The structure of executive function (EF), as it pertains to distinct "hot" (affectively salient) and "cool" (affectively neutral) dimensions, in early childhood is not well understood. Given that the neural circuitry underlying EF may become increasingly differentiated with development and enriched experiences, EF may become more dissociable into hot and cool factors with age and advantaged socioeconomic circumstances. We used confirmatory factor analysis (CFA) to compare a multidimensional hot and cool EF model with a unidimensional model in early childhood, and to investigate model invariance across age and maternal education. Participants were 1900 children (2–5 years of age) from socioeconomically diverse families in an urban area in the southern United States. We aggregated data from four previously collected studies that included EF tasks, thus this study includes secondary data analysis. We tested model fit across (1) children older and younger than 4 years of age and (2) higher (college experience) versus lower (no college) maternal education. Results indicated that a two-factor hot and cool EF model provided the best fit to the data across all groups. Although the number of factors was invariant, only partial metric invariance was met for age, suggesting that how certain tests represent EF changes with age. For maternal education, partial scalar invariance was met, with higher maternal education associated with higher scores on certain EF tasks. Findings with this large sample suggest that EF includes two factors characterized as hot and cool. However, the study raises questions about model invariance, particularly across age.