Background: Adaptation to learning styles has been proposed to enhance learning. Objective: We hypothesized that learners with sensing learning style would perform better using a problem-first instructional method while intuitive learners would do better using an information-first method. Design: Randomized, controlled, crossover trial. Setting: Resident ambulatory clinics. Participants: 123 internal medicine residents. Interventions: Four Web-based modules in ambulatory internal medicine were developed in both "didactic" (information first, followed by patient problem and questions) and "problem" (case and questions first, followed by information) format. Measurements: Knowledge posttest, format preference, learning style (Index of Learning Styles). Results: Knowledge scores were similar between the didactic (mean [plus or minus] standard error, 83.0 [plus or minus] 0.8) and problem (82.3 [plus or minus] 0.8) formats (p = 0.42; 95% confidence interval [CI] for difference, -2.3 to 0.9). There was no difference between formats in regression slopes of knowledge scores on sensing-intuitive scores (p = 0.63) or in analysis of knowledge scores by styles classification (sensing 82.5 [plus or minus] 1.0, intermediate 83.7 [plus or minus] 1.2, intuitive 81.0 [plus or minus] 1.5; p = 0.37 for main effect, p = 0.59 for interaction with format). Format preference was neutral (3.2 [plus or minus] 0.2 [1 strongly prefers didactic, 6 strongly prefers problem], p = .12), and there was no association between learning styles and preference (p = 0.44). Formats were similar in time to complete modules (43.7 [plus or minus] 2.2 vs 43.2 [plus or minus] 2.2 minutes, p = 0.72). Conclusions: Starting instruction with a problem (versus employing problems later on) may not improve learning outcomes. Sensing and intuitive learners perform similarly following problem-first and didactic-first instruction. Results may apply to other instructional media.