<p>Segregating target speech in a multi-talker background would require organizing speech fragments into a coherent stream of speech. The current investigation explored the intelligibility of temporally interrupted and checkerboard (interrupted in time and frequency) speech stimuli with one and two talkers. A series of experiments with one-talker stimuli revealed that the intelligibility of interrupted speech became lowest at around a 255-ms segment duration. In contrast, for two- and four-band checkerboard speech, the minimum intelligibility occurred at around 113–160-ms segment durations. Two-talker experiments revealed that no phonemic restoration was found, and that signal-to-noise ratios (SNRs) between target and competing talkers had no clear effect on the target intelligibility of interrupted (and staggered) speech and two- and four-band checkerboard speech. Still, SNRs dominated the intelligibility of two-talker checkerboard speech with eight or more frequency bands. The probability summation model may account for the intelligibility minimum in both interrupted and checkerboard speech with a single talker. The results suggest that the four speech cue channels explain the similarity of the intelligibility curves for two- and four-band checkerboard speech.</p>