In Drosophila embryos, each motoneuron is accurately matched to one or more singly identifiable muscle fibers. In this article we altered the number and pattern of the embryonic muscle fibers using genetic, heat shock, and laser ablation methods to test whether motoneuron growth cones are able to recognize specific targets. The choices made by two motoneurons were assayed using both intracellular dye fills and immunocytochemistry. The motoneurons RP1 and RP3 have nearly identical central and peripheral axonal trajectories. However, RP3 innervates the two most ventral longitudinal muscle fibers, 7 and 6, while RP1 grows past these fibers to innervate only muscle fiber 13. In rhomboid mutants muscle fiber 7 does not develop. Despite the loss of one of its targets, RP3 faithfully innervated the remaining muscle fiber 6 in over 80% of the observed cases. Furthermore, neuron RP1 accurately innervated muscle fiber 13, although it traversed one fiber fewer to reach it. Laser ablation of muscle fiber 7 confirmed the target choices shown by the motoneurons. In numb mutants, multiple muscle fibers, including 7, 13, and 12, fail to develop. This allowed us to test whether fibers distal to the target are involved in muscle fiber recognition, possibly by halting the growth cone advance. In mutant embryos, RP3 innervated muscle fiber 6 at the same frequency regardless of the absence of the distal muscle fiber 13. By contrast, RP1, which had lost its target entirely, frequently failed to innervate any muscle fiber during the period examined. Finally, muscle fiber 13 can be duplicated in wild-type embryos by means of a brief heat pulse during myogenesis. Presented with two targets, RP1 innervated both fibers in each case examined, while RP3 synapsed with muscle fibers 7 and 6 normally. Neuron-specific antibodies revealed that the embryonic growth cone choices were not transient, but persisted into the larval neuromuscular projections. These results indicate that each motoneuron growth cone has a primary target preference, which is retained even when the numbers of the muscle fibers, and therefore their relative positions, are altered. We therefore suggest that synaptic recognition by Drosophila motoneuron growth cones relies on unique features of the individual muscle fibers.