The distinctiveness effect reverses when using well-controlled distractors

The distinctiveness effect in memory holds that distinctive items (e.g., unusual objects, infrequent words, or atypical faces) have a recognition advantage over typical items (e.g., Geraci & Rajaram, 2002; Schmidt, 1991; Valentine, 1991). The recognition advantage is usually operationalized as a higher hit rate and lower false alarm rate for distinctive items. Although the effect has been replicated numerous times in various domains, the conclusion that distinctive items are remembered better than typical items seems at odds with other findings in cognitive psychology. In particular, typicality effects, such as the recognition advantage for prototypical colours (Lucy & Shweder, 1979) and the improved discrimination of own-race faces (e.g., Walker & Tanaka, 2003), suggest that we have better mental representations for frequently encountered, typical items. Akey factor that isoftenoverlooked instudiesondistinctiveness is theroleof foils,ordistractors.Infacerecognitionexperiments,subjectsstudyasetoftarget facesand later try topickthe target faces fromamongdistractor faces, either ina forced-choice or old/new paradigm. Although experimenters usually select distractorsrandomly, thisdoesnotguaranteeanequitablecomparisonbetween thetwotypesoffaces.Thecentraldistributionoffacespace (seeValentine,1991) actually predicts a systematic relationship between the distinctiveness of a face and its overall similarity to randomly chosen distractors. A simple geometric analysis shows that random distractors will be statistically more similar to typical faces (located more centrally in face space) than to distinctive faces (located more peripherally). This asymmetry implies that distinctive faces will be overall more distinguishable from random distractors, and therefore more recognizable when pitted against them. We propose that the well-reported distinctiveness advantage in face recognition is primarily due to this asymmetry. We further propose that if the asymmetry is eliminated, typical faces should be recognized more accurately than distinctive faces, consistent with other typicality effects. To this aim, we describe two studies that measure the accuracy of face