Learning Innate Face Preferences

Abstract

Newborn humans preferentially orient to face-like patterns at birth, but months of experience with faces is required for full face processing abilities to develop. Several models have been proposed for how the interaction of genetic and evironmental influences can explain this data. These models generally assume that the brain areas responsible for newborn orienting responses are not capable of learning and are physically separate from those that later learn from real faces. However, it has been difficult to reconcile these models with recent discoveries of face learning in newborns and young infants. We propose a general mechanism by which genetically specified and environmentdriven preferences can coexist in the same visual areas. In particular, newborn face orienting may be the result of prenatal exposure of a learning system to internally generated input patterns, such as those found in PGO waves during REM sleep. Simulating this process with the HLISSOM biological model of the visual system, we demonstrate that the combination of learning and internal patterns is an efficient way to specify and develop circuitry for face perception. This prenatal learning can account for the newborn preferences for schematic and photographic images of faces, providing a computational explanation for how genetic influences interact with experience to construct a complex adaptive system.

Constructing visual function through prenatal and postnatal learning

“…For simplicity, the simulations bypassed V1, but including it leads to similar results (Bednar 2002;Bednar and Miikkulainen 2003). As for orientation, a variety of prenatal training conditions were simulated, to determine how different training patterns can lead to face preferences.…”

“…These RFs cause the network to respond to facelike schematic images (Figure 6), with total activity levels that rank the patterns in the order preferred by infants (Goren et al 1975;. When tested on 18 schematic patterns from newborn studies (Goren et al 1975;Valenza, Simion, Cassia, and Umiltà 1996), the full version of the model (including both V1 and the FSA) ranked them in the same preference order for all 22 of the statistically significant preferences found in newborns (Bednar 2002;Bednar and Miikkulainen 2003). These results demonstrate that a network exposed to three-dot patterns is sufficient to explain the experimental results with newborns tested with schematic stimuli.…”

“…Given a database of 150 top-lit images of adult males, the model responded in the correct location (the center of the face) to 88% of the images (Bednar and Miikkulainen 2003). Conversely, it responded to only 4% of the images in a database of natural scenes.…”

“…In addition to the three-dot pattern proposed by , we tested a variety of other prenatal training pattern shapes (Bednar 2002;Bednar and Miikkulainen 2003). For all training patterns, test faces that matched the training pattern size gave higher responses than did other natural images.…”

Self-Organization in a Parametrically Coupled Logistic Map Network: A Model for Information Processing in the Visual Cortex

“…Computational units of LISSOM are cortical columns that continuously adapt to afferent and lateral inputs, and the units synchronize and desynchronize their activity. This model has been partially successful in interpreting neurobiological facts such as columnar map organization as well as patchy connectivity, recovery from retinal and cortical injury, psychophysical phenomena such as tilt after effect [7], contour integration, and preference for faces [5].…”