Insects have evolved sophisticated flight control mechanisms permitting a remarkable range of manoeuvres. Here, I present a qualitative analysis of insect flight control from the perspective of flight mechanics, drawing upon both the neurophysiology and biomechanics literatures. The current literature does not permit a formal, quantitative analysis of flight control, because the aerodynamic force systems that biologists have measured have rarely been complete and the position of the centre of gravity has only been recorded in a few studies. Treating the two best-known insect orders (Diptera and Orthoptera) separately from other insects, I discuss the control mechanisms of different insects in detail. Recent experimental studies suggest that the helicopter model of flight control proposed for Drosophila spp. may be better thought of as a facultative strategy for flight control, rather than the fixed (albeit selected) constraint that it is usually interpreted to be. On the other hand, the so-called ‘constant-lift reaction’ of locusts appears not to be a reflex for maintaining constant lift at varying angles of attack, as is usually assumed, but rather a mechanism to restore the insect to pitch equilibrium following a disturbance. Differences in the kinematic control mechanisms used by the various insect orders are related to differences in the arrangement of the wings, the construction of the flight motor and the unsteady mechanisms of lift production that are used. Since the evolution of insect flight control is likely to have paralleled the evolutionary refinement of these unsteady aerodynamic mechanisms, taxonomic differences in the kinematics of control could provide an assay of the relative importance of different unsteady mechanisms. Although the control kinematics vary widely between orders, the number of degrees of freedom that different insects can control will always be limited by the number of independent control inputs that they use. Control of the moments about all three axes (as used by most conventional aircraft) has only been proven for larger flies and dragonflies, but is likely to be widespread in insects given the number of independent control inputs available to them. Unlike in conventional aircraft, however, insects' control inputs are likely to be highly non-orthogonal, and this will tend to complicate the neural processing required to separate the various motions.