Principal Investigator: Moshe Bar
Studying Cortical Mechanisms of Visual Object Recognition and Memory by Combining Psychophysics, fMRI, and MEG
One of the most challenging questions in perception is how we process, represent, and recognize visual objects. My background in psychophysical and computational studies of these issues, together with recent methodological progress, have led me to the conclusion that imaging the activity that is elicited in the brain during performance of these tasks is crucial for understanding the underlying mechanisms. While imaging studies have already mapped the anatomy of some of the participating areas, little is known about the function and mechanisms involved. Empirical results from cognitive sciences, however, suggest intriguing hypotheses that could be studied directly using imaging.
Here I propose a novel combination of psychophysical experiments, functional magnetic resonance imaging (fMRl; high spatial resolution), and magnetoencephalography (MEG; high temporal resolution), to address aspects of visual object recognition and memory.
A central goal of the proposed research is to distinguish between two general views of visual information processing in the course of visual object recognition. In the more common view, information flow leading to object recognition is essentially bottom-up, from low to high level visual areas. A recent model (Ullman, 1995), on the other hand, suggests that flexible object recognition, and in particular the generalization to novel viewing conditions, requires the integration of bi-directional processing, top-down as well as bottom up. Psychophysical results I have obtained during my Ph.D. research, together with theoretical work that will be presented here, lend initial support for a bi-directional model.
Distinguishing between these two views requires (a) identifying the areas participating in object recognition, and by this also extending previous mapping studies, and (b) tracking the sequence of events in the visual cortex during recognition tasks in order to compare conflicting predictions of the two alternatives. The unique integration of imaging techniques with different strengths and weaknesses (spatial and temporal) is critical for exploring such processes which are extremely brief and localized.
In addition, studies of visual recognition naturally involves visual priming. This is supported by empirical findings regarding the role of priming in object recognition, as well as by theoretical consideration. Priming is a perceptual facilitation that stems from previous experience. Although it likely plays an important role in visual learning and memory, little is known about its neural basis. Therefore, a second goal of my proposal is to use imaging to (a) distinguish between different types of priming (e.g., semantic and visual); (b) test invariances of object recognition priming to transformations in size and orientation to learn more about the underlying representations; and (c) suggest a neural account of priming that will also include its role in object recognition and memory.
Extending my training to include imaging with fMRl and MEG will allow
me to address these questions, and would therefore be invaluable to my
future career as a cognitive neuroscientist. My choice of grantee institution,
together with the opportunity to work with, and learn from, leading figures
in fMRl (Dr. Tootell), cognitive neuroscience (Prof Schaeter), and MEG
(Prof. Halgren), has the potential of providing me with this training,
and to yield significant new information about the flow of information
in the cortex, visual memory, and object recognition.
