Because variation in fat reserves (i.e., condition) is expected to contribute to variation in survival and reproductive success, zoologists often wish to estimate the condition of the animals they study. The conventional condition estimates used for snakes are the residuals from a regression of body mass on body length. Because this estimate of condition is not independent of the variables used to estimate it (i.e., fat is a component of body mass), estimates derived in this fashion will be confounded whenever fat varies nonrandomly with length. To avoid this problem we used total lipid extraction to estimate percent body fat in a representative sample of northern water snakes (Nerodia sipedon). The conventional condition index explained less than half the variance in the measured percent body fat in this sample. An improved estimate of condition calculated as the difference between total body mass and predicted lean mass (based on the results of the lipid extraction) explained 70% of the variation in percent body fat in the original sample of snakes. This improved estimate also revealed that condition declined with body size in a large sample of male water snakes measured in early spring over a 4-year period. This last result, coupled with theoretical expectations that condition will not vary randomly with body size in snakes, suggests that researchers interested in condition should derive indices from direct measurements of body fat.