The present results demonstrated that all 3 factors --lexical difficulty, stimulus variability, and word length--significantly influenced spoken word recognition by children with multichannel cochlear implants. Lexically easy words were recognized significantly better than lexically hard words, regardless of talker condition or word length of the stimuli. These results support the earlier findings of Kirk et al(12) obtained with live-voice stimulus presentation and suggest that lexical effects are very robust. Despite the fact that listeners with cochlear implants receive a degraded speech signal, it appears that they organize and access words from memory relationally in the context of other words. The present results concerning talker variability contradict those previously reported in the literature for listeners with normal hearing(7,11) and for listeners with mild-to-moderate hearing loss who use hearing aids.(14) The previous investigators used talkers and word lists different from those used in the current study and found that word recognition declined as talker variability increased. In the current study, word recognition was better in the multiple-talker condition than in the single-talker condition. Kirk(15) reported similar results for postlingually deafened adults with cochlear implants who were tested on the recorded word lists used in the present study. Although the talkers were equally intelligible to listeners with normal hearing in the pilot study, they were not equally intelligible to children or adults with cochlear implants. It appears that either the man in the single-talker condition was particularly difficult to understand or that some of the talkers in the multiple-talker condition were particularly easy to understand. Despite the unexpected direction of the talker effects, the present results demonstrate that children with cochlear implants are sensitive to differences among talkers and that talker characteristics influence their spoken word recognition. We are conducting a study to assess the intelligibility of each of the 6 talkers to listeners with cochlear implants. Such studies should aid the development of equivalent testing conditions for listeners with cochlear implants. There are 2 possible reasons the children in the present study identified multisyllabic words better than monosyllabic words. First, they may use the linguistic redundancy cues in multisyllabic words to aid in spoken word recognition. Second, multisyllabic words come from relatively sparse lexical neighborhoods compared with monosyllabic tokens. That is, multisyllabic words have fewer phonetically similar words, or neighbors, competing for selection than do monosyllabic stimuli. These lexical characteristics most likely contribute to the differences in identification noted as a function of word length. The significant lexical and word length effects noted here may yield important diagnostic information about spoken word recognition by children with sensory aids. For example, children who can make relatively fine phonetic distinctions should demonstrate only small differences in the recognition of lexically easy versus hard words or of monosyllabic versus multisyllabic stimuli. In contrast, children who process speech using broad phonetic categories should show much larger differences. That is, they may not be able to accurately encode words in general or lexically hard words specifically. Further study is warranted to determine the interaction between spoken word recognition and individual word encoding strategies.