The effect of air viscosity on the flow around an insect wing increases as insect size decreases. For the smallest insects (wing length R below 1 mm), the viscous effect is so large that lift-generation mechanisms used by their larger counterparts become ineffective. How the weight-supporting vertical force is generated is unknown. To elucidate the aerodynamic mechanisms responsible, we measure the wing kinematics of the tiny wasp Encarsia formosa (0.6 mm R) in hovering or very slow ascending flight and compute and analyze the aerodynamic forces. We find that the insects perform two unusual wing-motions. One is "rowing": the wings move fast downward and backward, like stroking oars; the other is the previously discovered Weis-Fogh 'fling'. The rowing produces 70% of the required vertical force and the Weis-Fogh 'fling' the other 30%. The oaring wing mainly produces an approximately up-pointing drag, resulting in the vertical force. Because each oaring produces a starting flow, the drag is unsteady in nature and much greater than that in steady motion at the same velocities and angles of attack. Furthermore, our computation shows that if the tiny wasps employed the usual wing kinematics of the larger insects (flapping back and forth in a horizontal plane), vertical force produced would be only 1/3 of that by the real wing kinematics; i.e. they must use the special wing movements to overcome the problem of large viscous effects encountered by the commonly used flapping kinematics. We for the first time observe very small insects using drag to support their weight and explain how a net vertical force is generated when the drag principle is applied.Although the wing of an insect beats at high frequency (usually above 100 Hz), the velocity of the wing relative the undisturbed air is small, owing to the small wing-length. As a result, the vertical force coefficient of the wing required to balance the weight is relatively high; the mean vertical force coefficient required is around 2