Axonal regression is utilized to refine neuronal circuits during development, but the dynamic properties of such regression remain largely unknown. We used confocal time-lapse imaging to examine the regression of single enhanced yellow fluorescent protein-labeled axons in corticospinal slice cocultures. By acquiring images at long (1 day) and short (30–60 min) intervals on days 5–13 in vitro, we detected three types of regressive events: 1) retractions, 2) amputations (referred to as autoaxotomy), and 3) degeneration. Retractions proceeded at some constant rate for up to 3 hours and then paused or switched to another constant rate, apparently shifting stepwise among three retraction rates (6, 12, 17 μm/hours). Autoaxotomy was a previously unreported strategy for regression. It occurred spontaneously, either at a proximal branch neck or at a distal end. Axons also underwent a form of degeneration that had several novel characteristics. Degenerating axons showed bright bead-like spots arranged at 3–9-μm intervals. The gaps were much larger than the spot size, and there was no prior sign of damage (e.g., swelling). Each spot's fluorescence intensity often waxed and waned, with its position unchanged. Degeneration progressed without clear proximal-to-distal directionality and was complete within 3–4 hours. Quantitative analysis of daily branch regression showed that branches almost always regressed up to their branch point or stopped before it, thereby keeping the branch point stable. This branch-point stability was retained for all three regression strategies observed, suggesting that the fate of each branch is determined relatively independently during the development of axonal arborization. J. Comp. Neurol. 513:164–172, 2009. © 2009 Wiley-Liss, Inc.