Recognition memory is more accurate when faces are inverted than when they are upright

The face inversion effect (FIE) indicates a dramatic impairment of recognition for upside-down faces (Yin, 1969). The FIE has been attributed to memory encoding, but it has also been found in the simultaneous presentation of face images (e.g., Farah, Wilson, Drain, & Tanaka, 1998). Here, we investigate the FIE by using perceptually matched upright and upside-down faces. Method. Experiment 1. Face continua were generated by morphing between two faces having different identities. A psychometric procedure was used to select pairs of morphed faces, either upright or upside-down, that produced 55, 65, 75, and 85% correct performance in a perceptual discrimination task (i.e., when two face images were simultaneously presented). In Experiment 2, the upright and upside-down face pairs, which had been found to be equivalent in terms of perceptual discriminability, were used in an old/new recognition memory paradigm. Results. Experiment 1: Equivalent levels of perceptual performance required larger morph distances for the upside-down faces than for the upright faces (this result replicates the face inversion effect). Experiment 2: Recognition memory performance was better for the upside-down than for the (perceptually matched) upright faces. The advantage of upside-down faces in recognition memory decreased with increasing perceptual dissimilarity between the memory and the probe image. Conclusions. When upright and upside-down faces are equally discriminable in a perceptual task, performance in memory recognition is better for the upside-down than for the upright faces. This result is the opposite than expected by the FIE. We discuss this finding in the light of the different principles that may underline the organization of upright and upside-down faces in the similarity face-space (Valentine, 1991).