Mental modeling ranges from pure categorization, for example, of linguistic concepts, to cognitive representation of complex decision tasks involving stochastic uncertainty and strategic interaction. In the tradition of consequentialistic bounded rationality, we assume to choose among choice alternatives by anticipating their likely implications. Such deliberation basically requires causal relationships linking own choices (means) and determinants beyond own control, such as chance events and choices by others (scenarios), to the relevant outcome variables (ends). We suggest a general framework of mental representation whose aspects are illustrated for stochastic choice and strategic interaction tasks. We also discuss how this framework can be experimentally implemented, showing how experimental research can shed light on mental modeling and—more generally—cognitive processes, in addition to eliciting the usual choice data.