A long-standing problem in the behavioural assessment of personality is the individual specificity of responses. Often, different persons externalize the same trait in different responses. One solution to this problem is to aggregate many different responses. The paper compares the power of response aggregation for predicting self-and other-ratings of personality with two alternative strategies of response selection: the nomothetic strategy of selecting the response with the highest overall predictive power, and the more idiographic strategy of selecting, for each individual, the most extreme response. Seventy subjects were videotaped in a sequence of social situations inducing shyness to various degrees. Five different nonverbal measures of shyness all correlated significantly with the subjects' self- or other-rated shyness, and showed low correlations across subjects and a substantial cross-situational consistency of response profiles. Response aggregation and both strategies of response selection were found to be equally powerful in predicting the subjects' self- and other-ratings of shyness from the five behavioural responses. Therefore, these findings somewhat dampen the hope often expressed in recent theoretical discussions of personality assessment that more respect for the individual case may improve nomothetic assessment procedures.