Movements of six basking sharks (4.0-6.5 m total body length, L(T)) swimming at the surface were tracked and horizontal velocities determined. Sharks were tracked for between 1.8 and 55 min with between 4 and 21 mean speed determinations per shark track. The mean filter-feeding swimming speed was 0.85 m s(-1) (+/-0.05 S.E., n=49 determinations) compared to the non-feeding (cruising) mean speed of 1.08 m s(-1) (+/-0.03 S.E., n=21 determinations). Both absolute (m s(-1)) and specific (L s(-1)) swimming speeds during filter-feeding were significantly lower than when cruise swimming with the mouth closed, indicating basking sharks select speeds approximately 24% lower when engaged in filter-feeding. This reduction in speed during filter-feeding could be a behavioural response to avoid increased drag-induced energy costs associated with feeding at higher speeds. Non-feeding basking sharks (4 m L(T)) cruised at speeds close to, but slightly faster ( approximately 18%) than the optimum speed predicted by the Weihs (1977) [Weihs, D., 1977. Effects of size on the sustained swimming speeds of aquatic organisms. In: Pedley, T.J. (Ed.), Scale Effects in Animal Locomotion. Academic Press, London, pp. 333-338.] optimal cruising speed model. In contrast, filter-feeding basking sharks swam between 29 and 39% slower than the speed predicted by the Weihs and Webb (1983) [Weihs, D., Webb, P.W., 1983. Optimization of locomotion. In: Webb, P.W., Weihs, D. (Eds.), Fish Biomechanics. Praeger, New York, pp. 339-371.] optimal filter-feeding model. This significant under-estimation in observed feeding speed compared to model predictions was most likely accounted for by surface drag effects reducing optimum speeds of tracked sharks, together with inaccurate parameter estimates used in the general model to predict optimal speeds of basking sharks from body size extrapolations.