Functional morphologists commonly study feeding behavior in vertebrates by recording electrical activity from head muscles during unrestrained prey capture. Rarely are experiments designed to permit a partitioning of variation in muscle electrical activity patterns. Analysis of muscle activity during aquatic prey capture in two morphologivally distinct species of salamanders, Ambystoma dumerilii and A. mexicanum, is conducted to assess variation at four levels: between species, among individuals within species, among experiments conducted on different days, and among feedings. The results show that (1) mean correlations among the 11 electromyographic variables measured for each feeding are low and vary considerably among individuals, (2) many of the variables show significant differences among experimental days, (3) only one variable, the difference in timing between the depressor mandibulae and sternohyoideus muscles, showed significant variation between species, and (4) seven of the 11 variables showed significant variation among individuals within species. Overall, the variation between feedings (trials) was high, and there was some variation between days on which the experiments were conducted. Neither electrode position within the muscle nor satiation contributed to the high trial variance. The results suggest that functional analyses of feeding behavior should include an assessment of variation due to individuals, days, and trials, because the amount of variation at these levels may render differences between species nonsignificant.