The traditional approach to determine which psychophysical variables are important for speech perception by cochlear implant (CI) users has been to perform correlational analyses. However, the finding of a statistically significant correlation between psychophysical and speech perception variables does not explain the actual mechanisms CI users employ to identify speech sounds. A mathematical modeling approach is proposed that does provide a detailed description of how CI users identify speech sounds based on the acoustic information they receive. To use the proposed approach, it must be hypothesized what the relevant psychophysical dimensions are for a given stimulation strategy. In this study, it is hypothesized that users of the SPEAK stimulation strategy use three types of cues to identify consonants: first, the average cochlear locations stimulated in response to the first three formants; second, the silent gap duration (if present); and third, the ratio between total energy above and below 1000 Hz. Based on a subject’s performance along these psychophysical dimensions, the model outputs a predicted consonant confusion matrix. Preliminary comparisons between model predictions and actual data indicate that the model was able to predict which pairs of consonant sounds would be most frequently confused by CI users. [Work supported by NIH.]