Self-esteem, self-certainty and desirable responding: Cognitive load does not always increase extremity

With an eye toward couples\u27 cognitions, five studies were conducted to study effects of cognitive load on endorsements of trait adjectives as self-descriptive. Following Paulhus, Graf and Van Selst (1989, hereafter PGV), subjects make yes-no decisions about undesirable, neutral and desirable trait adjectives under high or low load. PGV found extremity responding under high load: desirable traits were endorsed more often and undesirable traits less often. The present studies investigated this phenomenon considering social desirability a continuous variable and using logistic regression rather than ANOVA. Study 1 (N = 14) duplicated the PGV procedures, but found no effects for nomothetic (Anderson 1968) ratings; idiographic ideal-self ratings yielded a reliable effect in the opposite direction to predictions. Study 2 (N = 58) used time pressure to manipulate load and found no effects. Study 3 (N = 55) returned to the PGV a computerized procedure, but again found no reliable effect. When individual differences in self-esteem (mean desirability of traits one has or real-ideal discrepancy) were considered, low-self-esteem subjects were found to show extremity effects with respect to idiographic actual-self ratings, while high-self-esteem subjects showed anti-extremity effects. Study 4 (N = 38) used an 8-digits-in-memory load manipulation and replicated the finding that those low in desirability self-esteem showed extremity effects, while those high in desirability self-esteem showed anti-extremity effects, regardless of whether nomothetic or idiographic predictors were used. Subjects were found to have mean Rosenberg Self-Esteem above that reported on samples from the PGV population, although Rosenberg Self-Esteem did not reliably moderate the effect. Self-Clarity (Campbell 1990) also mediated an extremity effect for actual-self ratings. Study 5 assessed self-certainty on visual analog scales (N = 110; cf Baumgardner 1990), in addition to Study-4 variables (N = 60). An actual-self anti-extremity effect and a weak ideal-self extremity effect, but no individual-differences, were found. Possible effects of the preceding questionnaire and sex differences are considered. Future research is outlined, emphasizing types of cognitive load and individual difference variables. Extensions to the study of marital cognitions are discussed