Analogical transfer in perceptual categorization Michael B. Casale & Jessica L. Roeder & F. Gregory Ashby Published online: 20 December 2011 # Psychonomic Society, Inc. 2011 Abstract Analogical transfer is the ability to transfer knowledge despite significant changes in the surface features of a problem. In categorization, analogical transfer occurs if a classification strategy learned with one set of stimuli can be transferred to a set of novel, perceptually distinct stimuli. Three experiments investigated analogical transfer in rule-based and information-integration categori- zation tasks. In rule-based tasks, the optimal strategy is easy to describe verbally, whereas in information-integration tasks, accuracy is maximized only if information from two or more stimulus dimensions is integrated in a way that is difficult or impossible to describe verbally. In all three experiments, analogical transfer was nearly perfect in the rule-based conditions, but no evidence for analogical transfer was found in the information-integration condi- tions. These results were predicted a priori by the COVIS theory of categorization. Keywords Analogy . Categorization . Cognitive neuroscience . Implicit learning . Perception The ability to transfer knowledge from a familiar context to a new, unfamiliar context is an extremely important skill. Without it, we would spend much of our lives relearning previously acquired knowledge. Analogical transfer is the ability to transfer knowledge despite significant changes in the surface features of the problem (e.g., Catrambone, Craig, & Nersessian, 2006; Cho, Holyoak, & Cannon, 2007; Duncker, 1945; Gick & Holyoak, 1980, 1983; Holyoak & Koh, 1987; Redington & Chater, 2002). Analogical transfer has not been studied much in percep- tual categorization. In a categorization domain, analogical transfer occurs if a participant is able to apply a classification strategy learned with one set of stimuli to a set of novel, perceptually distinct stimuli. In some cases, it seems certain that analogical transfer would succeed. For example, once children learn to sort toy blocks into small and large categories, it is likely that they could quickly learn to sort toy automobiles into small and large categories, even though toy automobiles and toy blocks share few surface features. Less clear is whether analogical transfer occurs with catego- rization rules that are more difficult to describe verbally. In rule-based (RB) category-learning tasks, the catego- ries can be learned via an explicit hypothesis-testing procedure (although they could also be learned in other ways). In most cases, the rule that maximizes accuracy is easy to describe verbally (Ashby, Alfonso-Reese, Turken, & Waldron, 1998). In the simplest examples only one stimulus dimension is relevant, and the participant’ s task is to discover this dimension and then to map the different dimensional values to the relevant categories. More difficult RB tasks require attention to two or more dimensions. For example, the correct rule might be a conjunction of the type “The stimulus is in Category A if it is large and bright.” The key requirement is that the correct categorization rule in RB tasks is one that can be discovered by an explicit hypothesis-testing procedure. In contrast, accuracy in information-integration (II) tasks is maximized only if information from two or more stimulus dimensions is integrated at some predecisional stage. Typically, the optimal strategy is difficult or impossible to describe verbally (Ashby et al., 1998). M. B. Casale (*) Department of Psychology, University of California, 9500 Gilman Drive #0109, La Jolla, San Diego, CA 92093-0109, USA e-mail: mbcasale@gmail.com J. L. Roeder : F. G. Ashby University of California, Santa Barbara, CA, USA Mem Cogn (2012) 40:434–449 DOI 10.3758/s13421-011-0154-4