OBJECTIVE The purpose of this study was to examine the ability of phallometry to discriminate among homicidal child molesters, nonhomicidal child molesters, and a comparison group of nonoffenders. METHODS Twenty-seven child molesters who had committed or had attempted a sexually motivated homicide, 189 nonhomicidal child molesters, and 47 nonoffenders were compared on demographic variables and psychometrically determined responses to aural descriptions of sexual vignettes. Two phallometric indexes were used: the pedophile index and the pedophile assault index. The pedophile index was computed by dividing the subject's highest response to an aural description of sex with a "consenting" child by his highest response to description of sex with a consenting adult. The pedophile assault index was computed by dividing the subject's highest response to an aural description of assault involving a child victim by his highest response to description of sex with a "consenting" child. RESULTS Homicidal child molesters, nonhomicidal child molesters, and nonoffenders were not significantly different in age or IQ. Homicidal and nonhomicidal child molesters had significantly higher pedophile index scores than nonoffenders. Significantly more homicidal child molesters (14 [52%] of 27) and nonhomicidal child molesters (82 [46%] of 180) than nonoffenders (13 [28%] of 47) had pedophile index scores equal to or greater than 1.0, but homicidal and nonhomicidal child molesters did not differ from each other. Significantly more homicidal child molesters (17 [63%] of 27) than either nonhomicidal child molesters (71 [40%] of 178) or nonoffenders (17 [36%] of 47) had pedophile assault index scores equal to or greater than 1.0, and nonhomicidal child molesters and nonoffenders were not significantly different from each other. Within-group analyses revealed that of the three groups, only the nonhomicidal child molesters exhibited a significant difference between their pedophile index scores and their pedophile assault index scores; their pedophile index scores were higher. CONCLUSIONS Consistent with past research, the authors found that the pedophile index is useful in differentiating homicidal and nonhomicidal child molesters from nonoffenders and that the pedophile assault index is able to differentiate homicidal child molesters from nonhomicidal child molesters and nonoffenders.