A fundamental issue in speech science concerns the extent to which speech sounds are mentally represented by articulatory-motor and/or auditory-acoustic features. This dissertation aims to expand upon the current literature by investigating changes in production and perception following visual feedback training with either articulatory or acoustic speech targets. Eleven second language (L2) learners of English participated in a single session of pronunciation practice in which they produced either an English vowel (/ae/-/[open-mid front unrounded vowel]/) or consonant (/s/-/[voiceless postalveolar fricative]/) contrast while their speech movements and acoustics were recorded using an electromagnetic articulography system. Participants were randomly assigned to one of two training conditions (visual feedback or control) and one of two feedback conditions (articulatory or acoustic). Articulatory-based visual feedback was provided by a talker-controlled tongue-avatar, while acoustic-based visual feedback was provided by a real-time sound spectrograph. Changes in production for vowel contrasts were acoustically analyzed by measuring the Euclidean distance between the two vowels. A subset of vowel and consonant tokens were additionally judged by native English listeners in a forced-choice perceptual discrimination task. In general, the results showed that talkers who received visual feedback training moderately improved their production accuracy when compared to those exposed to the control condition, although this result did not reach significance. For vowel contrasts, acoustic and perceptual data demonstrated that articulatory-based visual feedback led to a similar magnitude of improvement as acoustic-based visual feedback. Of the two talkers who trained on consonant contrasts, the talker who practiced with articulatory-based visual feedback showed a greater increase in production accuracy than the talker who practiced with acoustic-based visual feedback. An analysis of the relationship between the changes in talkers' L2 production and perception following training revealed a significant positive correlation between production of the trained sound contrast and its perceptual discrimination. Overall, these findings do not support the hypothesis that a single session of visual feedback modifies talkers' internal representation for L2 speech sounds. These data are limited by the small number of participants and may reflect learning constraints imposed by a single training session. In addition, planned analyses of the kinematic data may yet reveal covert contrasts between the speech sound contrasts that are not evident in the acoustic and perceptual data. Future work, including additional analyses of individual subject data, is needed to fully understand the mechanisms underlying visual feedback instruction. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://bibliotheek.ehb.be:2222/en-US/products/dissertations/individuals.shtml.]