Both offender and victim data were used. The offender data consisted of written accounts of rapes transcribed from interviews with 41 imprisoned serial rapists. The victim data were derived from rape cases submitted for analysis by local law enforcement agencies. Each rape in both data sets were coded according to 33 verbal, sexual, and physical scales as well as an overall classification of rape type. The analyses suggest that behavioral scales can be used to classify rapes, according to both a fourfold typology of rape and the offender's propensity for increasing violence in a succession of rapes. The analysis produced a subset of scales classifying rapes according to the type of rape with an overall accuracy of 83 percent in the offender data and 89 percent in the victim data. Also, a subset of scales was produced classifying rapists as to their increaser/nonincreaser status with an overall accuracy of 92 percent in the offender data and 89 percent in the victim data. 2 tables and 14 references (Author abstract modified)