Evidence of fixed capacity in visual object categorization

Abstract

How is visual object perception limited by divided attention? Whereas some theories have proposed that it is not limited at all (unlimited capacity), others have proposed that divided attention introduces restrictive capacity limitations or serial processing (fixed capacity). We addressed this question using a task in which observers searched for instances of particular object categories, such as a moose or squirrel. We applied an extended simultaneous-sequential paradigm to test the fixed-capacity and unlimited-capacity models (Experiment 1). The results were consistent with fixed capacity and rejected unlimited capacity. We ascertained that these results were due to attention, and not to sensory interactions such as crowding, by repeating the experiment using a cuing paradigm with physically identical displays (Experiment 2). The results from both experiments were consistent with theories of object perception that have fixed capacity, and they rejected theories with unlimited capacity. Both serial and parallel models with fixed capacity remain viable alternatives.

Fig. 1

Schematic of experimental conditions used in Experiment 1. Each condition begins with category cues, indicating the two relevant categories for that trial. Stimuli are then presented. After the presentation the category cues return, and the observer responds by indicating which category was present in the display

Fig. 2

Illustration of the predictions of unlimited-capacity and fixed-capacity models. In the unlimited-capacity model, performance is a function of exposure duration but not of the number of competing stimuli. Thus, the predicted pattern is simultaneous = sequential < repeated. In the fixed-capacity model, performance is a function of both exposure duration and competition among stimuli. The predictions of the model are demonstrated in terms of a serial model that allows two stimuli to be analyzed per frame. In the example shown, only two stimuli can be analyzed in the simultaneous condition, but four stimuli can be analyzed in both the sequential and repeated conditions. This yields the predicted pattern: simultaneous < sequential = repeated

Fig. 3

Example stimuli used in the study. Examples of all eight categories are shown. Each category included 30 images, of which 2 are shown

Fig. 4

Results of Experiment 1: Percent correct in the simultaneous, sequential, and repeated conditions, plotted with error bars representing ± 1 standard error of the mean. The dotted and dashed lines represent predictions from the fixed-capacity models (sequential = repeated) and unlimited-capacity models (sequential = simultaneous), respectively

Fig. 5

Schematic of the experimental conditions used in Experiment 2. Experiment 2 repeated the simultaneous–sequential comparison using physically identical stimulus displays. Each condition included eight unique images, four in each frame, with only one target among them. Cues preceding each frame indicated possible target positions. The neutral condition indicates that the target can appear in any location in either frame. The simultaneous cues indicate that the target can appear in a particular frame (the second frame, in this example). The sequential cues indicate that the target can occur within a particular subset of each frame

Fig. 6

Results of Experiment 2: Percent correct in the neutral, cued-simultaneous, and cued-sequential conditions. The dashed line represents the prediction from unlimited-capacity models (cued-sequential = cued-simultaneous). The dotted line represents the prediction from fixed-capacity models (cued-sequential = repeated condition from Exp. 1). The neutral condition was included as a check for cue effectiveness; ineffective cuing would be indicated by equivalent performance in all three conditions