The present study investigated how positive versus negative self-schematics on a friendly-unfriendly or intelligent-unintelligent dimension should process person information differently in impression formation tasks. Sixty female undergraduates served as subjects. They were asked to rate two targets on several trait scales based on a list of 12 behavioral descriptions. As a result, unfriendly-schematics rated the targets more favorably on friendliness-related scales than aschematics, while friendly-schematics rated them the same way as did aschematics. On the other hand, intelligent-schematics tended to evaluate the target's intelligent behaviors more positively compared with unintelligent-schematics and aschematics. The recall test showed that negative self-schematics on either of the two dimensions were more likely to cluster schema-related behaviors in memory. It was suggested that some motivational accounts other than self-esteem theory should be needed for interpreting the present results.