The question of whether lay attributors are biased in their discounting of 1 cause given an alternative cause has not been resolved by decades of research, largely due to the lack of a clear standard for the rational amount of discounting. The authors propose a normative model in which the attributor's causal schemas and discounting inferences are represented in terms of subjective probability. The analysis examines Kelley's (1972b) proposed causal schemas and then other schemas for multiple causes (varying in assumptions about prior probability, sufficiency, correlation, and number of causes) to determine when discounting is rational. It reveals that discounting is implied from most, but not all, possible causal schemas, albeit at varying amounts. Hence, certain patterns of discounting previously interpreted as biases may, in fact, reflect coherent inferences from causal schemas. Results of 2 studies, which measured causal assumptions and inferences, support this interpretation.