MEG correlates of visual perception : the M170 response to degraded face and building images

Abstract

Introduction: Knowing how the brain processes faces is valuable to scientists, clinicians, and industry How does seeing a face eventually lead to recognition: e.g. "that's Jill." Many reports (including this one) are interested in this question, due to its intrinsic scientific appeal and application-based benefits. From a science perspective, there is a natural desire to know how the brain does something it is good at: recognize faces. Humans are extremely adept at discriminating between faces, and there is evidence for specialized face-processing brain machinery (Kanwisher 2000). Additionally, knowing how the brain recognizes faces provides insight into how the brain generally works. It lays down some of the operations the brain is able and likely to perform. From a clinical perspective, knowing the steps the brain takes to recognize faces provides an opportunity to ask, at which step do things go wrong for people with face processing disorders, such as prosopagnosia and autism? Pinpointing a disorder's defective step illuminates the extent of the disorder, and provides a straightforward test for diagnosis (this step is fine, but the next step is affected). If this test can be done with non-invasive measurement on young children, it will provide an early diagnosis leading to preemptive treatment. From an industry perspective, computerized face identification systems are wanted for airports and casinos to catch unwanted trespassers. Computer scientists have been working diligently to produce such a product, but the current systems perform far worse than normal people (Willing 2003; Zhao et al. 2003). A possible solution is to create a computer system that copies the human brain. Theoretically, a computer using the same steps a brain uses to recognize faces will be just as good as a person. This provides strong motivation for discovering the "neural algorithm" underlying face processing. The M1 70 represents at least one step in the face processing neural mechanism Previous research has uncovered the basic route through the brain subserving face recognition (for a review see Haxby et al. 2002). Major structures have been identified, and their functions have been proposed. However, further refinement of the pathway is needed. Namely, the major structures need to be better characterized, and non-major structures need to be identified. Researchers have established that one of the important brain areas for face recognition, the lateral fusiform gyrus (AKA the fusiform face area, FFA), produces a neural signal called the M170. The M170 is a signal released from the brain about 170 milliseconds after viewing an image, and is measured with magnetoencephalography (MEG). The remainder of this report will focus on what neural step the M170 may reflect. It will include a comprehensive review of past M170 research (preceded by some important facts about MEG), and a new experiment addressing two pivotal questions: 1) Why does the M170 respond to buildings if it is face-selective? 2) Does the M170 reflect categorization (that's a face or not) or identification (that's Jill's face)?