I examined the effect of a feedback system on the performance of two graduate teaching assistants (GTAb) in writing conferences with their students. Two sets of conferences were taped and coded for the interactions that took place and for the content transacted. I established inter-coder reliability of .84 on interactions and .87 and .89 on content categories, using Scott's Pi Coefficient, with two trained coders. The coded tapes were processed using the “Real Time observational Data Collection” microcomputer program. This program provided printouts of the frequencies and percentages of interactions and content in each of the GTAs’ conferences with students in this study. After the printouts were generated, I conducted feedback interviews following the first and second set of text-based conferences. Transcripts of these taped feedback interviews were also analyzed. I found that there were important differences between the first and second sets of conferences. One GTA was able, as a result of the feedback system, to alter his performance in his second set of conferences, becoming more student-centered, clarifying his expectations for students, and modeling appropriate responses to writing. The content of his second set of conferences did not change markedly; however, the content of both sets of conferences dealt with rhetorical issues as the literature recommended. While the second GTA’s performance did not change a great deal, her attitude became more consistent with the literature suggested. The content of her first conferences tended to focus on mechanical or sentence level concerns. Her second set of conferences, however, contained more discussion of rhetorical issues. Both GTAs attributed most of the changes in their conferences to the feedback system, particularly the printouts that quantified the interactions and content. Finally, the GTAs intentions for their first set of conferences matched their actual performance. When they saw their first printouts, however, tht,ey chanljed their intentions for their second set of conferences, becoming more consistent with what experts contend should be happening in writing conferences. / Doctor of Education