ATTITUDES AND ATTRACTION : A NEW VIEW ON HOW TO DIAGNOSE THE MODERATING EFFECTS OF PERSONALITY

Abstract

It is now well-established that the greater the attitudinal similarity between two persons, the greater the attraction between them (Byrne, 1971 ). Surprisingly, evidence for the moderating role of personality variables in this similarity-attraction relationship has not been consistent. The present research, therefore, examined how the moderating effect of a personality variable should be studied. In past research, a statistically significant interaction between the personality variable (P) and the attitudinal conditions (C) at ? of .05 was the basis of accepting the moderating role of personality in the similarity-attraction link. Most of such tests were unfortunately plagued by their low statistical power due to a restricted range of similarity manipulations, small sample sizes, and the dichotomization of the continuous personality scores. Even when these methodological precautions are taken, the test of the P x C effect at ? of .05 cannot ensure high statistical power (Cohen, 1988, 1990). It is proposed, therefore, that research in this area should (a) use existing theory and literature to generate predictions of personality effects in specific conditions, and (b) rely on the replication of those predictions regardless of whether the overall P x C effect is significant or not. To test the proposal, the need for affiliation (nAff) as a personality variable was selected, and measures of both social (i.e., liking and enjoyment of company) and intellectual (i.e., intelligence and general knowledge) attraction toward a similar or dissimilar "bogus" stranger (Byrne, 1971) were taken from college students in Singapore. To improve power of the P x C test, eff01is were also made to increase (a) effect size by using extreme proportions of similarity (0 vs. 1 ), (b) n per cell of the P x C effect, and (c) ? to .20. In addition, multiple regression (MR) was used. Two independent experiments with Ns of 180 (Experiment 1) and 290 (Experiment 2) were conducted. To improve power further, the data of Experiment 1 were pooled together with those of the same-gender stranger condition of Experiment 2 (Experiment 3, N = 321 ). Moreover, the P x C effect was tested using three other conventional methods of median split, 3-category split, and extreme groups. The data were analyzed using the proposed theory-based approach which predicted the moderating effect in social attraction alone in the condition of similar attitudes. The basis for this prediction was that similar attitudes are weighted less than dissimilar ones in social attraction (Singh, & Ho, 1998; Singh, Teoh, & Loo, in press). Despite conscious efforts to improve the statistical power of the P x C test, no consistent results emerged from the four methods used. MR detected the effect at ? of .05 only when N was 321. In contrast, the proposed approach of a priori predictions and the replication of results confirmed the presence of the moderating effect of nAff in the similarity-social attraction link. In social attraction, regression analyses consistently detected the moderating effect of nAff in the condition of similar attitudes at ? of .05. More specifically, participants high in nAff were more socially attracted toward a similar stranger than those low in nAff. In intellectual attraction, the weighting of similar and dissimilar attitudes was nearly equal and hence there was no moderating effect of nAff. Estimates of the strength of the P x C effect, albeit small, confirmed the moderating effect of nAff on the similarity-social attraction relationship. These results not only confirm Chaplin's (1991) finding of a small moderating effect of personality but also suggest that power of the conventional methods of analysis cannot be satisfactory. Therefore, the use of a theory-based approach and the estimation of effect size can be more useful than past methodology in future work in this area.