Evidence for rule versus exemplar learning strategies as stable individual differences independent from working memory

Evidence for two qualitatively different learning strategies has emerged from the function- and category-learning literatures: a rule-based and an exemplar-based strategy. With a rule-based strategy, learners abstract some common principle from training items, which allows extrapolation to novel instances. With an exemplar-based strategy, learners encode training items without abstraction, which facilitates generalisation based on surface similarity to trained items. Previous studies offer preliminary evidence that strategies are stable; that is, convergent performance was found across pairs of disparate tasks. The current paper advances this work by examining whether performance across a battery of tasks converges, providing evidence for a latent variable underlying learning strategy. Subjects completed five learning strategy and three working memory tasks. Using data reduction and latent structure modelling methods, we found evidence for a general strategy construct that was unrelated to working memory. This is important because it shows that differences in learning strategy are not simply due to differences in working memory.