We study the pattern of correlations across a large number of behavioral regularities, with the goal of creating an empirical basis for more comprehensive theories of decision making. We elicit 21 behaviors using an incentivized survey on a representative sample (n = 1,000) of the U.S. population. Our data show a clear pattern of high and low correlations, with important implications for theoretical representations of social and risk preferences. Using principal components analysis, we reduce the 21 variables to six components corresponding to clear clusters of correlations. We examine the relationship between these components, cognitive ability, demographics, and qualitative self-reports of preferences.