Abstract
Incremental theorists, who believe intelligence can improve, may be more resilient to failure than entity theorists, who believe intelligence is fixed. Three studies explored whether incremental theory reduces self-handicapping and self-esteem vulnerability in students who do and do not invest their self-worth in academics. In Studies 1 and 2, contingent incremental students self-handicapped by choosing to listen to performance-impairing music and by avoiding practice before a difficult task. In Study 3, contingent incremental students who could not self-handicap reported greater ability attributions and lower self-esteem following failure. These studies suggest that when self-worth is contingent on academics, incremental theorists remain concerned about their self-worth and self-handicap to protect their self-esteem from the ego-threat associated with failure.
Acknowledgments
This research was supported by National Science Foundation grant BCS 0446567. Yu Niiya was supported by a Fulbright graduate fellowship and the Institute for Social Research Innovation in Social Research Fellowship, Amara Brook was supported by funds provided by Santa Clara University, and Jennifer Crocker was supported by National Institute of Mental Health grant K02 MH01747.
We thank Katarzyna Bundyra, Fabian Castaneda, Julie A. Patterson, and Carmen Rivera for their assistance in data collection and Nobuyuki Futai for developing the computer program used in Studies 2 and 3.
Notes
1. Although statisticians often claim that median splits inflate Type I error rates (e.g., Bissonette, Ickes, Bernstein, & Knowles, Citation1990), Baumeister (Citation1990) found that median splits usually do not cause problems in real datasets because item variances and trait extremity are only moderately correlated at best. Regression analyses using contingency and incremental theory as continuous variables did not show a significant Contingency × Self-Theories × Task Difficulty interaction (β = .15, p = .63) but plotting the interactions showed the same pattern of interaction as with the median splits method.
2. In Studies 1 and 2, we selected the easy RAT items from the questions that more than 90% of participants in our previous studies (Brook, Citation2005; Niiya & Crocker, Citation2007) answered correctly and the difficult items from those for which 5% or fewer participants answered correctly.
3. For analyses of simple effects following interactions, we used the mean square within of the overall model instead of the mean square within of each subgroup because the former provides a more accurate estimate of the population error variance (Maxwell & Delaney, Citation2000).
4. Eighteen participants scored exactly on the median (5.60). The results remained identical when we included those participants in the low contingent group or deleted them from the analyses. For example, the Contingency × Self-Theories × Practice interaction for the internal attribution was F(1, 160) = 4.29, p < .05 when we included the median group in the low contingent group and F(1, 152) = 4.01, p < .05 when we deleted them from the analysis.