The power expenditure of flapping flight in birds is characterised by a U-shaped function of speed through the air. From this relationship and the assumption of limited power available from flight muscles, it is possible to predict changes in the birds' airspeed in relation to external factors such as wind. These predictions are derived from flight mechanical theory and optimality criteria concerning migration or transport flight economy. Using tracking radar we measured flight speeds of migrating birds at 12 sites along the Northwest Passage in arctic Canada. We analysed variation in airspeed (V-a) in relation to the wind effect (V-g-V-a, where V-g is the groundspeed), vertical speed (V-z), altitude (z) and the compensation for the amount of side wind (1/cosalpha, where cc is the angle between track and heading). We found significant effects on the variation in V-a for all four variables, revealed by multiple regression analysis, but the total variation explained was relatively small suggesting that other factors might be involved. The signs of the regression coefficients were as predicted, except for the effect of side wind where we found a negative relationship between V-a and 1/cosalpha, possibly because our sample included an unknown mixture of bird species. We also compiled information from the literature from studies reporting analyses of the effects of the four variables on V-a. Adjustment of V-a in relation to the wind effect seems nearly omnipresent among birds, while the effects of vertical speed and altitude have been reported surprisingly few times. An increased V-a with increasing alpha (and 1/cosalpha) has not been found yet, perhaps due to the lack of critical observation conditions.