Despite the use of visual habituation over the past half century, relatively little is known about its underlying processes. We analyzed heart rate (HR) taken simultaneous with looking during infant-controlled habituation sessions collected longitudinally at 4, 6, and 8 months of age with the goal of examining how HR and HR-defined phases of attention change across habituation. There were four major findings. First, the depth and topography of decelerations and proportion of sustained attention (SA) did not vary across habituation at any age, which suggested (in contrast to the tenets of comparator theory) the persistence of substantial cognitive activity at the end of visual habituation. Second, attention termination (AT) robustly declined across trials, suggesting that, contrary to prior thinking, AT might be a sensitive indicant of visual learning. Third, infants at all ages showed an HR increase (startle) to stimulus onset on the first trial, the magnitude of which was associated with subsequent delayed HR deceleration and less SA; thus, stimulus events affect processing during trials. Finally, mean overall HR reliably increased across trials for all ages. This last finding implies the need to distinguish between "phasic" HR changes (e.g., decelerations during looks) and longer term "tonic" HR changes (mean increases across trials) during habituation, and raises the question of what processes the tonic increases might reflect within the habituation paradigm. (Contains 4 tables and 3 figures.)