Under the assumption that similar findings by independent investigators imply a useful level of association between a feature in a drawing and a behavioral or personality trait, the human figure drawings of 200 aggressive or withdrawn male delinquents were evaluated and compared with both undifferentiated delinquents and nondelinquents with the expectation that the former groups would show a higher incidence of those features which have been associated with these respective traits. It was also hypothesized that using the features collectively would increase the accuracy of prediction over using individual characteristics. Significant differences were found for 6 of the 16 features, and predictive power was substantially increased by using all the features as indicators of their respective traits. These drawing characteristics have diagnostic value, but their relatively infrequent occurrence and the presence of most features in the drawings of all four groups require cautious clinical assessment of drawings exhibiting these features.