(Dis)connecting perception and production: Training native speakers of Spanish on the English /i/-/I/ distinction

This dissertation features three experiments that investigated how perception and production are connected in the acquisition of second language (L2) phones by comparing the effectiveness of two modality-specific trainings and their respective potential for cross-modality gains. Participants were native speakers of Spanish with advanced English proficiency, and the targets were the English vowels /i/ and /I/.;In Experiment 1, participants (n=15) received perception-only training; they heard auditory exemplars of the target phonemes but never produced the sounds. In Experiment 2, two variations of a production-only training were compared that either allowed or denied access to the auditory feedback loop. A first group (n=14) underwent training using a computer program that provided real-time visual representations of spoken vowels. They never heard any other-produced auditory tokens of the target sounds, although they could hear the sound of their own voices. A second group (n=15) underwent the same training, but wore noise-cancelling headphones and listened to white noise. This ensured that they never heard other- or self-generated tokens of the target phonemes, for the first time in the literature truly isolating production from all auditory influence. All participants in both experiments completed a battery of pre- and posttests in perception and production, and they were also compared against a control group (n=15) and two baselines: a group of native speakers of English (n=20), and a bilingual group (n=16) who was deemed to have acquired /i/ and /I. In Experiment 3, the two baselines were directly compared in order to test their efficacy as benchmarks for phonetic training experiments.;Results revealed that: (1) perception-only training led to large gains in perception and no sizeable improvements in production; (2) production-only training led to variable results for production, and medium-sized improvements in perception; (3) access to the auditory feedback loop provided a benefit to production; (4) access to or denial of the auditory feedback loop did not affect cross-modal learning in perception; and (5) bilinguals are a fitting, and for many purposes likely sufficient, comparison baseline group in L2 speech training experiments.;The dissertation contributes novel theoretical, methodological, and educational insights to the L2 speech training literature.