PRESENTATION TITLE AND DESCRIPTION

Rolf Kotter: Exploring structure-function coupling in cerebral cortical networks.

The cerebral cortex is a complicated network composed of many specific areas, which communicate through a dense network of axonal fibre projections. Crudely, connectivity defines the flow of information, whereas area characteristics help define what the components do with it and what the intrinsic computations are. Equipped with the global wiring diagram and with simple areas models we use multivariate statistics and computer simulations to explore the coupling of structure and function in cortical networks. Both matches and apparent mismatches between different empirical data sets provide clues to cortical mechanisms.

Back

Karl Friston: The nature of representations and the role of non-linear coupling.

I will compare and contrast two perspectives on how representations in the brain are constructed. These perspectives are a constructive perspective where representations are assembled in a bottom-up fashion through successive non-linear transformations of sensory inputs. The alternative view is provided by generative models where the sensory input is predicted or generated by high level representations of their underlying cause. These models imply fundamentally different neuronal architectures and respective roles for non-linear transformations and coupling among cortical areas. The generative model fits much more comfortably with known anatomy and physiology. In particular it relies upon backwards connections that emulate the non-linear mixing of real sensory causes to produce sensory input. If valid, the generative or predictive perspective means that the latencies of evoked neuronal responses, and their temporal dynamics, should show a paradoxical inversion where late components of early sensory responses are contingent on evoked transients in higher cortical levels. Furthermore, the generative framework suggests that non-linearity is a feature of backward as opposed to forward connections. These conclusions rest upon an integration of generative models in unsupervised learning and non-linear dynamics.

Back

Alex Martin: Semantic primitives and the representation of object concepts.

Object concepts are represented by networks of semantic primitives. These primitives are stored within the processing systems active when objects are perceived and manipulated (i. e., during learning). They include information about features and attributes such as an object's typical shape, motion, and use-associated motor movements. These representations are distributed in that: 1) They involve multiple regions (e.g., ventral occipitotemporal, lateral temporal, premotor cortices), and, 2) Within a region, different object categories elicit complex and overlapping patterns of activity. Thus, there are no category-specific" areas. These feature-based representations are "semantic" in that they are associated with the object concept, independent of stimulus format (pictures, words, mental images, etc.). This type of feature-based model can provide the combinatorial power needed to represent an infinite variety of object concepts, and a Foundation for more abstract conceptualizations (e.g., social interaction).

Back

Robert Desimone: Translation of cellular responses to mechanisms of attention.

Abstract from recent work

Back

Steve Bressler: The emergence of cognitive function from the operation of large-scale cortical networks.

Humans are able to rapidly and flexibly adapt to an almost infinite variety of changing environments in a manner that consistently integrates sensory, motor, and higher-order information domains. A growing body of evidence from a number of fields indicates that this dynamic adaptability derives from the ability of the cerebral cortex to repeatedly change the state of coordination among its constituent areas. It is now known that the evolution of cognitive state on a sub-second time scale is characterized by a progression of coordination states in the cortex, in which distributed sets of cortical areas become coordinated by phase synchronization in large-scale networks. Theoretical analysis suggests that the coordination of large-scale networks is accompanied by the mutual constraint of local activity patterns within the coordinated areas. The sum of constraints imposed by the network on the pattern of activity in any given area may dynamically create a variable local context for its information processing. A mutual pattern constraint mechanism in the cortex is proposed to satisfy large-scale processing demands and direct behavior to specified goals. The versatility of this mechanism may help to explain how the cortex overcomes the kind of processing rigidity exemplified by many artificial network models.

Back

Barry Horwitz: Neural modeling: The interplay between data-fitting and simulation approaches.

Modern cognitive neuroscience methods produce vast and exceedingly rich data sets at numerous spatial and temporal scales - from single unit neuronal recordings to functional brain imaging to neuropsychological investigations of behavior. The richness and complexity of these data preclude easy understanding and engender the need for equally rich computational approaches to data analysis, and equally important, data interpretation. Data-fitting models permit the extraction of conceptually defined parameters for the spatiotemporal scale appropriate for each data set used. Simulation neural models can be used to bridge spatiotemporal scales and thus relate parameters obtained from different types of data to one another. These points will be illustrated using functional brain imaging data.

