From Kepler data, we show that the incidence of flares on stars drops by only a factor of 4 from K-M dwarfs to A-F stars. Allowing for visibility effects, this implies that the true relative number of flare stars does not change very much from cool dwarfs to hot A stars. The idea that flares on A stars can be attributed to a cool companion has to be rejected because it leads to flare amplitudes two orders of magnitude smaller than actually observed. We confirm that spots on flare stars are generally larger than those on non-flare stars and that flare stars rotate significantly faster than non-flare stars. Analysis of 209 flare stars observed in Kepler short-cadence mode allows accurate measurements of flare shapes and duration. We find that about one-third of the flares have a bump or slope discontinuity on the decaying branch and that flares of long duration are to be found in stars with low surface gravities. Flare energies are strongly correlated with stellar luminosity and radius. The correlation with radius leads to a rough estimate of several tens of gauss for the typical magnetic field associated with a flare. The correlation with stellar luminosity can be understood if the typical flare loop length-scales approximately as the stellar radius. We examined the flare frequency as a function of orbital phase in three eclipsing binaries in which a large number of flares are visible. There appears to be no correlation of flaring with orbital phase, which weakens the hypothesis that flares in close binaries could be a result of reconnection of field lines connecting the two stars.