Thirty experimental and 30 control group subjects were selected. The Carkhuff empathy model was presented to experimental groups of young inmates for either 12, 24, or 36 hours of training conducted in consecutive 4-hour blocks. Eleven subjects received pretraining and posttraining evaluation. The primary dependent measures were scores from Carkhuff's Discrimination Index (D.I.) and Empathy Communication Index (C.I.). Independent raters assessed subject levels of empathy. Ancillary measures were scores from the Minnesota Multiphasic Personality Inventory Deviance Scale and offender prognosis for parole success-failure. The analyses of variance for the primary measures revealed that experimental subjects significantly increased their ability to discriminate levels of empathy, compared to control subjects. Duration of training was not a significant factor in any of the analyses of variance, nor were interactions with that factor significant. Supplementary analyses revealed no significant relationship between the ancillary measure scores and outcome of empathy training as measured by the D.I. and C.I. The supplementary analyses suggest that either the supplementary analyses suggest that either the supplementary measures, as expected, were not sufficiently valid, or strong treatment effects were responsible for the lack of significant correlation with empathy outcome instruments. Secondary measures used in the experiment need to be further evaluated and refined, since these measures indirectly determine the length of time served by offenders. Instruments used, the trainer manual, tabular data, and other supplementary materials are appended. Approximately 140 references are provided. (Author abstract modified)