UCLA

This dissertation investigates the development of cross-linguistic transfer in the use of strategies to repair preconsonantal word-external empty onsets (i.e., /C#V/) in Spanish heritage speakers. The main focus of this work is to analyze the use of /ʔ/-epenthesis as a strategy to repair empty onsets, as its use is asymmetrical in the Spanish and English phonologies.

A total of 190 participants (i.e., monolingually raised Spanish speakers, monolingually raised English speakers and Spanish heritage speakers) took part in Spanish and/or English word production tasks that elicited function + content word sequences in real words (e.g., Sp. un elefante 'an elephant', En. an umbrella) and in phonotactically legal novel words (e.g., Sp. un anbo, En. all embos). Participants are divided in three main age groups: younger children, older children, and adults.

First, I examine whether the use of glottal phonation is a repair strategy employed more often in English than in Spanish, and whether it is conditioned by age. Overall, I find that in English glottal phonation is the preferred strategy to repair prosodically-prominent onsetless syllables, whereas, in Spanish, the preferred repair strategy is modal phonation in both prosodically and non-prosodically prominent syllables. For the Spanish real words, I find no effects of language development in the experiments with real words. In the experiment with novel words, I show that older children and adult children produce a greater rate of glottal phonation than younger children, which I explain as an effect of the unpredictability of the novel words in the context. In English, the rate of glottal phonation in the real words is not affected by age. For the English novel words, older children produce greater rates of glottal phonation than the adults, demonstrating a U-shaped behavior during language development.

Next, I formalize the Spanish and English results using MaxEnt grammars and accounting for three repair strategies: /ʔ/-epenthesis, no repair (i.e., coda consonant), and resyllabification/ambisyllabicity. I introduce the effect of prosodic prominence in the grammars using constraints on phonological enhancement.

I then compare the results of the Spanish heritage speakers to those of the monolingually raised Spanish and English speakers. For the Spanish real and novel words, I find that the younger and older child HS produce greater rates of glottal phonation than the monolingually raised Spanish speakers, but no significant difference is observed between the adult monolingually raised Spanish speakers and the adult heritage speakers. To account for these results, I claim that child heritage grammars are more permeable to cross-linguistic transfer than adult heritage grammars. In addition, in the Spanish real words, my results show that heritage speakers with a higher amount of Spanish output glottalize less often than heritage speakers with a low amount of Spanish output, which brings to the forefront the importance of language use to control transfer from the majority language. In English, my results show that, overall, heritage speakers produce lower rates of glottal phonation than monolingually-raised English speakers (i.e., the dominant language may also suffer from negative transfer from the heritage language).

Finally, I argue that the results for the heritage grammars can be formalized using a model of language coactivation during input evaluation � la Goldrick et al. (2016). The two sets of language-specific constraints are scaled by the speaker-specific activation scores, which set the intended language and reflect the speaker's language dominance. To account for the apparent development of language control during language maturation (i.e., greater permeability of language transfer during adulthood), I posit the existence of a prior applied to the constraint weights during input evaluation.