Testing for underlying representations: Segments and clusters in Sevillian Spanish

Finished · Phonological Learning

Summary

Gilbert examines [Ch] sequences present in Sevillian Spanish, conducting a perception experiment to demonstrate that Sevillian speakers, compared to speakers of other Spanish dialects, are sensitive to the duration of [h] in these sequences, in that they treat a high-duration [h] as if it were an /sC/ cluster and not /Ch/. Diachronically, /s/ has debuccalized to /h/ in many Spanish dialects, including Sevillian. Furthermore, in Sevillian, the debuccalized /h/ has metathesized with the following consonant (with variation in this metathesis). Gilbert conducts a perception study with Sevillian, Mexican, and Argentinian Spanish speakers to see whether the duration of the /h/ in [Ch] clusters is interpreted as a preceding s or not. Mexican speakers do not debuccalize or metathesize /s/, while Argentinian speakers debuccalized but do not metathesize. The study examined the cross-morphemic effects of debuccalization and metathesis on the verb tener `to have'. She had speakers listen to a sequence like [* tjene phali], where the asterick is a beep to be filled in by the participant, and the h following the p was manipulated to vary in duration. For Sevillian speakers, Gilbert predicted that they would be more likely to choose a 2SG subject as the duration of the h in [phali] increased because they would reconstruct /tjenes pali/ if the /h/ was analyzed as underlying /s/ in the preceding verb. These are the results she got, and it was very obvious. The other dialects of Spanish did not show this pattern, which is expected since they don't metathesize.

Her conclusion was that Spanish [Ch] clusters are underlyingly /sC/. Moreover, she did touch briefly on whether theories that derive representations from statistical averaging over surface forms could arrive at this representation. It is difficult for me to see why they couldn't, since there is variation in the realization of [Ch] with [CC], [sC] and others. While the methathesized [Ch] is the most frequent by far, listeners could categorize all of these realizations within the same lexical category and retrieve the 2SG interpretation from all of them. Thus, it doesn't necessarily entail that some UR must exist.