Saito and Shimazaki (2012) found that people rely upon covariation information rather than temporal order information as cues to causal structure, whereas Lagnado and Sloman (2006) reported an opposite finding, indicating relatively greater influence of temporal order cues. The present research examines the hypothesis that such conflicting findings result from differences in task complexity. Specifically, it is proposed that covariation information becomes less influential as the number of variables increases. Experiment 1 investigated the relationship between the judgment strategy (i.e., covariation vs. temporal order) and the number of variables comprising a causal structure. As a result, people favored covariation cues primarily in tasks with simple causal structure. Experiment 2 used more complex causal structure. The results demonstrated that the tendency to emphasize covariation cues or to rely upon temporal order cues changes as a function of task complexity. These results were consistent with both previous findings and discussed in terms of causal Bayes net theories and heuristic models.