Kinds of individuals defined by patterns of variables

This paper argues that investigators should consider replacing the popular practice of comparing individuals varying in gender, social class, and/or ethnicity on one or more continuous measures with a search for kinds of individuals defined by patterns of properties that include not only their values on outcome measures but also their gender, social class, and ethnicity. Investigators who believe that a particular predictor contributes to an outcome independent of the gender, class, or ethnicity of the participants often implement statistical procedures that promise to remove the contributions of the above categories. These analyses lead to misleading conclusions when the controlled category is correlated with the dependent measures. The final sections summarize the properties of genders, classes, and ethnic groups that make distinctive contributions to many psychological outcomes. The paper ends by noting that a society's ethical beliefs constitute a defensible basis for ignoring the biological properties associated with these categories in order to allow members of these groups access to whatever educational or occupational goals they desire.