The two main questions this dissertation is concerned with are (i) How does perceptual sensitivity develop along a dimension that contrasts two unknown speech sounds, and (ii) Does perceptual development vary with the learner's age? With respect to the first question, two hypotheses were tested. One of these, acquired distinctiveness, states that listeners initially have difficulty perceiving differences along a dimension that contrasts speech sounds that are non-phonemic in their native language. As a result of learning, perceptual sensitivity to speech sounds that are categorized differently increases. Acquired similarity, on the other hand, assumes that listeners are initially good at perceiving differences along a dimension that contrasts speech sounds that are non-phonemic in their native language. As a result of learning, perceptual sensitivity to speech sounds that are categorized together decreases. It was expected that school-aged children as well as adults would learn to perceive nonnative phoneme contrasts through acquired distinctiveness. The main methodology of this research was a pretest-training-posttest design in which listeners from different age groups were tested on their perceptual development of nonnative phoneme contrasts. Listeners were trained on speech continua from five different speakers, thus introducing inter-speaker variation. They were tested on a phoneme continuum from a sixth speaker to address language learning instead of the learning of a single speech continuum. Moreover, posttests were run one day after the last training session, and not directly after training, in order to address effects of linguistic learning rather than phonological priming. If training resulted in robust category formation, perceptual sensitivity along the test speaker's phoneme continuum was expected to change. With respect to the question of how perception develops, it was found that perceptual sensitivity along a relevant acoustic dimension changes through acquired distinctiveness. Earlier research suggested that learning occurs through acquired distinctiveness, but no prior attempt was made to provide direct evidence for this learning strategy with respect to the development of phoneme contrasts. The increases in perceptual sensitivity near the learned phoneme boundaries remained small, however, and did not - with short-term training - develop into native-like discrimination peaks. In response to the second question it was shown that 12-year-old children and adults learn to perceive nonnative phoneme contrasts in similar ways. The relatively small effect of short-term training on perceptual sensitivity that was found in several experiments within this study contrasts with the claim that short-term training can change phonemic perception in learners. Whereas the conclusions in this study were based on a comparison of both classification and discrimination experiments, the earlier conclusion was drawn on the basis of classification results only. That less complete design may have resulted in an overestimation of the learner's L2 proficiency. Only with long-term training, native-like peaks in perceptual sensitivity seemed obtainable. Short-term training is likely to improve listeners' performance, but it is likely that the category representations established in learners during such training differ from the phoneme representations of native listeners.