Love, Rouder, and Wisniewski (1999) obtained interesting results showing that, in a same/different task on abstract visual scenes, subjects were able to process global properties quickly, even before local properties were identified. Our aim in this work is to explore more fully the complex relationships that exist between local processing and global processing. In our first experiment, we tested the robustness and generality of these global and local effects by using another, very different kind of local element. We showed that the global effects remain strong even when the local elements are neither conventional nor easily discriminable. In the second experiment, we showed that there exists an intermediate level of similarity between purely local and purely global similarity. Furthermore, we found that even when a stronger form of local dissimilarity is manipulated (through the introduction of different local elements), global effects were still observed. We conclude with a discussion of the respective roles of global and local properties in light of our findings.
The topic of whole/part perception has a long, controversial history with many ramifications in several domains of cognitive psychology. It is probably in the field of visual perception that this topic has been studied the most. Since the pioneering work of Navon (1977), the central question has been whether people perceive the whole first and then decompose it into parts, or perceive the parts first and then integrate them into a whole. Despite the strong interest in this question both for visual cognition and for cognition in general, many contradictory answers have been proposed (Kimchi, 1992).
In the area of visual scene comparison, Love, Rouder, and Wisniewski (1999) used an original manipulation of local/global properties and obtained very interesting results that are relevant to two classically separate domains of cognition: reasoning by analogy and visual cognition. In their experiments, subjects performed a discrimination task in which they had to indicate very quickly whether a standard matrix (i.e., a source) and a comparison matrix (i.e., a target), each composed of simple geometric elements (squares, circles, and triangles), were the same or different. The significance of this study is related to the relationship between local and global properties. In many experiments conducted to explore the local/global distinction (for example, classical nested-letters experiments), global precedence can be explained by the difference in the conspicuousness of local and global properties. However, with the material used in the present study (see Figure 1), the conspicuousness of local and global properties was controlled because it was impossible to identify global structures (i.e., to form class equivalencies between local elements) without previously identifying local elements. Consequently, any global precedence effect cannot be explained by differences in the conspicuousness of global forms. Other, more relational or structural explanations can be proposed.
Interestingly, although the subjects could respond "different" only after processing local elements (source and target matrices always differed locally on at least six local elements), they were strongly influenced by global properties. Indeed, the subjects were slower to respond "different" when the matrices were globally the same (Matrices A and B)-that is, when they were analogue. Moreover, the degrees of global and local match interacted, because local matches slowed the subjects down only when the two matrices matched globally. The latter result suggests a global precedence: When matrices were globally different, these global differences allowed the subjects to quickly respond "different" even when there were local similarities. Local similarities had a clear and strong effect only when the matrices were globally the same.
Different Kinds of Local Properties and Their Influences on Global Processing
Several crucial questions remain about the complex relationships that exist between local and global processing. …