Every branch of the military has called for the need to embrace personalized, adaptive learning methods tailored to each learner’s unique profile of strengths and weaknesses (Department of the Army, 2011; Roberson & Stafford, 2017; U.S. Fleet Forces Command, 2017). While the effectiveness of adaptive learning is well-established (VanLehn, 2011), there may be boundary conditions under which adaptive learning methods are most effective. Therefore, the purpose of this study was to examine such conditions. Participants (N = 76) recruited from Amazon Mechanical Turk engaged in a mathematical task to train and assess their skills in order of operations. A pre-test administered to all participants was used to classify each participant at a beginner, intermediate or advanced level. The pre-test was followed by multiple problem sets of varying difficulty levels, and then a post-test. To progress through the problem sets, participants were randomly assigned to a control learning condition (n = 49) or an adaptive learning condition (n = 27). For the former, participants progressed linearly through a set of math problems. For those participants in the adaptive learning condition, the difficulty of each problem set was determined based on their prior performance. The criterion measure of learning was operationalized as the gain between the participants’ pre-test and the post-test scores. The results revealed no main effect of instructional condition (adaptive versus control). However, moderator analyses demonstrate the presence of an interaction between instructional condition and prior knowledge. Specifically, the greatest learning gains were observed for beginners in the adaptive learning condition. Smaller gains were also observed by advanced learners in the adaptive learning condition. By comparison, intermediate performers learned most in the control condition. These results suggest that for “average” learners, the traditional, linear path to learning may be sufficient. Lessons learned and implications for future research are discussed.