedback, only of certain secondary characteristics of this compensation. Speakers’ overall compensation is likely impacted more by higher-level factors which may influence sensitivity to auditory and somatosensory feedback.