Accounts of learning and generalization typically focus on factors related to lasting changes in representation (i.e., long-term memory). The authors present evidence that shorter term effects also play a critical role in determining performance and that these recency effects can be subdivided into perceptual and decisional components. Experimental results based on a probabilistic category structure show that the previous stimulus exerts a contrastive effect on the current percept (perceptual recency) and that responses are biased toward or away from the previous feedback, depending on the similarity between successive stimuli (decisional recency). A method for assessing these recency effects is presented that clarifies open questions regarding stimulus generalization and perceptual contrast effects in categorization and in other domains.

Available online at www.sciencedirect.com

Stimulus generalization is often regarded as a fundamental component of category learning, yet it has not been directly studied in this context. Here we develop a technique for measuring generalization based on sequential effects in subjects ’ responses. We find that patterns of generalization can adapt to global properties of the task, but only when the category structure is defined by perceptually primitive and separable dimensions. Implications are discussed for attentional learning and the nature of both perceptual and

Similarity is often regarded as a fundamental construct underlying stimulus generalization in category learning and many other domains. The key assumption of this approach is that multidimensional differences between stimuli are summarized by a single value before entering the decision process. The present study challenges this assumption by showing that category judgments depend on the full relationship between present and past stimuli, in a way that cannot be mediated by a unidimensional similarity measure. Approaches based on response generalization, knowledge partitioning, and distributional representations are also shown to be insufficient to account for our findings.

Sequential effects reflect parallel learning of multiple

Writing this thesis would not have been possible without the help of a number of people. In particular, I would like to acknowledge my supervisor Jarl Giske for believing he could turn a field ornithologist into a marine modeller. You did not quite succeed, partially because you have given me the freedom to explore my interest in general ecological questions. Your ideas and enthusiasm have kept me going and I appreciate that you have always found time to discuss and comment on my work. I am grateful to my co-supervisor Marc Mangel. The few months in your dynamic lab group at University of California, Santa Cruz, were most enjoyable. Your sharp thinking and insightful comments have inspired and focused my own work. A special thank to Christian Jørgensen for paying such great interest in the progress of my thesis. I have really appreciated our numerous discussions, your genuine enthusiasm and generosity. Your creative thinking and clever comments have improved the work a lot. I am grateful to my other co-authors Øyvind Fiksen and Josefin Titelman for introducing me to the challenges of the marine ecosystem. I would also like to thank

Exemplar sequencing effects in incidental and intentional unsupervised category learning were investigated to illuminate how people form categories without an external teacher. Stimuli were perfectly separable into 2 categories based on 1 of 2 dimensions of variation. Sequencing of the first 20 training stimuli was manipulated. In the blocked condition, 10 Category A stimuli were followed by 10 Category B stimuli. In the intermixed condition, these 20 stimuli were ordered randomly. Experiment 1 revealed an interaction between learning mode and sequence, with better intentional learning for intermixed sequences but better incidental learning for blocked sequences. Experiment 2 showed that manipulating trial-to-trial variability along each dimension can impact intentional learning. Training sequences that emphasized variation along the category-relevant dimension resulted in better performance than sequences that emphasized variation along the category-irrelevant dimension. The results suggest that unsupervised category learning is influenced by the mode of learning and the order and nature of encountered exemplars.