We sent a questionnaire survey to a random sample of 125 correspondents to the BMJ who had previously sent a letter which had been rejected. The objective was to evaluate the policy of sending on some unpublished letters to the authors of the articles to which they referred. There were 94 replies, a response rate of 75%. The key finding was that although most respondents agreed with the policy, a third thought it unconstructive. A quarter of the respondents said that (...) the BMJ policy would discourage them from sending a letter to the journal for publication. This survey has led to a change of policy at the BMJ. Letters which are not published are not now sent on to the authors of the original articles. (shrink)