A figure-ground model of cognitive balance

Beider's balance model was recast through triad-specific application of the Gestalt figure-ground distinction. The resulting model effectively identifies the cognitive biases, reciprocity, positivity and symmetry (or agreement) as balance "levels". The most positive sentiment relationship in a configuration was designated the "reference relationship" and treated theoretically as the basic ground against which the remaining system was evaluated. The Figure-Ground (FG) model holds that imbalance in p-o-x or p-o-q triads may be conceptualized as discrepancy in identified conceptual sets which may occur at any of the three levels. The model allows for degrees of tension (or imbalance), within the figure-ground substructure (measured by the "sumdis" variable), but postulates that the single most discrepant relation ("maxdis"), regardless of level, is the major source of cognitive imbalance. It is assumed that Attention is directed toward unbalanced aspects of the social field and that these relationships are linked to balanced aspects of the field in much the same way figure is related to ground in visual perception. In study I, twenty-four three-person groups of college women (N = 72) were instructed to role-play positive and negative sentiments. Role-play instructions were coordinated so that representative p-o-q configurations could be tested. Subjects gave ratings on pleasantness (PL) and tension (T), and changes In ratings (Dn) were also recorded. Previous multivariate analysis indicated that positivity — the tendency for all relationships to become more positive -— was the single most significant variable in accounting for results. Re-analysis indicated positivity was more powerful than the FG variables. In study II, 41 city firemen rated 422 p-o-x, 123 P-o-q and 41 four-entity configurations for PL, T and consistency (C). Entities in the structures were fellow firemen who had shared a previous watch with S. Data from studies I and II were analyzed using a multiple regression approach. Generalizing from both studies, results indicated: (a) across dependent variables, the sumdis index was more powerful than the maxdis index, suggesting that the less-than-maximum discrepancies in a structure may influence experienced imbalance; (b) the positivity variable was more powerful than maxdis for p-o-q groups, although the latter did account for significant variance; (c) the maxdis index was differentially sensitive to the dependent variables as a function of group type. The index account ed for significant PL variance in p-o-x groups (F = 125.67, df = 1/82, p < .01), but did not do so for p-o-q groups (F = 2.80, df = 1/82, N.S.). In p-o-q groups, maxdis was significantly correlated with C (F = 11.14, df = 1/82, p < .01), Findings suggested p-o-x and p-o-q configurations contained both cognitive and affective components, the p-o-q groups being more sensitive to the affective component. The maxdis index was primarily sensitive to cognitive balance while positivity was most associated with affective dynamics. Among dependent variables, PL reflected cognitive balance in p-o-x groups, but tended to be more sensitive to affective dynamics in p-o-q groups. In the latter configurations, C appeared to measure cognitive tension. The T variable was sensitive to both affective and cognitive dynamics, but did not effectively differentiate the two. In the four-man groups, positivity was the most powerful variable, but maxdis (based on the single most discrepant relationship) accounted for significant variance. Contrasted with the p-o-q groups, the index evidenced a stronger correlation with PL (r = .48, p < .01), while the correlations with C and T remained fairly constant. The data pattern suggested that group size may have little effect on cognitive dynamics once it exceeds two; however, the cognitive component may begin to exert more influence on PL as group size continues to increase. Directions for future research were discussed.Psychology, Department o