Addressing phonological questions with ultrasound

Abstract

Ultrasound can be used to address unresolved questions in phonological theory. To date, some studies have shown that results from ultrasound imaging can shed light on how differences in phonological elements are implemented. Phenomena that have been investigated include transitional schwa, vowel coalescence, and transparent vowels. A study of consonant cluster phonotactics is presented as an example of how ultrasound methodology can be used to examine phonological issues. Five English speakers presented with phonotactically illegal non‐words (e.g., /zgomu/) typically repaired these sequences with vowel insertion (e.g., [z▒gomu]). Using ultrasound imaging, the production of these words is compared to legal sequences that are articulatorily similar, such as succumb and scum to assess the nature of the schwa found between /z/ and the following consonant. Results indicate that for some speakers, production of schwa in /zC/ sequences is not consistent with the phonological epenthesis of a schwa. Instead, speakers appear to be failing to sufficiently overlap the consonant gestures.

• Ultrasound
• phonology
• phonotactics
• tongue

Acknowledgements

Thanks to Maureen Stone, Stefan Benus, David Goldberg, Christine Mooshammer and Amanda Miller‐Ockhuizen for their insightful comments and/or help with the data. Parts of this work were presented at WCCFL 22 and submitted as part of the author's doctoral dissertation (Johns Hopkins University, 2003). This research was supported by the IGERT program in the Cognitive Science of Language at Johns Hopkins University, National Science Foundation Grant 997280 and by the National Institutes of Health Grant DC01758 to Dr. Maureen Stone.

Notes

Because superfluous has alternate pronunciations of either [sup]erfluous or [s▒p]erfluous, only speakers who clearly produced the schwa variant were recorded.

Attempts to demonstrate that schwa is not a segment but an interpolation between the surrounding segments have been forced to conclude that schwa does have a gestural target (Browman & Goldstein, 1992; Kondo, 1994). For example, Kondo's acoustic data indicate that while schwa might not be specified for backness (i.e., F2 is affected by the surrounding environment), it does have a consistent F1 value indicating that schwa does have a target height.

Another possible measurement is the root mean square, which is very similar to the mean distance measure in equation 1, except that it uses squared values. This amplifies the larger distances and reduces the influence of the smaller ones. In general, the two measures return very similar results, but the absolute value of the distance is chosen as being a more direct measure.

Although each speaker produced two tokens for each of the /zC/ targets, only the stimulus with the best image for each speaker was measured.

One reviewer suggests that an alternative explanation for all of the speakers' behavior is that they are producing lexical schwas, but that they are undershooting the targets. The theory of undershoot was proposed by Lindblom (1963), who demonstrated that as Swedish vowels decrease in duration, they also tended to fail to reach their target formant frequencies. However, subsequent studies failed to show similar effects in other languages, including American English (Gay, 1978; Fourakis, 1991) and Dutch (van Son and Pols, 1992). Moon and Lindblom (1994) addressed these critiques by demonstrating that undershoot occurs when there is a large difference between the tongue position for C1 and the tongue position for the following vowel (as in the [wI] of will and willing). In this study, all C1s are coronal fricatives, and all following vowels are schwas. Even if some speakers are inserting schwa, this combination of sounds, in addition to the short duration of schwa, is very unlikely to lead to undershoot.