This study examined the effect of critical opinions on discourse production. In experiment I, forty-one university undergraduates were divided into three groups (control group, self-criticized group, and criticized group), and were asked to write their opinions. After they had received critical opinions, they were asked to write their opinions once again. Although the groups did not differ significantly in the opinions expressed in their essays, more subjects in the treatement groups significantly changed their main opinions than did subjects in control group. In experiment II, twenty-nine university undergraduates were tested as in experiment I except that, instead of second essays, they were asked to elaborate their original essays using a think-aloud method. Protocol analysis showed that the elaborations “as writer” and “as reader” appeared significantly more frequently than “others (e. g., correction of the erratum)”, and that the groups that received critical opinion had longer protocols (by one elaboration).