A Computational Model of the Development of Hemispheric Asymmetry of Face Processing


Extensive research effort has been invested in building neurocomputational models for face and object recognition. However, the relationship between the recognition model and the development of the visual system is rarely considered. Research on the development of contrast sensitivity shows that human infants can only perceive low spatial frequency information from visual stimuli, but their acuity improves gradually with age. Also, the right hemisphere (RH) develops earlier than the left hemisphere (LH), and is dominant in infants. Here we show that these constraints, coupled with a desire on the part of the infant to individuate its caretakers and family, leads naturally to the right hemisphere bias for face processing. We propose a developmental model for face and object recognition using a modular neural network based on Dailey and Cottrell (1999). This neural network represents the two hemispheres using two modules, with a competitive relationship between them mediated by a gating mechanism. The strong RH and low spatial frequency bias for face recognition emerges naturally in the model from the interaction of the slow development of acuity and the early dominance of the right hemisphere. Remarkably, this strong asymmetry does not appear to hold for the other object categories that we tried.

Back to Table of Contents