The purpose of this study was to evaluate the feasibility of a multi-source feedback system in a public sector organization and, more specifically, to examine performance appraisals provided by nontraditional rating sources (i.e., subordinates, peers, and sell) both psychometrically and on the basis of user-reactions. Intraclass correlations (i.e., inter-rater reliabilities) for subordinate evaluations were found to be relatively high (49). The pattern of correlations between ratings provided by different source pairs were similar to those reported in the literature (e.g., supervisor-peer = .31, self-peer = .30, subordinate-supervisor = .26, self- supervisor = .17). Results of a series of ANOVAs indicated that while self and peer ratings were most prone to leniency, supervisory ratings were relatively less prone to this bias. Self ratings were less open to halo than the ratings provided by the other sources. Along with an evaluation of the obtained user reactions, applicability of a multi-source performance appraisal system within the Turkish context is discussed.