Abstract
The composite face effect (CFE, Young, Hellawell, & Hay, 1987) has, in recent years, been suggested as one possible empirical signature of the holistic (configural, gestalt, etc.) characteristics of facial perception and cognition. In CFE people's performance with one part of a composite face appears to be dependent on the other. Theoretical analyses of the CFE using multidimensional signal detection theory (general recognition theory, GRT) has suggested that the behavioral regularities can potentially have both perceptual and decisional sources, with recent empirical studies documenting the influence of decisional factors in the CFE. However, GRT (like classical univariate signal detection theory) addresses behavioral regularities without assuming specific mechanisms. Consequently, the present study investigated one possible source for the decisional factors that can be involved in the DFE: differential attentional weighting. Our hypothesis was that observers will distribute visual attention to the two components of a component in accord with the statistical regularities of the presentation frequencies, and that shifts in the distribution of attention will drive shifts in response in criteria. We tested this hypothesis using a composite face task, in which we varied the base rates (e.g., prior frequencies) for the two halves of the composite stimuli. The base rate manipulation imposed correlational structure on the dimensional space (Garner, 1974), and thus allowed for the emergence of various decisional criteria within individual observers. This enabled us to relate statistical regularities in the stimulus space to various GRT constructs, including those that tap the decisional components. Based on our results, we highlight the need for further theorizing and experimentation on the role of attentional mechanisms in configural face perception.