Instructors traditionally lecture on new content before providing practice problems, but learning is often superficial. Exploratory learning before instruction deepens conceptual understanding by giving students a novel activity to explore before direct instruction. We examined how increasing the salience of contrasting cases in exploration versus practice activities impacts these learning benefits. Undergraduate students in statistics courses (Experiments 1-2) or a lab study (Experiment 3) completed a problem-solving activity either before (explore-first condition) or after (instruct-first condition) instruction about statistical variance. In Experiment 1 (N = 116), the problem-solving activity included a dataset with minimally contrasting cases. In Experiment 2 (N = 143), the activity increased the contrasts between cases to highlight important problem features. In Experiment 3 (N = 225), students were randomly assigned to complete problems with either minimal or increased contrasts. Students completed a posttest measuring conceptual understanding. When contrasts were minimal (Exps. 1&3), posttest scores were equal between explore-first and instruct-first conditions. When contrasts were increased (Exps. 2&3), posttest scores were higher in the explore-first condition compared to the instruct-first condition. However, in Experiment 3, minimal contrasts led to middling scores that were neither better nor worse than increased contrasts. Students in the instruct-first conditions used fewer correct problem steps on the learning activity when given increased compared to minimal contrasts, suggesting disengagement. Exploratory learning before instruction may help deepen cognitive engagement, showing benefits only when students might otherwise disengage.