The face-inversion effect as a deficit in the encoding of configural information: direct evidence.

We report four experiments leading to conclusions that: (i) the face-inversion effect is mainly due to the deficits in processing of configural information from inverted faces; and (ii) this effect occurs primarily at the encoding stage of face processing, rather than at the storage stage. In experiment 1, participants discriminated upright faces differing primarily in configuration with 81% accuracy. Participants viewing the same faces presented upside down scored only 55%. In experiment 2, the corresponding discrimination rates for faces differing mainly in featural information were 91% (upright) and 90% (inverted). In experiments 3 and 4, the same faces were used in a memory paradigm. In experiment 3, a delayed matching-to-sample task was used, in which upright-face pairs differed either in configuration or features. Recognition rates were comparable to those for the corresponding upright faces in the discrimination tasks in experiments 1 and 2. However, there was no effect of delay (1 s, 5 s, or 10 s). In experiment 4, we repeated experiment 3, this time with inverted faces. Results were comparable to those of inverted conditions in experiments 1 and 2, and again there was no effect of delay. Together these results suggest that an 'encoding bottleneck' for configural information may be responsible for the face-inversion effect in particular, and memory for faces in general.