Two experiments are reported which measure the increase in intelligibility occasioned by listening binaurally to running speech imbedded in interfering signals. In the first experiment, the frequency range of the speech is restricted to one of three ranges: 200–1600, 880–2200, 1660–6100. The interfering signal is broad‐band random noise. Under difficult listening (27% intelligibility for homophasic listening) the lowest range shows a binaural improvement of 33% for a change in interaural polarity, and of 28% for an interaural time disparity of 500 μsec. The middle‐ and high frequency ranges show less binaural gain, but do afford some advantage to two eared listening under difficult conditions. The second experiment compares the effects of different speech waves masking the wanted speech in binaural listening. For the maskers used, the binaural system helps most when the interfering wave is the speaker's own voice or a multiple mixture of many voices. The measured differences between homophasic and binaural listening are small but statistically reliable. The interaural differences employed (time delay and polarity reversal) are no help when the interfering wave is a female voice, another male voice, or the multiple mixture played backward.