Background Performance-based personality assessment has advantages, and the sentence completion method is one of the most reliable of these techniques. We sought to derive rating scales for three major personality traits (Extraversion, Neuroticism, and Psychoticism) based on Rotter Incomplete Sentences Blank (RISB) protocols. Participants and procedure Four successive samples of students (N = 231) completed the RISB and other measures. Rating scales were derived empirically and cross-validated against the Eysenck Personality Questionnaire (EPQ-R). Additional validity evidence was obtained regarding symptomatic distress (SCL-90-R scales) and creativity (based on originality scores and expert judgments of student-generated drawings and poetry). Results The scales could be applied reliably by undergraduate research assistants (mean individual intraclass correlation coefficients [ICCs] of .75, .79, and .71, respectively) and correlated well with self-reported traits (mean r values of .51, .57, and .58). Reliability of the finalized system was higher, with ICCs of .84, .83, and .79. RISB-rated traits correlated predictably with symptomatic distress. RISB-rated Psychoticism correlated strongly with the originality and judged creativity of drawings and poems. Psychoticism ratings were more strongly related to criterion measures than was the original self-report. Conclusions The sentence completion method can be used to rate personality traits reliably. These ratings correlate fairly strongly with the results of self-report inventories but sometimes correlate more strongly with theoretically relevant variables than do self-report scores.