Between- and within-channel auditory selective attention were examined by presenting subjects with tone pips randomly to opposite ears; some pips had a slightly different pitch. Subjects were instructed to count rare, deviant tone pips at one ear and ignore all input to the other ear. Heart rate was sampled twice: once for the attended tone pips and once for the nonattended stimulus series. Heart rate responded differently to attended tone pips. While subjects were waiting for the rare stimulus to occur, heart rate slowed until the deviant stimulus was detected, which was followed by heart rate acceleration. Anticipatory heart rate deceleration was largely absent for nonatended series, and rare tone pips presented at the nonattended ear were not followed by acceleratory recovery. All tone pips elicited cardiac cycle time effects, that is, stimuli presented at short delays after the R wave prolonged the concurrent interbeat interval more than stimuli presented later. The cardiac cycle time effect was not altered by stimulus relevance (attended vs. nonattended) or significance (standard vs. rare). These results suggest that all stimuli receive preliminary perceptual analysis, but only attended stimuli are processed for further evaluation.