In limulus sperm an actin filament bundle 55 mum in length extends from the acrosomal vacuole membrane through a canal in the nucleus and then coils in a regular fashion around the base of the nucleus. The bundle expands systematically from 15 filaments near the acrosomal vacuole to 85 filaments at the basal end. Thin sections of sperm fixed during stages in spermatid maturation reveal that the filament bundle begins to assemble on dense material attached to the acrosomal vacuole membrane. In micrographs fo these early stages in maturation, short bundles are seen extending posteriorly from the dense material. The significance is that these short, developing bundles have about 85 filaments, suggesting that the 85-filament end of the bundle is assembled first. By using filament bundles isolated and incubated in vitro with G actin from muscle, we can determine the end "preferred" for addition of actin monomers during polymerization. The end that would be associated with the acrosomal vacuole membrane, a membrane destined to be continuous with the plasma membrane, is preferred about 10 times over the other, thicker end. Decoration of the newly polymerized portions of the filament bundle with subfragment 1 of myosin reveals that the arrowheads point away from the acrosomal vacuole membrane, as is true of other actin filament bundles attached to membranes. From these observations we conclude that the bundle is nucleated from the dense material associated with the acrosomal vacuole and that monomers are added to the membrane-associated end. As monomers are added at the dense material, the thick first-made end of the filament bundle is pushed down through the nucleus where, upon reaching the base of the nucleus, it coils up. Tapering is brought about by the capping of the peripheral filaments in the bundle.