Three experiments investigated the role of binaural cues in concurrent vowel segregation, and their relation to harmonicity cues. Subjects identified pairs of vowels that differed in amplitude (0 and 15 dB), fundamental frequency (\fo) (0 and 6\%), harmonic state (harmonic or inharmonic), and binaural state [identical at both ears, delay of 450 \mus\ in one ear, reversed polarity in one ear ($\pi$ phase), and random interaural phase]. In Experiment 1, both vowels were harmonic. For equal amplitudes and an \fo\ difference (\deltafo) of 6\% , identification was better with an interaural time delay (ITD) of 450 \mus\ than 0 \mus\, as reported by previous authors. At \deltafo=0\%, the effect was small. When vowel amplitudes differed by 15dB, identification of the weaker vowel was best when the {\em competing}\ vowel was identical at both ears, unless the target was also identical at both ears, in which case identification was relatively poor. Contrary to what was observed at 0dB, binaural cues were also effective at \deltafo=0\%. In Experiment 2, both vowels were inharmonic. Effects were similar to those observed in Exp. 1 at \deltafo=6\%: identification was best when the competing vowel was identical at both ears, and did not depend much on the binaural state of the target vowel. In Experiment 3, the harmonic state (harmonic, inharmonic) and binaural state (identical at both ears, random interaural phase) of both vowels were crossed. The effect of binaural state was similar to that of harmonic state, but smaller. In general, effects of binaural state were smaller than those of harmonicity and \deltafo. Overall, results are consistent with a binaural cancellation model of sound segregation.