Purchase this article with an account.
Ezra Y. Rosen, Maximilian Riesenhuber; A simple and not so special model of face processing in cortex. Journal of Vision 2004;4(8):902. doi: 10.1167/4.8.902.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Understanding how objects are represented in cortex is one of the key questions in visual neuroscience. While data from physiology support a generic shape-based model of visual processing in which the recognition and representation of all objects is based on the same computational mechanisms, data from behavior as well as fMRI have been interpreted to suggest that the representation and recognition of some object classes is based on special computational mechanisms. In particular, the Face Inversion Effect (FIE) has been widely interpreted as a result of face-specific processing, involving not just shape-based (featural) but also configural information. Similarly, the existence of a face-specific locus of activation in fMRI studies has been cited as support for face-specific processing. However, while the data are trivially compatible with a model that allows for face-specific processing, it has not been shown that a simple shape-based model cannot equally explain the observed effects without having to postulate object-specific computational mechanisms. Using our recent shape-based model of object recognition in cortex, we performed quantitative simulations to explore this hypothesis. In particular, we simulated a set of model face units. We find that the units' fine tuning to upright faces can produce a sizeable FIE. We also modeled fMRI data on object-specific cortical activation by averaging activity over the population of face units. We find that average activations are very similar for upright and inverted faces (while being significantly lower for non-face objects), as observed experimentally (Haxby et al., 1999; Kanwisher et al., 1998). Finally, the simulation results offer a possible explanation why some fMRI studies in normals and patients find a dissociation between FFA activation and face discrimination performance, with interesting experimental predictions. Thus, face-specific mechanisms might play a lesser role in human vision than thought previously.
McDonnel-Pew Award in Cognitive Neuroscience
This PDF is available to Subscribers Only