Analyses were performed on the spontaneous speech samples produced by profoundly hearing-impaired children who used the 3M/House single-channel cochlear implant (n = 7), the Nucleus multichannel cochlear implant (n = 7), or the two-channel tactile aid, the Tactaid II (n = 12). Speech recordings were obtained in the predevice condition, and after 6- and 12-months use. The speech of the children in the three experimental groups was compared to that of a control group of subjects who used a tactile aid inconsistently and received minimal speech training. Several new analysis procedures were developed to describe adequately the characteristics of the children's speech. The results revealed that all three experimental groups and the control group showed an increase in their production of sounds classified as English phonemes in the postdevice conditions. It was apparent at both the 6- and 12-months postdevice intervals that the most dramatic improvements were made by the users of the Nucleus device. The changes made by the Tactaid and 3M/House users were similar. The subjects with the Nucleus device also developed more diverse phonetic repertoires over time than did the other experimental subjects. More subjects with the Nucleus device acquired fricatives, liquids, glides, voiceless consonants, and high front vowels than did the subjects in the other groups. The subjects in the control group developed the most limited phonetic repertoires.