Recent findings suggest that when solving problems involving cognitive flexibility (CF), individuals who approach alearning task using reinforcement learning (RL), outperform those who approach the task using supervised learning (SL).Based on these data, we hypothesized that CF is a function of individual differences in learning preference and taskdemands. Healthy native English speakers were administered three CF tasks that incorporated (i) shifting, (ii) divergentthinking, or (iii) both shifting and divergent thinking elements. Participants response selection history on a reward-basedlearning task, which could be approached either through SL or RL, was used to determine each participants learning styleand predict CF performance. Results showed that different CF task components (i.e., whether the task involved divergentthinking) interacted with participants learning preferences as measured by the independent learning task. We discuss howlearning preferences might capture individual differences in CF.