The influence of feature conjunction on object inversion and conversion effects.

ABSTRACT

The face inversion effect is considered by some as a marker ofholistic or configural processing. In an effort to understand the relations between object orientation and the use of configuration for recognition, we explored the extent to which the combination of information across discrete, spatially separated object features contributed to the impact of transforming the orientation of faces and other objects. Subjects performed a matching task on four different sets of faces and novel objects. For two of the sets (diagnostic), the match could be made based on information contained in a single feature. For the remaining two sets (conjunction and disjunction), the information useful for successful matching was divided between two spatially separated features (at the top and at the bottom of the object). For OR disjunctions the top and bottom features provided redundant information such that subjects could rely on either the top or the bottom feature, whereas for AND conjunctions the top and bottom features provided unique information such that subjects needed to use both the top and the bottom features. Experiment 1 assessed the cost of inversion on these stimulus sets with the typical inversion task, where both study and test stimuli were presented in the same orientation. Consistent with previous work, significant inversion effects were found for all of the face sets, but none of the novel object sets. Experiment 2 assessed the cost of 'conversion'--that is, the effect of transforming the orientation of the stimulus between study and test. Consistent with previous work, significant conversion effects were found for all face sets; however, a significant conversion effect was also found for the AND novel set. That only the AND novel set was significant suggests that conversion effects are reliant on combination of information across features, which is compromised when the stimulus configuration is altered between study and test. The results suggest that configural processes, like those purportedly used for face recognition, may be domain-general processes that can also be recruited strategically with other types of objects.