The best supported hypothesis for the evolutionary origin of insect wings is that they evolved from articulated, leg-derived respiratory structures of aquatic ancestors. However, there are no fossils of the immediate ancestors of winged insects, and it is difficult to imagine how a functional transition from gills to wings could have occurred. Recent studies of surface-skimming locomotion in stoneflies and mayflies offer a plausible solution by showing how rudimentary wings and muscle power can be used to accomplish two-dimensional aerodynamic locomotion on the surface of water. Here we extend that line of research by examining the phylogenetic distribution and mechanistic diversity of surface skimming in stoneflies, along with a limited examination of mayflies. These investigations reveal both a broad taxonomic occurrence and a fine gradation of mechanically distinct forms. Distinct forms of wing-flapping surface skimming include (1) stoneflies that flap their wings weakly while maintaining their body in contact with the water and undulating their abdomen laterally in a swimming-like motion, (2) stoneflies that skim while elevating their body above the water and maintaining all six legs on the surface, (3) stoneflies and mayflies that skim with only four legs on the water surface, (4) stoneflies that skim with only their two hind legs on the surface, and (5) stoneflies that, beginning with a series of leg motions nearly identical to hind-leg skimmers, use their hind legs to jump from the water into the air to initiate flapping flight. Comparisons across these forms of skimming show that wing-beat amplitude, horizontal velocity, and the verticality of aerodynamic force production increase as the body orientation becomes more upright and contact with the water is minimized. These behaviors illustrate a mechanical pathway by which flying insects could have evolved from swimming ancestors via a series of finely graded intermediate stages. The phylogenetic distribution of skimming and flight in stoneflies does not indicate any clear directionality toward either greater or lesser aerodynamic abilities; however, the broad and apparently basal phylogenetic distribution of skimming taxa supports the hypothesis that the common ancestor of stoneflies was a surface skimmer. This may also be true for the common ancestor of stoneflies and mayflies, that is, the first winged insects. We combine these data with fossil evidence to form a synthetic model for the evolution of flying insects from surface skimmers.