When inferring causal relationships, people are often faced with ambiguous evidence. Models of causal inference have taken different approaches to explain reasoning about such evidence. One approach – epitomized by Bayesian models of causal inference – defers judgment by representing uncertainty across multiple explanations. Another approach – usually adopted by associative models – approximates uncertainty by positing within-compound associations, a special type of association that forms between simultaneously presented cues. Although these approaches explain many of the same experimental findings, we note some limitations of the latter approach. In two experiments, we tested the predictions of these approaches. The results were consistent with models that represent uncertainty across multiple explanations and inconsistent with models that use within-compound associations.