This study is one of a series that examines stimulus features important for cochlear implant function. Here, we examine effects of stimulus level. In subjects with cochlear implants, a number of psychophysical tests of temporal discrimination (pulse rate discrimination, gap detection, etc.) show marked improvement as a function of stimulus level through most or all of the dynamic range, while electrode-place discrimination can improve or degrade as a function of level. In this study, effects of these combined potential influences were studied by examining the effects of stimulus level on syllable identification. We tested two hypotheses: that syllable identification varies as a function of stimulus level and that level and electrode configuration interact in affecting syllable identification. We examined vowel and consonant identification as a function of stimulus level for bipolar and monopolar electrode configurations. We used experimental processor maps where upper and lower stimulation limits of each electrode pair were equated to eliminate confounding effects of dynamic range, which varies across subjects and electrodes. For each channel, stimulation amplitude was set to a fixed percentage of its dynamic range. Eight adult subjects with Nucleus CI24M implants were tested using the SPEAK processing strategy. With each electrode configuration, stimulus levels were tested from 0% to 90% of the dynamic range in nine steps. The effects on consonant and vowel identification were similar. Phoneme identification was usually better for monopolar than for bipolar stimulation. In the lower half of the dynamic range, syllable identification usually increased as a function of stimulus level. In the upper half of the dynamic range, syllable identification continued to increase as a function of level to 90% of the dynamic range for some subjects, while for others there was no appreciable change or a decrease as a function of level. Decreases in performance at high levels were more common with monopolar than bipolar stimulation. These results suggest that if speech processors are programmed to optimize level for each individual, speech perception performance could be improved.