Fig.6. Sample instantaneous forces for various combinations of flip start ₀ and flip duration . In all kinematic patterns, stroke amplitude was 180° and angle of attack was 45°. The format for each panel is that described for Fig.3. As in Fig.3, the radial forces for all these kinematics are zero and have not been plotted. (A) Forces generated with a slow flip (=0.5), symmetrical with respect to stroke reversal (₀=-0.25, flip timing _f=0). Under these conditions, the quasi-steady model (broken lines) accurately predicts measured lift, but not drag. (B) Forces generated with moderate flip duration (=0.25), advanced with respect to stroke reversal (₀=-0.25, _f=-0.125). With these kinematics, the augmentation of lift by rotational circulation and wake capture is evident. (C) Forces generated with a long, advanced flip (=0.5; ₀=-0.5, _f=-0.25). This pattern of kinematics produced elevated drag due to wake capture at the start of each stroke. (D) Same kinematics as in C, but with a delayed flip (=0.5; ₀=0, _f=+0.25). The delay in flip timing causes a small decrease in mean drag, but an enormous decrease in lift. (E–H) The influence of rotational timing on a short-duration flip. (E) Forces generated by a short flip advanced by almost a full half-cycle with respect to stroke reversal (=0.1; ₀=-0.5, _f=-0.45). Note that the angle of attack is negative during most of translation because the wing flips much too soon. As a consequence, the pattern generates negative lift. (F) Forces generated by a slightly advanced short flip (=0.1; ₀=-0.1, _f=-0.05). This near-optimal pattern augments lift by both rotational mechanisms. (G) Forces generated by a short symmetrical flip (=0.1; ₀=-0.05, _f=0). (H) Forces generated by a slightly delayed short flip (=0.1; ₀=0, _f=0.05). The small delay of 0.05 decreases the mean lift coefficient by 20% compared with the symmetrical case shown in G. , mean drag coefficient; , mean lift coefficient.