Evaluators usually assume that the data they receive are valid, or at least, that the data reflect that which is being evaluated. In a recent evaluation in a large southern school district, evaluators compared their raw interview notes with interview notes edited by the teachers interviewed or project staff. The program evaluated was Project SEED, an additional mathematics class for students of low socioeconomic status. In the present study, 24 graduate students with no knowledge of the background compared 14 original and edited pairs of comments/notes in terms of whether the edited notes: (1) contained additional information; (2) were more positive in tone and intent; (3) were less concrete; and (4) changed meaning or content. Edited responses contained additional information and were more positive about the project, in the opinion of naive raters. Recommendations are made to control the collection of data so that the evaluation will be useful to decision makers. Allowing those who are being evaluated to control and possibly modify the data raises the danger of bias. The raw and edited responses, the survey form administered to the graduate student raters, and a table of the results are provided. (SLD)