- 1. In a tethered cockroach (Periplaneta americana) whose wings have been cut back to stumps, it is possible to elicit brief sequences of flight-like activity by puffing wind on the animal's body. 2. During such brief sequences, rhythmic bursts of action potentials coming from the thorax at the wingbeat frequency, descend the abdominal nerve cord to the last abdominal ganglion (A6). This descending rhythm is often accompanied by an ascending rhythm (Fig. 2). 3. Intracellular recording during flight-like activity from identified ascending giant interneurons, and from some unidentified descending axons in the abdominal nerve cord, shows that: (a) ventral giant interneurons (vGIs) remain silent (Fig. 3); (b) dorsal giant interneurons (dGIs) are activated at the onset of the flight-like activity and remain active sporadically throughout the flight sequence (Fig. 4); (c) some descending axons in the abdominal nerve cord show rhythmic activity phase locked to the flight rhythm (Fig. 5). 4. Also during such brief sequences, the cereal nerves, running from the cerci (paired, posterior, wind sensitive appendages) to the last abdominal ganglion, show rhythmic activity at the wingbeat frequency (Fig. 6). This includes activity of some motor axons controlling vibratory cereal movements and of some sensory axons. 5. More prolonged flight sequences were elicited in cockroaches whose wings were not cut and which flew in front of a wind tunnel (Fig. 1 B). 6. In these more prolonged flight sequences, the number of ascending spikes per burst was greater than that seen in the wingless preparation (Fig. 8; compare to Fig. 2). Recordings from both ventral and dorsal GIs show that: in spite of the ongoing wind from both the tunnel and the beating wings, which is far above threshold for the vGIs in a resting cockroach, the vGIs are entirely silent during flight. Moreover, the vGIs response to strong wind puffs that normally evoke maximal GI responses is reduced by a mean of 86% during flight (Fig. 9). The dGIs are active in a strong rhythm (Figs. 11 and 12). 7. Three sources appear to contribute to the ascending dGI rhythm (1) the axons carrying the rhythmic descending bursts; (2) the rhythmic sensory activity resulting from the cereal vibration; and (3) the sensory activity resulting from rhythmic wind gusts produced by the wingbeat and detected by the cerci. The contribution of each source has been tested alone while removing the other two (Figs. 13 and 14). Such experiments suggest that all 3 feedback loops are involved in rhythmically exciting the dGIs (Fig. 15).