Many studies of causal judgments have dealt with the relation between the presence and the absence of a cause and an effect. However, little is known about causal learning with a continuous outcome. The present study adopted Cohen’s d as an objective standard for effect size in situations where a binary cause influenced a continuous effect and investigated how people use means and standard deviations in the estimation of effect sizes. The experimental task was to read a scenario where the performance of two groups was compared and to infer the causal effect. Whereas means were manipulated while holding standard deviations constant in the mean difference group, standard deviations were varied with holding means constant in the standard deviation difference group. The results demonstrate that participants could respond appropriately to the difference in two means, and that they gave a higher estimate of effect size in large standard deviation situations than in small standard deviation situations. Judgments about standard deviations are in contrast to Cohen’s d, indicating disproportionate attention to different kinds of data samples.

Previous studies of causal learning heavily focused on binary outcomes; little is known about causal learning with continuous outcomes. The present paper proposes a qualitative extension of the causal power theory to the situation where a binary cause influences a continuous effect, and induces causal power under various ceiling situations with the continuous outcomes. To test the predictions, we systematically manipulated the type of outcome (continuous vs. percentage vs. binary) and the contingency information. The experiment shows that people estimate causal strength based on the linear-sum rule for continuous outcomes and the noisy-OR rule for binary outcomes. In the partial ceiling situation where causal power is partially inferred but not precisely estimated, the distribution of participants’ judgments was bimodal with one mode at the minimum value and the other at the maximum value, suggesting some participants made conservative estimates while others made optimistic estimates. These results are generally consistent with the predictions of the causal power theory. Theoretical implications and future directions are discussed.