Journal of Personality and Social Psychology 1974, Vol. 30, No. 3, 378-381 EFFECT OF VALENCE ON RELATIVE WEIGHTING IN IMPRESSION FORMATION 1 BERT H. HODGES 2 Vanderbilt University This experiment assessed relative weighting of traits in three types of person- ality descriptions; those with only favorable traits (PP), with favorable and unfavorable traits (PN), and with only unfavorable traits (NN), Predictions that traits in PP descriptions would be weighted equally and that the more negative trait in PN and NN descriptions would receive greater weight were supported by the results. The results were generally consistent with the initial impression averaging model, although a possible difficulty with respect to the relative size of set-size effects for PP and NN descriptions was noted. Stimulus traits in impression formation may vary in polarization or in valence. Polarization refers to distance from the neu- tral point, while valence refers to evaluative sign, favorable or unfavorable, of the trait. This study explores the effect of valence on relative weighting in impression formation. Three valence combinations are possible: those in which all traits are favorable (PP); those in which some traits are favorable and some unfavorable (PN); and those in which all are unfavorable (NN). PP descriptions have generally shown equal weighting of traits, as well as set-size effects when traits have the same polarization (e.g., Anderson, 1965, 1967, 1968a). However, in PN descrip- tions, numerous studies indicate higher weight to unfavorable than favorable traits (e.g., Briscoe, Woodyard, & Shaw, 1967; Levin & Schmidt, 1969; Miller & Rowe, 1967; Wyer, 1969, 1970). In NN descriptions, moreover, more negative (more polarized) traits seem to receive more weight than less negative traits (e.g., Anderson, 1965, 1968a; Wyer, 1 This article is based on a master's thesis done under the direction of Donald L. Thistlethwaite and submitted to Vanderbilt University, 1970. The au- thor is deeply appreciative of Professor Thistle- thwaite's extensive help on earlier versions of this article. Also, the author is grateful to Norman Anderson for criticisms and editorial suggestions. Parts of this article were presented to the meeting of the Southeastern Psychological Association, May 1971. 2 Requests for reprints should be sent to Bert H. Hodges, Department of Psychology, Gordon College, Wenham, Massachusetts 01984. 1969). Set-size effects also occur in NN descriptions (e.g., Anderson, 1967). The purposes of this study were (a) to replicate in a single study the above findings concerning relative weighting in PP, PN, and NN descriptions; and (b) to ascertain the implications of these findings for algebraic models of impression formation. The most successful algebraic model has been the weighted averaging model proposed by Anderson (1965). The model assumes the psychological scale value of any given trait is unknown but is related to a behavioral evaluative rating as follows: + + Wo [1] where T\ is the rating of a particular trait, AI is the psychological scale value of the trait, 7 0 is an initial impression, and Wi and WQ are raw weights associated with AI and /o, respectively. For the two-trait descriptions used in this study, the model may be stated symbolically as: E = + Wi)7\ Wo + Wi + W2 [2] where E is the evaluation of the person de- scription. (The general case of Formula 2 is explained in Hodges, 1973.) The relative weight or importance of T\ is (w 0 + w\)/ (w 0 + w i+ w 2 ), and similarly for T 2 and 7 0 . These relative weights sum to 1 according to the model. Based on the results reviewed above, it was expected that Wi and w- 2 would be approxi- 378