In light of evidence that multiple‐goal messages tend to be characterized by longer speech‐onset latencies and higher pause‐phonation ratios, Greene, McDaniel, Buksa, and Ravizza have suggested that it is not the pursuit of multiple goals, per se, that results in less fluent speech. Rather, these authors suggest that the slower speech production characteristic of such messages is the result of difficulties in assembling, or integrating, incompatible message features. Recent evidence, however, indicates that this account is incomplete and in need of revision. Toward this end, we advance a complexity account that suggests that there are increased processing‐capacity and temporal demands associated with formulating and maintaining more complex message representations. The article then reports four experimental studies of this complexity account. Experiments 1 and 2 operationalize complexity as the amount of information to be communicated in a message. These studies indicate that complexity does affect pause‐phonation ratio and average pause duration. Experiments 3 and 4 operationalize complexity as the coherence of the information to be conveyed in a message. The results of Experiment 4, but not Experiment 3, again indicated that complexity influences speech fluency. The results of these studies, then, are taken to provide considerable support for the complexity h