A significant loss in attraction effect has never been obtained from subjects who are the direct recipients of another's evaluation. Rather, the effect has been obtained only from those subjects who assume the role of recipient of another's evaluation. In a discussion context, the present study compared direct recipients (DR) and role playing subjects (RP) in terms of their attraction toward a person who evaluated them in either a negative-negative (NN) or positive-negative (PN) sequence. Ten subjects participated in each condition of the 2 (DR or RP) x 2 (NN or PN) factorial design. A 2 x 2 analysis of variance on the attraction measure revealed a significant interaction between mode of receiving information (DR or RP) and sequence of information (NN or PN). Orthogonal comparisons revealed that DR subjects showed greater attraction toward PN confederate than the NN confederate, with the reverse being obtained for the RP subjects. The results showed that while RP subjects produced a loss of effect similar to that found by Clore, Wiggins, and Itkin (1975), DR subjects produced an information averaging effect similar to Tognoli and Keisner (1972). This difference was discussed in terms of "sender-oriented, receiver-oriented, and message-oriented explanations". (Author)