Explanations of language learning often involve appeals to distinct learning mechanisms. On one hand, learners’ innate characteristics are emphasized, with learning tied to a limited time period, a critical period, when the brain is predisposed for success in language learning. This view is often contrasted with a position emphasizing the role of the environment in shaping language learning, highlighting the contribution of feedback mechanisms and the nature of the speech input. One approach has been to examine how second languages are learned in order to directly examine change due to learning. To this end, the present paper documents the behavioral and cortical changes resulting from learning a novel language contrast, specifically Mandarin tone. Hemispheric differences in the processing of language contrasts are observed, with significant left hemispheric participation in native listeners and no hemispheric preference for non-native listeners. Additional experiments examined the training of non-native listeners, revealing that tone perception accuracy can be improved with minimal exposure. Furthermore, it can be generalized to new stimuli and talkers, retained for at least six months, and transferred to production. Native listeners identify post-training productions more accurately than pretraining productions and acoustic analyses of the post-training F0 contours show better approximation to native speaker norms. These behavioral changes due to training can also be observed cortically, with the learning of Mandarin tone contrasts associated with significant increases in activity in the traditional language areas (left hemisphere superior temporal gyrus) as well as the recruitment of neighboring neural areas. Implications for theories of language learning will be addressed.