Research report Changing patterns of brain activation during category learning revealed by functional MRI Deborah M. Little a, * , Raymond Klein a,b , Donna M. Shobat a , Erik D. McClure a , Keith R. Thulborn a a Center for Magnetic Resonance Research, College of Medicine, University of Illinois, Room 1193, MC 707, 1801 W. Taylor Street, Chicago, IL 60612, United States b Department of Psychology, Dalhousie University, Halifax, Nova Scotia, Canada Accepted 23 July 2004 Available online 2 September 2004 Abstract Functional magnetic resonance imaging (fMRI) was used to investigate neural changes as a function of category learning in normals (n =8). Subjects were trained to classify patterns of dots into four categories over 4 consecutive days. fMRI monitored the changes that occurred during learning prior to training and then following each training session. During fMRI, subjects determined whether two patterns of dots were members of the same category. The behavioral changes that occurred as a result of the training were observed as increases in response accuracy within shortened response times. fMRI illustrated initial increases in volumes of activation distributed across the known visuospatial processing networks. The regions affected by learning were identified as those involved in the planning and execution of eye movements (frontal and supplementary eye fields, FEF and SEF), spatial attention (superior and inferior parietal lobules, SPL and IPL) and visual processing (primary, secondary, and tertiary visual cortices). The volumes of activation then decreased as training progressed further. Of the two proposed mechanisms for learning, that of strengthened connectivity on a given network and that of selection of different networks, our data supports the former. D 2004 Elsevier B.V. All rights reserved. Theme: Neural basis of behavior Topic: Learning and memory: systems and functions Keywords: Category learning; Prototype distortion; Human brain networks; fMRI 1. Introduction The neural changes within the human brain during category learning have yet to be fully documented [3,22]. Category learning has been well-characterized behaviorally [28,41], with a variety of different materials and methods that range in breadth from the more ecologically valid (i.e., classification of birds; Ref. [4]) to more experimentally defined (i.e., random dot patterns; Ref. [36]) materials. Although category learning is an important skill, the neuronal response that underlies it is not well understood. Learning is the acquired behavioral change associated with improved performance on a given task [12,16,33]. The behavioral changes associated with learning are commonly observed as increases in accuracy within shortened response times [2,25]. The rate or speed of learning can be manipulated by changing task difficulty and by including or excluding feedback [1,27]. Unlike category learning, the neural changes associated with other areas of skill learning (such as learning a motor sequence) have been fairly well explored (for reviews, Refs. [16,33]). The neuronal response to successful skill learning can be described by and mapped onto multiple neural networks. The specific network is dependent upon the skill required to complete the task (motor, visual–spatial or 0926-6410/$ - see front matter D 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.cogbrainres.2004.07.011 * Corresponding author. Tel.: +1 312 355 3086; fax: +1 312 355 3085. E-mail address: little@uic.edu (D.M. Little). Cognitive Brain Research 22 (2004) 84 – 93 www.elsevier.com/locate/cogbrainres