Back

John Gabrieli: Refining memory theory with imaging.

Ideas about the functional neural architecture of human memory used to depend on happenstance rare and complicated lesions. Functional neuroimaging provides powerful new tools to systematically delineate the anatomy and function of memory systems of the human brain. Imaging research on memory, however, has often either confirmed interpretations of the consequences of lesions or provided unexpected but controversial findings. It is still a challenge; therefore, to think about how imaging research may truly inform theories of memory in a way that can sway fundamental views in the field.

Back

Martha Farah: What does the brain have to say about the general kind of thing that the mind is?

Cognitive neuroscientists are hard at work finding out whether we have specialized face processors, working memory maintenance-only buffers, plasticity over this or that time frame, etc. I'd like to call our attention to a more general set of questions about the mind. Like the specific ones mentioned here, they are empirical questions, for which neuroscience provides relevant data. But unlike the specific ones mentioned here, these cannot be answered by individual critical experiments, no matter how well designed. Instead, they can only be answered by considering the overall pattern of data that is accumulating in our field. These questions concern the general kind of thing the mind is. Is it sensible to call it computational, and what type of computational architecture is used? Is it optimized for flexible, general-purpose problem-solving, or specific problems? What is the balance of genetic preprogramming versus experience-driven organization in the development of its organization (be it general-purpose or specific mechanisms)? I hope others will suggest more questions along these lines. As someone who attended graduate school in the 70's, I can tell you how these questions would have been answered then. Living history in your midst! The answers are very different than today's, and as you'll see, the change was driven mainly by neuroscience.

Back

Giulio Tononi: Consciousness integrated and differentiated.

A useful way of identifying the neural basis of consciousness is to consider the kinds of neural processes that could account for its most fundamental properties. Two fundamental properties of consciousness are integration or unity, and differentiation or complexity. Integration is evident in that each conscious state is experienced as a whole and cannot be subdivided into independent components. Differentiation is evidenced by our ability to access, in a fraction of a second, any one out of countless numbers of conscious states. To understand these properties of consciousness and their neural substrates, a novel theory is developed that accounts at the same time for the integration and the differentiation of conscious experience. According to this theory, encapsulated in the dynamic core hypothesis, consciousness does not arise as a property of brain cells as such, but rather as a consequence of dynamic interactions of a continually changing functional cluster of nerve cells in the thalamus and cerebral cortex. The formulation of this theory has required the development of new theoretical concepts and measures, such as those for functional clustering and complexity, and the construction of large-scale computer models of brain function. A series of experiments using modern methods of magnetoencephalography has shown that neural correlates of conscious experience are consistent with the notion of a dynamic core and involve distributed brain areas which are different in different individuals.

Back

Mel Goodale: Visual Duplicity: Representation versus action.

Visual systems first evolved not to enable animals to see, but to provide distal sensory control of their movements. Vision as 'sight' is a relative newcomer on the evolutionary landscape, but its emergence has enabled animals to carry out complex cognitive operations on perceptual representations of the world. In the more ancient visuomotor systems, there is a basic isomorphism between visual input and motor output. In representational vision, there are many cognitive 'buffers' between input and output. Thus, the relationship between what is on the retina and the behavior of the organism cannot be understood without reference to other mental states. The implications of all of this for the organization of the visual pathways in the primate brain and the emergence of visual experience will be discussed.

Back

Scott Kelso: Self-organizing systems and nonlinear dynamics.

Dr. Kelso's homepage: http://www.ccs.fau.edu/~kelso/

Back

Randy McIntosh: Reflections on the day

The purpose of this last presentation is to give an overview (totally objective, of course) of the issues presented during the day and suggest how these issue may or may not impact on how we understand the relation between the brain and the mind.

Back