This Title All WIREs
How to cite this WIREs title:
WIREs Cogn Sci
Impact Factor: 3.476

Visual category learning

Full article on Wiley Online Library:   HTML PDF

Can't access this content? Tell your librarian.

Visual categories group together different objects as the same kinds of thing. We review a selection of research on how visual categories are learned. We begin with a guide to visual category learning experiments, describing a space of common manipulations of objects, categories, and methods used in the category learning literature. We open with a guide to these details in part because throughout our review we highlight how methodological details can sometimes loom large in theoretical discussions of visual category learning, how variations in methodological details can significantly affect our understanding of visual category learning, and how manipulations of methodological details can affect how visual categories are learned. We review a number of core theories of visual category learning, specifically those theories instantiated as computational models, highlighting just some of the experimental results that help distinguish between competing models. We examine behavioral and neural evidence for single versus multiple representational systems for visual category learning. We briefly discuss how visual category learning influences visual perception, describing empirical and brain imaging results that show how learning to categorize objects can influence how those objects are represented and perceived. We close with work that can potentially impact translation, describing recent experiments that explicitly manipulate key methodological details of category learning procedures with the goal of optimizing visual category learning. WIREs Cogn Sci 2014, 5:75–94. doi: 10.1002/wcs.1268 This article is categorized under: Psychology > Theory and Methods Psychology > Learning Neuroscience > Cognition
Examples of objects used in visual category learning experiments: (a) Gabor patches varying in orientation and spatial frequency; (b) simple objects varying in shape, size, and shading; (c) simple line drawings of ships varying in the shape of the wings, porthole, tail, and nosecone; (d) novel contour shapes defined by Fourier descriptors; (e) random dot patterns; (f) greebles that vary in body shape and the shapes of three appendages; (g) ziggerins that vary in shape and style; (h) novel cars created my morphing.
[ Normal View | Magnified View ]
Schematic illustration of a possible visual category learning experiment. Some category learning experiments are conducted in a single session (e.g., Refs ) and some over multiple sessions (e.g., Refs ). Most category learning experiments include explicit training blocks, where typically there is a set of objects that are used as training items and corrective category feedback is provided on every trial. These may be followed by test blocks that include training items as well as transfer items and no corrective feedback is provided (e.g., Refs ). Sometimes transfer blocks are also interspersed at various key time points over the course of category learning (e.g., Ref ). A common structure of a single training trial is to present an object, either for unlimited time or for an experimentally determined amount of time, record the subject category response, and provide feedback, sometimes with the object still visible and sometimes not. As discussed in the text, the timing of the feedback and whether intervening tasks are used within a trial can be manipulated (e.g., Refs ). An intertrial interval (ITI) separates one trial from the next. The trial‐by‐trial contingencies can influence performance, as discussed in the text (e.g., Refs ).
[ Normal View | Magnified View ]
Examples of a prototype dot pattern, low‐level distortion of the prototype, high‐level distortion of the prototype, and new unrelated random dot pattern used in dot pattern category learning studies.
[ Normal View | Magnified View ]
Examples of psychological spaces of objects and category structures that are learned. (a) Objects are semicircles that vary in size (four levels) and the angle of a radial line drawn from the center to the edge (four levels). Subjects learn objects as members of two categories (A and B) and are tested after category learning on the learned objects (As and Bs) and new transfer objects (Ts). The top category structure is linearly separable; the bottom category structure is not. (b) Objects are Gabor patches that vary in orientation and spatial frequency. Each point is an object in psychological space. Subjects learn objects as members of category A (black squares) or category B (open circles). Top shows an information integration (II) category structure that requires both dimensions. Bottom shows a rule‐based (RB) category structure that requires only one dimension (in this case, spatial frequency). (c) Objects are morphs of cars. The two dimensions are themselves morphs between two parent cars (four different parent morphs in total). Each car in the morph space is a morph between a morph along the horizontal dimension and a morph along the vertical dimension. Two possible category structures are shown, one with a vertical category boundary defining horizontal morph dimension as relevant, and another with a horizontal category boundary defining the vertical morph dimension as relevant.
[ Normal View | Magnified View ]
(a) Six types of category structures tested by Shepard et al.; see also Refs . These types reflect the six possible ways of dividing up a set of stimuli defined along three binary‐valued dimensions into categories of equal size (see also Ref ). The three dimensions of the cube represent the three psychological dimensions, with each corner of the cube representing a possible object. For each type, the color coding (white or black) denotes members of one category versus the other category. Example of an assignment of logical dimensions to physical dimensions shown below. (b) Category learning data as P(Error) as a function of learning block for the six types when object dimensions are separable (shape, size, and color). (c) Category learning data as P(Error) as a function of learning block for the six types when object dimensions are integral (hue, saturation, and brightness of color patches). Note the different qualitative ordering of category learning difficulty depending on whether separable or integral dimension objects are used.
[ Normal View | Magnified View ]

Browse by Topic

Neuroscience > Cognition
Psychology > Theory and Methods
Psychology > Learning

Access to this WIREs title is by subscription only.

Recommend to Your
Librarian Now!

The latest WIREs articles in your inbox

Sign Up for Article Alerts