Task-specific strategies of mental "rotation" of facial representations.

ed, coded, and/or incomplete reflection of the object itself. Mental transformations are viewed as operations on the structure of a propositional network that lead to changes in the way the representation can be used. Investigators have attempted to devise procedures that would permit inferences as to whether analog or propositional processing is being used in mental rotation tasks, despite the arguments of some that empirically based distinctions are logically impossible to make (e.g., Anderson, 1978). In fact, convincing empirical evidence that exclusively favors either propositional or analog representations has not been forthcoming. It is likely that the requirements of the tasks required of subjects in mental representation experiments affect the apparent form of the representations used by subjects in completing those tasks. Richman, MitchelI, and Reznick (1979) have argued that, in mental scanning and travel tasks, the instructions given subjects can determine whether results consistent with an analog model or a proposition al model are produced. Others (e.g., Corballis, Zbrodoff, Shetzer, & Butler, 1978; Pylyshyn, 1979) have found that the nature of the stimulus is important in mental rotation tasks. In fact, the requirements of the experimental task are often used by adherents of one side of the mental representation debate in criticizing research by adherents of the other side. For example, Pylyshyn (1979) designed an experiment to demonstrate that, contrary to then eurrent theories of mental imagery, subjects "rotate" simple stimuli more rapidly than complex stimuli. Copyright 1983 Psychonomic Society, Inc. 42 COCHRAN, PICK, AND PICK Cooper and Podgomy's (1976) experiment, which produced results that failed to support such a complexity effect, were thus called into question. Kosslyn (1981), on the other hand , criticized the task demands of Pylyshyn 's (1979) experiment, arguing that strategies were induced in subjects that resulted in data that should not be interpreted as denying the possibility of analog represen tation. Although the importance of task demands is increasingly acknowledged, task demands per se have not been thoroughly investigated. Presumably, different visual camparisan tasks induce different strategies in subjects, resulting in data that reflect the specific strategy that was induced. In order to determine the importance of task demands in producing specific effects, experiments are needed in which the nature of the task is changed, in a systematic way, while other factars are held constant. The present studies were designed to investigate the influence of task demands in mental representation experiments. Specifically, we investigated the extent to which the results of a typical mental rotation experiment changed as a function of the type of same-different judgment subjects were required to make. In addition, we hypothesized that the types of patterns subjects are asked to represent and manipulate affect the apparent nature of their representation. Although the effect of figural complexity on the speed of mental rotation has been investigated often (e.g., Cooper & Podgorny, 1976; Pylyshyn, 1979), the effect of figural organization has not. Shepard (e.g., Shepard, 1975) has referred to the impression of subjects performing his mental rotation task that the representation being rotated "breaks up," or loses its cohesiveness, if the rotation is attempted at tao great a rate. Consistent with Shepard's view is the hypothesis that the representations of weil organized patterns could be mentally rotated or manipulated faster, while still maintaining their cohesiveness, than could similar, but less highly organized patte rns. Human faces represent one type of pattern in which the configuration of features is important. Adults recognize patterns that are facial configurations much more rapidly than formally similar patterns composing a nonfacial configuration (e.g., Bradshaw & Wallace, 1971), and the special nature of faces is evident even for infants, who prefer to gaze at them instead of other types of patterns (e.g., Fantz, 1961). The perception of misoriented faces has been exarnined by others. Rock (1973) has noted that faces are exceedingly difficult to recognize when inverted, perhaps because the individual inverted features cannot be processed simultaneously. Carello (Note 1) attributes the difficulty of tasks requiring the recognition of inverted faces to our tendency to analyze faces in terms of invariants that do not withstand physical rotation. Unfortunately, since the mental manipulation of faces has not been experimentally compared to the manipulation of nonfaces of equivalent complexity, it is difficult to interpret findings that the recognition of inverted faces is difficult. The difficulty could result from the special nature of faces, or from the requirements imposed by a recognition task involving faces, or from same other, as yet undetermined factor. It is, in fact, likely that the nature of the stimulus and the requirements of the task interact: Mental rotation of faces may not be possible if information necessary to carry out some types of perceptual tasks (e.g., recognition) is to be preserved. For other tasks (e.g., same-different judgments), the presence of facial structure in a stimulus may facilitate mental rotation and/or perceptual matching. Dur experiments examined this hypothesis. Specifically, we hypothesized that, for our same-different matehing tasks, the presence of facial organization would facilitate mental rotation: Representations of facial configurations, because of their inherent organization, should be rotated more rapidly than representations of comparable nonfacial configurations. That is, the time required to match faces should be less than that to match nonfaces, and the difference should increase with the angular difference between the configurations being compared. For other matehing tasks, particularly those that discourage mental rotation, the advantage of facial organization should be less