To gain insight into how temperature affects locomotory performance, we measured power output of scup red muscle during oscillatory length changes at 10 degrees C and 20 degrees C. When we optimized work loop parameters, we found that maximum power was 27.9 W kg-1 at 20 degrees C and 12.8 W kg-1 at 10 degrees C, giving a Q10 of 2.29. Maximum power was generated at a higher oscillation frequency at 20 degrees C (5 Hz) than at 10 degrees C (2.5 Hz), and the Q10 for power output at a given oscillation frequency ranged from about 1 at 1 Hz to about 5 at 7.5 Hz. An analysis of the results in terms of crossbridge kinetics suggests that the higher power output at 20 degrees C is associated with both a higher Vmax (i.e. more power per cycling crossbridge) and faster activation and relaxation (i.e. a greater number of cycling crossbridges). To obtain a more realistic understanding of the functional importance of temperature effects on muscle, we imposed on isolated muscle in vivo length changes and oscillation frequencies (measured during previous experiments on swimming scup) and the in vivo stimulus duty cycle (measured from electromyograms in this study). At 20 degrees C, muscle bundles tested under these in vivo conditions produced nearly maximal power, suggesting that the muscle works optimally during locomotion and, thus, important contractile parameters have been adjusted to produce maximum mechanical power at the oscillation frequencies and length changes needed during swimming. At 10 degrees C, muscle bundles tested under in vivo conditions produced much less power than was obtained during the 'optimized' work loop experiments discussed above. Furthermore, although 'optimized' work loop experiments showed that maximum power output occurs at 2.5 Hz, scup do not appear to swim with such a low tailbeat frequency. Although the reason for these apparent discrepancies at 10 degrees C are not known, it shows that simple extrapolation from isolated muscle to the whole animal, without knowledge of how the muscle is actually used in vivo, can be misleading.