The L2 Production of Spanish Taps and Trills by L1 Romanian and L1 French Speakers

This dissertation explores two learning problems in L2 phonology: (1) acquiring a new L2 distribution when both sounds exist in the L1 as allophones; (2) acquiring the same L2 distribution when neither sound has a direct L1 counterpart. Previous literature has investigated these problems separately, focusing more on the first situation than the second. To explore these scenarios, I analysed the L2 production of Spanish taps and trills, which are contrastive word-medially intervocalically, by 10 L1 Romanian and 10 French speakers. 13 native Spanish controls were also included. These language pairings were chosen because Romanian has both rhotics as allophones (albeit in free variation), with the tap being the most common, and the trill sometimes occurring word-initially, while French has neither. Data was elicited using two tasks: a sentence reading (more formal, orthography) and a picture description task (less formal, no orthography). The stimuli included nonce and real words containing word-initial and word-medial rhotics. Because a secondary objective was to provide empirical evidence and acoustic analyses for rhotics in Romanian, participants also completed similar tasks in their L1. The results confirm previous observations that the most common Romanian rhotic is the tap. The trill mainly appears word-initially. The L2 data showed that the tap was less difficult than the trill, with both learner groups producing taps at high rates. However, both learner groups had low trilling rates. Though the L1 Romanian group produced more trills than the L1 French group overall, when the position of trills was considered, there were no differences between groups. Specifically, both learner groups had significantly higher trilling rates word-medially than initially, corroborating previous research, and there were no group differences. Furthermore, speakers produced more of the expected rhotics in the sentence reading than in the picture description task. Overall, given the low trilling rates, these findings suggest that recategorizing L1 allophones may be just as challenging as learning a new L2 distribution that does not have L1 counterparts, though the difficulties may be due to different reasons. Finally, this study also provides acoustic data supporting previous observations regarding Romanian rhotics.Ph.D