Cognitive self-instructional (CSI) programs have been successful in improving problem-solving skills in many, but not all, children. The importance of understanding the influence of subject characteristics in self-control studies, while often ignored in actual research, has been repeatedly advocated verbally. This paper presents a study designed to explore and confirm the relevance of subject characteristics such as language level, attribution of personal causality, and relationship with the therapist to CSI training outcome. In a treatment analogue study, 15 children received CSI training and 15 were in an attention control group. The children, aged 72 to 139 months, were from a non-clinical population attending a summer day camp. Each child participated in a subject measure assessment, pre- and post-training assessments, and two 20 minute training sessions. Training materials for both groups included tasks which varied in their similarity to the assessment measure. The CSI group improved more than the control group on Porteus Mazes; both groups improved over time on other measures. Cognitive maturity, personal attributions of causality, private speech, and therapist ratings were related for the CSI group to improvement on the Porteus Mazes but not the other measures. These same subject variables, especially the attribution measure, were related for the control group to all outcome measures. Prediction and detection of treatment versus practice effects are discussed. (Author/RH)