We propose that heterogeneous asset trading behavior is the result of two distinct, non-convertible mental dimensions: analytical (“quantitative”) capability and mentalizing (“perspective-taking”) capability. We develop a framework of mental capabilities that yields testable predictions about individual trading behavior, revenue distribution and aggregate outcomes. The two-dimensional structure of mental capabilities predicts the existence of four mental types with distinguishable trading patterns and revenues. Individuals will trade most successfully if and only if they have both capabilities. On the other hand, subjects who can mentalize well but have poor analytical capability will suffer the largest losses. As a consequence, being able in just one dimension does not assure trading success. We test these implications in a laboratory environment, where we first independently elicit subjects’ capabilities in both dimensions and then conduct a standard asset market experiment. We find that individual trading gains and patterns are consistent with our theoretical predictions. Our results suggest that two mental dimensions are necessary to encompass the complex heterogeneous behaviors in asset markets; a one-dimensional measure of mental capability will lead to biased conclusions. The findings have potential implications for financial institutions, which can use the measures to select successful traders, or for policy-makers, helping them to prevent the formation of asset bubbles. Finally, our conceptual framework and the empirical screening method could be applied to explain heterogeneous behavior in other games.