Date of Award


Document Type


Degree Name

Master of Arts (MA)




One important assumption of information processing models of human perception is that such processing occurs as a sequence of events or stages which take place over time. Recent investigations (Navon 1977, 1981) have sug gested that the global aspects of a stimulus are perceived at an earlier stage in the time course of processing than are the local elements of the stimulus. Other studies (Kinchla § Wolfe 1979; Martin 1979; Hoffman 1980) have sug gested that neither global forms nor local elements are necessarily perceived prior to the other structural level, but that other factors such as the size of the stimulus, the sparcity of the local elements, and the goodness of the form in the stimulus mediate the perceptual precedence of global and local levels.

Attempts to generalize the results of the above studies to real-world perception are very tentative. The nature of the stimuli used in global precedence research is markedly different from that of real-world stimuli. Pomerantz (1981) has defined two types of relationships that exist between the global and local levels in visual stimuli: In one type there is no predictive relationship between the global and local levels (i.e., their identities are independent), while the second type contains mutually predictable (dependent) global and local levels. Real-world viewing involves the processing of stimuli of the latter type, but the global precedence studies to date have all utilized stimuli with independent global and local levels. It is not known whether generalizations can be made across these two configural types with respect to the perceptual precedence of the global or local levels.

The present study investigated perceptual precedence in pictorial stimuli in which the identity of global and lo cal levels were dependent on one another. Perceptual pre cedence was measured through the use of a Stroop-like inter ference task, similar to that used by Navon C1977). The task required that subjects direct their attention to either the global or local level, as cued by the experimenter prior to each trial, and then respond "yes" or "no" to the pres ence of the cued object (local level) or scene (global level). Response latencies were recorded for each trial. Half of the trials contained inconsistent global and local levels, and the interference produced by the irrelevant level was taken as evidence for the processing precedence of that level. Stimulus display size was also varied across trials.

The results indicated that for small scenes, subtend ing 4° of visual angle, global precedence did occur, but for large scenes (16°) the opposite effect ("local prece dence") was found. The pattern very closely parallels the findings of Kinchla and Wolfe (1979), using global local independent stimuli, suggesting that the relationship between the global and local levels is not critical in de termining perceptual precedence.

A model was proposed in which the structural level first perceived is determined by the spatial frequency or size of the stimulus display. The model suggests that a critical sampling bandwidth exists and that the initial pro cessing of a stimulus occurs at that level whose spatial frequency falls within the bandwidth. A post-hoc examina tion of the spatial frequencies present in the stimuli used in this study suggests that the band is centered at about 4 contour changes per degree of visual angle and ranges from 2 to 8 changes per degree.