A growing body of literature suggests that experts are little if at all better than novices in terms of the quality of decision outputs. To explain this counter-intuitive finding, the authors propose a conceptual framework that focuses on initial problem structure as a key moderator of the effect of expertise on performance. Specifically, they argue that the expert-novice performance differential should be greatest at moderate levels of problem structure and weakest at both extremes. To examine this central hypothesis, the authors conduct a controlled experiment that compares experts with novices when solving a complex problem that had characteristics of a moderately ill-structured problem. Relative to novices, the authors find that experts select fewer, but more diagnostic, information inputs and are more consistent when evaluating nonquantified inputs. As a result, they make more accurate and tightly clustered judgments than do novices, and also are more confident in their decisions. To examine the moderating influence of problem characteristics, certain task variables are manipulated to increase or decrease initial problem structure. As hypothesized, the benefits of expertise are less pronounced when solving a problem with increased initial structure.