Insects in free flight were filmed at 5000 frames per second to determine the motion of their wings and bodies. General comments are offered on flight behaviour and manoeuvrability. Changes in the tilt of the stroke plane with respect to the horizontal provides kinematic control of manoeuvres, analogous to the type of control used for helicopters. A projection analysis technique is described that solves for the orientation of the animal with respect to a camera-based coordinate system, giving full kinematic details for the longitudinal wing and body axes from single-view films. The technique can be applied to all types of flight where the wing motions are bilaterally symmetrical: forward, backward and hovering flight, as well as properly banked turns. An analysis of the errors of the technique is presented, and shows that the reconstructed angles for wing position should be accurate to within 1-2<latex>$^\circ$</latex> in general. Although measurement of the angles of attack was not possible, visual estimations are given. Only 11 film sequences show flight velocities and accelerations that are small enough for the flight to be considered as `hovering'. Two sequences are presented for a hover-fly using an inclined stroke plane, and nine sequences of hovering with a horizontal stroke plane by another hover-fly, two crane-flies, a drone-fly, a ladybird beetle, a honey bee, and two bumble bees. In general, oscillations in the body position from its mean motion are within measurement error, about 1-2% of the wing length. The amplitudes of oscillation for the body angle are only a few degrees, but the phase relation of this oscillation to the wingbeat cycle could be determined for a few sequences. The phase indicates that the pitching moments governing the oscillations result from the wing lift at the ends of the wingbeat, and not from the wing drag or inertial forces. The mean pitching moment of the wings, which determines the mean body angle, is controlled by shifting the centre of lift over the cycle by changing the mean positional angle of the flapping wings. Deviations of the wing tip path from the stroke plane are never large, and no consistent pattern could be found for the wing paths of different insects; indeed, variations in the path were even observed for individual insects. The wing motion is not greatly different from simple harmonic motion, but does show a general trend towards higher accelerations and decelerations at either end of the wingbeat, with constant velocities during the middle of half-strokes. Root mean square and cube root mean cube angular velocities are on average about 4 and 9% lower than simple harmonic motion. Angles of attack are nearly constant during the middle of half-strokes, typically 35<latex>$^\circ$</latex> at a position 70% along the wing length. The wing is twisted along its length, with angles of attack at the wing base some 10-20<latex>$^\circ$</latex> greater than at the tip. The wings rotate through about 110<latex>$^\circ$</latex> at either end of the wingbeat during 10-20% of the cycle period. The mean velocity of the wing edges during rotation is similar to the mean flapping velocity of the wing tip and greater than the flapping velocity for more proximal wing regions, which indicates that vortex shedding during rotation is comparable with that during flapping. The wings tend to rotate as a flat plate during the first half of rotation, which ends just before, or at, the end of the half-stroke. The hover-fly using an inclined stroke plane provides a notable exception to this general pattern: pronation is delayed and overlaps the beginning of the downstroke. The wing profile flexes along a more or less localized longitudinal axis during the second half of rotation, generating the `flip' profile postulated by Weis-Fogh for the hover-flies. This profile occurs to some extent for all of the insects, and is not exceptionally pronounced for the hover-fly. By the end of rotation the wings are nearly flat again, although a slight camber can sometimes be seen. Weis-Fogh showed that beneficial aerodynamic interference can result when the left and right wings come into contact during rotation at the end of the wingbeat. His `fling' mechanism creates the circulation required for wing lift on the subsequent half-stroke, and can be seen on my films of the Large Cabbage White butterfly, a plume moth, and the Mediterranean flour moth. However, their wings `peel' apart like two pieces of paper being separated, rather than fling open rigidly about the trailing edges. A `partial fling' was found for some insects, with the wings touching only along posterior wing areas. A `near fling' with the wings separated by a fraction of the chord was also observed for many insects. There is a continuous spectrum for the separation distance between the wings, in fact, and the separation can vary for a given insect during different manoeuvres. It is suggested that these variants on Weis-Fogh's fling mechanism also generate circulation for wing lift, although less effectively than a complete fling, and that changes in the separation distance may provide a fine control over the amount of lift produced.