Three experiments studied the effect of stimulus polarity on the Electrically Evoked Compound Action Potential (ECAP) obtained with the masker-probe paradigm on different sites along the cochlea in cochlear implant users. Experiment 1 used a biphasic cathodic-1st (BIC) masker and showed that ECAP N1 peak latencies were longer for BIC than for biphasic anodic-1st (BIA) probes on all electrodes under test. Both the latency of each probe as well as the latency difference between BIA and BIC probes increased when the phase width (PW) of the masker and probe were increased together. Experiment 2 used maskers with long inter-phase gaps (IPGs), and, by manipulating the polarity of the second phase (closest in time to the biphasic probe), showed that only an anodic phase could mask the probe response. Experiment 3 used maskers and probes with long IPGs and measured ECAPs to the first phase of the probe; ECAPs could be measured when both this phase and the second phase of the masker were anodic, but not when they were cathodic. Our results extend those of a previous study, showing that the auditory nerve in humans is preferentially activated by anodic stimulation, to different sites along the cochlea.