When reasoning about events that occur in the world, two key types of judgments are counterfactual inferences—judging how the event might have turned out differently under different conditions---and causal selection—identifying which of multiple potential causes is responsible for an effect of interest. We conducted a novel experimental test that asked subjects to make both types of judgments about a realistic situation involving probabilistic causal relations and both proximal and distal causes. We found that a new model—the Exogenous Sampler (EXS) performed as well as the leading models of these judgment types but also specified the cognitive processes via which such judgments are computed. Yet, that all models failed to predict the full range of subjects' judgments points to the need of a new generation of models accounting for how people reason about events that arise from complex probabilistic causal structures.