A study was conducted to explore the generality of reconstructive processes of memory for social information. Sixty college students divided evenly into six experimental and control groups were asked to read one of two versions of a story: one in which a couple happily agreed not to have children, the other in which the man's desire to remain childless greatly upset the woman. Those in the experimental groups were then asked to write their impressions of the story. Subjects who had read about the disagreeing couple were then told they had gotten married, and those who read about the agreeing couple were told they had split up. Those in the control groups either received the biasing information before writing their impressions or did not write any impressions. At the second session, either 2 days or 2 weeks later, subjects were asked to recall the story they had read during the first session and to avoid adding their own impressions or thoughts to the account. The primary dependent measure was the number of reorganizing errors in the subjects' recall. Subjects who were tested after 2 days were more accurate in their overall recall than those tested after 2 weeks. There were no other differences in overall recall between groups. Contrary to prediction, those who learned the marriage outcome before writing their impressions did not commit a greater number of reorganizing errors than those who wrote no impression. They did, however, commit more errors than those who wrote an impression before receiving the biasing information, although the effect was only marginal. (HTH)