The flow structure that is largely responsible for the good performance of insect wings has recently been identified as a leading-edge vortex. But because such vortices become detached from a wing in two-dimensional flow, an unknown mechanism must keep them attached to (three-dimensional) flapping wings. The current explanation, analogous to a mechanism operating on delta-wing aircraft, is that spanwise flow through a spiral vortex drains energy from the vortex core. We have tested this hypothesis by systematically mapping the flow generated by a dynamically scaled model insect while simultaneously measuring the resulting aerodynamic forces. Here we report that, at the Reynolds numbers matching the flows relevant for most insects, flapping wings do not generate a spiral vortex akin to that produced by delta-wing aircraft. We also find that limiting spanwise flow with fences and edge baffles does not cause detachment of the leading-edge vortex. The data support an alternative hypothesis-that downward flow induced by tip vortices limits the growth of the leading-edge vortex.