Detectability of binaurally presented 400- and 800-Hz tonal signals was investigated in an adaptive, two-interval forced-choice experiment. A continuous 3150-Hz low-pass noise masker was presented either diotically (No), interaurally uncorrelated (NU), or interaurally phase-reversed (N pi), at an overall level of 70 dB SPL. Signal duration was either 100 or 1000 ms. The interaural phase difference (IAPD) of the signal was either fixed (0 degree-180 degrees) or time-varying (slightly different frequencies were presented to the two ears). The range of interaural phase variations was selected to yield the same varying interaural temporal differences that would be produced if real auditory targets moved through various arcs in the horizontal plane. In no case was a signal with varying IAPD any more (or less) detectable than would be expected from averaging subjects' performance in the corresponding fixed-IAPD conditions through which the variation occurred. However, in detecting these signals, subjects placed relatively more weight on the temporal central portion than on either the onset or offset. It is proposed that this weighting effect is based on two factors: (1) the signal's 20-ms rise-decay time (i.e., the onset and offset receive less binaural weight because of monaural attenuation); and (2) the very low-pass filtering effected by the binaural system, which results in some minimum time required for it to become "fully engaged." Another finding was that signal detectability became gradually worse as the antiphasic moment in a varying-IAPD signal was moved from the temporal midpoint toward the onset. No evidence was found that a signal's onset and offset were weighted differently in a binaural signal detection task.