Although the analysis did not find a significant number of demographic and psychological variables that distinguished men who completed treatment from those who dropped out, a logistic regression model was developed that correctly predicted treatment completion for 60 percent of the sample of 114 men. This was an 8.9-percent improvement over chance in the prediction of treatment completion for the sample. When treatment completion was modeled to be dependent on the six factors of the model, three of the predictive variables were significant: the man's use of sexual coercion with his partner in the last 12 months; the frequency of his use of sexual coercion; and having injured his partner in the last 12 months. The reduction of drop-out rates for batterer treatment programs may be possible if such programs can reliably predict treatment completion and dropping out. Program efforts could then provide treatment components appropriate for each of the two groups. The 114 men in the sample (56 treatment completers and 58 drop-outs) had been participants in the Alternatives to Violence Program, which is operated by a nonprofit counseling organization in Florence, SC. The batterer treatment program uses cognitive-behavioral treatment methods consistent in organization and focus for programs recommended in the literature. The instruments used in developing the model were the Revised Conflict Tactics Scales, which is a self-report measure of physical assaults in domestic relationships, and the CAGE questionnaire, which was used to identify excessive alcohol consumption. 4 tables and 46 references