In multi-channel cochlear implants, electrical current is delivered to appropriate electrodes in the cochlea to approximate the spatial representation of speech. Theoretically, electrode configurations that restrict the current spread within the cochlea (e.g., bi- or tri-polar stimulation) may provide better spatial selectivity, and in turn, better speech recognition than configurations that produce a broader current spread (e.g., monopolar stimulation). However, the effects of electrode configuration on supra-threshold excitation patterns have not been systematically studied in cochlear implant patients. In the present study, forward-masked excitation patterns were measured in cochlear implant patients as functions of stimulation mode, level and location within the cochlea. All stimuli were 500 pulses-per-second biphasic pulse trains (200 micros/phase, 20 micros inter-phase gap). Masker stimuli were 200 ms in duration; the bi-polar configuration was varied from narrow (BP+1) to wide (BP+17), depending on the test condition. Probe stimuli were 20 ms in duration and the masker-probe delay was 5 ms; the probe configuration was fixed at BP+1. The results indicated that as the distance between the active and return electrodes in a bi-polar pair was increased, the excitation pattern broadened within the cochlea. When the distance between active and return electrodes was sufficiently wide, two peaks were often observed in the excitation pattern, comparable to non-overlapping electric fields produced by widely separated dipoles. Analyses of the normalized data showed little effect of stimulation level on the shape of the excitation pattern.