Preattentive face processing: What do visual search experiments with schematic faces tell us?

Abstract

In recent research, several experiments have tested a preattentive threat-advantage hypothesis that threatening or negative faces can be discriminated preattentively, by using the visual search paradigm. However, supporting evidence is nonuniform, giving rise to the suspicion that stimulus factors rather than the stimuli's category of facial threat versus friendliness are responsible for sporadic demonstrations of a threat advantage. However, it is also possible that differences in experimental procedure contribute to the heterogeneous results. To test this possibility I selected examples from the past literature and presented them within the same constant experimental setting. I found a consistent advantage for negative face targets among positive face distractors with all stimulus pairs. Search slopes, however, mostly revealed inefficient search, questioning the preattentive discrimination of facial affect.

Acknowledgements

I am indebted to Christine Broermann and Sabine Dlugosch for conducting the experiments, to Lily-Maria Silny for her assistance in manuscript preparation, the Conscious and Nonconscious Processing Research Group for their valuable discussions of some of the results, and to Wulf-Uwe Meyer, and two anonymous reviewers, for their helpful comments on an earlier version of the manuscript.

Notes

¹A strict distinction between efficient and nonefficient processes (e.g., Treisman & Souther, 1985) has been convincingly criticized (cf. Wolfe, 1998), because search functions show a continuum of slopes, not a dichotomy. However, this critique does not imply that efficient search is not a necessary criterion for preattentive processing; it just says that it is not sufficient, because serial search can also be very efficient (e.g., Wolfe & Horowitz, 2004).

²Öhman et al. also obtained a crowd effect, with responses to happy crowds being faster and error rates being lower, at least for target present trials (Öhman et al. did not report the results for target absent trials separately for angry and happy crowds). For the sake of clarity, it should be emphasized that with a visual search paradigm, a crowd effect is not indicative of search efficiency: It is the slope (b) of the function relating RT to set size (y = bx + a) that reveals preattentive processing versus serial search, not the intercept term (a) of the function. Rather, the intercept reveals processes that occur before the beginning of the search or between the termination of the search and the production of the response, but not during search. For example, faster RTs to positive crowds may reveal less hesitation in beginning with the scanning of the positive crowd, a longer time in deciding that really no discrepant face is present for negative crowds, or a slower response execution with negative crowds. Either of these effects may be due to genuinely affective or to purely perceptual factors.

³Initially, we chose this rather stringent exclusion criterion because we feared that speed–accuracy tradeoffs that are specific for certain conditions would level out differences in RTs between these conditions. Later it turned out that the exclusion did barely change the patterning of the RTs. However, because the experimental design assumes that method factors (order of conditions and response mapping) are balanced across participants, we present the data as we originally collected them.

⁴As suggested by one reviewer, a continuous variation of facial threat (e.g., through morphing) in the targets might be helpful. If categorical perception of faces (Calder, Young, Perrett, Etcoff, & Rowland, 1996) affects search efficiency, one would expect a step-function relating the degree of facial threat to search efficiency, but not a linear function.