What might a theory of mental imagery look like, and how might one begin formulating such a theory? These are the central questions addressed in the present paper. The first section outlines the general research direction taken here and provides an overview of the empirical foundations of our theory of image representation and processing. Four issues are considered in succession, and the relevant results of experiments are presented and discussed. The second section begins with a discussion of the proper form for a cognitive theory, and the distinction between a theory and a model is developed. Following this, the present theory and computer simulation model are introduced. This theory specifies the nature of the internal representations (data structures) and the processes that operate on them when one generates, inspects, or transforms mental images. In the third, concluding, section we consider three very different kinds of objections to the present research program, one hinging on the possibility of experimental artifacts in the data, and the others turning on metatheoretical commitments about the form of a cognitive theory. Finally, we discuss how one ought best to evaluate theories and models of the sort developed here.

This paper describes an operational computer simulation of visual mental imagery in humans. The structure of the simulation was motivated by results of experiments on,how people represent information in. and access information from; visual images. The simulation includes a "surface representation." which is spatial and quasi-pictorial, and an underlying "deep representation." which contains "perceptual" information encoding appearance plus "propositional" information describing facts about an object. The simulation embodies a theory of how surface images are generated from deep representations. and how surface images are pmcessed when one accesses information embedded in them. The simulation also offers an account of various sorts of imagery transformations.Our simulation is a model of how people represent information in, and later retrieve information from, visual .mental images. That is, whe'n retrieving some fact, people sometimes report generating an image and inspecting it; for example. if asked what shape are a beagle's ears, one may recall this information by mentally picturing the dog's head, and "looking" at the ears. We wanted to understand how such processes operate, and felt that a simulation approach was useful for a number of reasons: First. it allows explicit modeling of a system of complex processes. Second. it provides a "sufficiency proof," a demonstration that one's model is in fact adequate to account for some range of data, if the program runs as claimed. Third, we felt that by attempting to motivate construction of the model with actual data about human cognition, we would be confronted by important empirical issues that may otherwise have been overlooked; we also felt that such an approach would lead us to collect a set of data that would build, that would accumulate to form a "big picture" (cf. Newell. 1973; see also Kosslyn. in press).

Theories of visual pattern recognition frequently assume that processing begins with an analysis of the pattern into component parts, which are often assumed to be line segments of particular orientations, lengths, positions, and curvatures. The present experiments measured discriminability of these simple parts when presented either in isolation or within configural contexts that provided no formal information useful for the discrimination. Certain contexts either impaired or did not affect performance. Other contexts were found, however, which dramatically improved discriminability. Thus, two patterns which differed only in a single part could be discriminated from each other more quickly than could their distinguishing parts shown in isolation. Further experiments suggest that this "configural superiority" effect influences perceptual components of processing rather than memorial components. The mechanism underlying configural superiority appears to be the detection of novel and distinguishing features, such as corners and intersections, which emerge when parts are placed in close proximity to each other. The outlines of a model for preattentive feature discrimination are presented.

The nature of processing demands during a letter-match task was investigated in an extension of the Posner and Boies (1971) paradigm. In Experiment I, a visual probe was employed in addition to an auditory probe in two different experimental conditions. The shape of the auditory probe reaction time (RT) function was similar to that found by Posner and Boies. However, in contrast to their findings, RT was greatly increased shortly after presentation of the first letter for the visual probe function. It was concluded that perceptual as well as postperceptual limitations on processing capacity exist. A second experiment provided further support for this hypothesis.The study of attention has become increasingly prominent in the field of information processing. Results of recent studies have suggested that attentional effects are specific to mental operations which take place in the "central processor" (e.g

No abstract