Typing according to the interview/voice analysis technique failed to discriminate assaultive from nonassaultive inmates. However, multivariate analysis indicates that assaulters were more likely to be Type-A as measured by the questionnaire, to be serving shorter sentences, and to have had a high number of previous convictions. A discriminant function using these three variables was able to differentiate assaulters from nonassaulters with 88-percent accuracy and could provide an instrument for identifying violence-prone inmates. 4 tables and 32 references (Author abstract modified)