We tested whether the spatial variation in resource depletion by Tundra Swans (Cygnus columbianus) foraging on belowground tubers of sage pondweed (Potnmogeton pectinatus) was caused by differences in net energy intake rates. The variation in giving up densities within the confines of one lake was nearly eightfold, the giving-up density being positively related to water depth and, to a lesser extent, the silt content of the sediment. The swans' preference (measured as cumulative foraging pressure) was negatively related to these variables. We adjusted a model developed for diving birds to predict changes in the time allocation of foraging swans with changes in power requirements and harvest rate. First, we compared the behavior of free-living swans foraging in shallow and deep water where they feed by head-dipping and up-ending, respectively. Up-ending swans had 1.3-2.1 times longer feeding times than,head-dipping swans. This was contrary to our expectation, since the model predicted a decrease in feeding time with an increase in feeding power. However up-ending swans also had 1.9 times longer trampling times than head-dipping swans. The model predicted a strong positive correlation between trampling time and feeding time, and the longer trampling times may thus have masked any effect of an increase in feeding power. Heart rate measurements showed that trampling was the most energetically costly part of foraging. However, because the feeding time and trampling time changed concurrently, the rate of energy expenditure was only slightly higher in deep water (1.03-1.06 times). This is a conservative estimate since it does not take into account that the feeding costs of up-ending are possibly higher than that of head-dipping. Second, we compared captive swans foraging on sandy and clayey sediments. We found that the harvest rate on clayey sediment was only 0.6 times that on sandy sediment and that the power requirements for foraging were 1.2-1.4 times greater. Our results are in qualitative agreement with the ...