How do people combine cues to form judgments? Recent debate has focused on whether and when individuals use heuristics versus linear models. We propose instead that people may rely on an understanding of the causal relationships between cues to determine how much weight to place on each one. Predictions of the causal model approach match those of linear models under certain circumstances and heuristic models under others, while making unique predictions in additional cases. In two experiments, we show that, as the causal relationships among cues changes, participant judgments consistently conform to predictions of the causal model approach while matching either heuristic or linear judgments in only a limited subset of cases.