Testing stereotype threat theory predictions for math-identified and non-math-identified students by gender
This study followed on a previous study, (Cullen, Hardison, and Sackett (2004)), that questioned whether stereotype threat effects emerge in real-world testing contexts. That work showed that there was little support for the prediction that stereotype threat emerges only among those highly identified with a domain. Specifically, under the assumption that domain identification would be highest for those individuals at the upper end of the performance distribution, it was predicted that regression lines characterizing the test-performance relation would be nonlinear for minorities but linear for Whites. Using two large data sets collected in real-life situations, the relation was linear for both groups. That study, however, did not utilize a direct measure of domain identification. In this study, SAT Math scores were compared between women who expressed the intent to major in math or related fields, and women who were interested in other fields. It was predicted that standardized scores in math should underpredict performance in another domain (English) of women highly identified in math, compared with women not identified with the domain, again producing a nonlinear test-performance relation for the former group. This prediction was tested using a large dataset of White men and White women who completed the Math SAT test, and for whom English grades could be computed. For both men and women, higher Math SAT scores were negatively related to grades in English, but the strength of the relation was equivalent (and linear) for both groups. These data reinforce the call for caution in assuming that stereotype threat effects emerge in real high-stakes testing environments.