Different from normal-hearing (NH) listeners, speech recognition thresholds (SRTs) in cochlear implant (CI) users are typically poorer with dynamic maskers than with speech-spectrum noise (SSN). The effectiveness of different masker types may depend on their acoustic and linguistic characteristics. The goal of the present study was to evaluate the effectiveness of different masker types with varying acoustic and linguistic properties in CI and NH listeners. SRTs were measured with nine maskers, including SSN, dynamic nonspeech maskers, and speech maskers with or without lexical content. Results showed that CI users performed significantly poorer than NH listeners with all maskers. NH listeners were much more sensitive to masker type than were CI users. Relative to SSN, NH listeners experienced significant masking release for most maskers, which could be well explained by the glimpse proportion, especially for maskers containing similar cues related to fundamental frequency or lexical content. In contrast, CI users generally experienced negative masking release. There was significant intercorrelation among the maskers for CI users' SRTs but much less so for NH listeners' SRTs. Principal component analysis showed that one factor explained 72% of the variance in CI users' SRTs but only 55% in NH listeners' SRTs across all maskers. Taken together, the results suggest that SRTs in SSN largely accounted for the variability in CI users' SRTs with dynamic maskers. Different from NH listeners, CI users appear to be more susceptible to energetic masking and do not experience a release from masking with dynamic envelopes or speech maskers.