We studied flying zebra finch (Taeniopygia guttata, N = 12), to provide a new test of a long-standing ;fixed-gear' hypothesis that flap-bounding birds use only intermittent non-flapping phases, instead of variation in muscle activity, to vary mechanical power output in flight. Using sonomicrometry and electromyography, we measured in vivo fascicle length and neuromuscular recruitment in the pectoralis as the birds flew in different flight modes (level, ascending, descending; mean velocity 1.6+/-0.3 m s(-1)) and across velocities in a new, variable-speed wind tunnel (0-12 m s(-1)). Synchronized high-speed digital video (250 Hz) provided a record of wing kinematics. Flight mode had a significant effect upon pectoralis strain, strain rate, fractional shortening and the relative timing of muscle activity (onset, offset and duration). Among flight velocities, we observed significant variation in pectoralis strain, fractional lengthening and shortening, strain rate, relative electromyographic (EMG) amplitude, and EMG duration and offset. In particular, variation in strain rate and relative EMG amplitude indicates that the fixed-gear hypothesis should be rejected. Instead, it appears that zebra finch vary work and power output within wingbeats by modulating muscle contractile behavior and between wingbeats using intermittent bounds. Muscle activity patterns and wing kinematics were similar between free flight and wind tunnel flight at similar speeds. Comparing flights with and without surgically implanted transducers and electrodes, zebra finch exhibited a reduction in maximum velocity (from 14 to 12 m s(-1)) and a significant increase in wingbeat frequency and percent time flapping. This identifies a potential limitation of in vivo flight measurements, and similar studies of bird flight should, therefore, include measurements of the extent to which flight performance is compromised by experimental protocol.