CNS*1994

The Annual Computational Neuroscience Meeting

July 1994, Monterey, California

CNS*1994 Abstracts

Author: David Golomb* l, Xiao-Jing Wang 2, and John Rinzel 3

1, 3 Mathematical Research Branch
NIDDK, Bldg. 31, Rm. 4B-54,
National Institutes of Health,
Bethesda, MD 20892, U.S.A

2 Department of Mathematics
University of Pittsburgh
Pittsburgh, PA 15260

Title: PARTIAL AND FULL SYNCHRONY IN THE RE DURING SPINDLE OSCILLATIONS

Abstract: We address through modeling the hypothesis of Steriade and colleagues that the Reticular Thalamic Nucleus (RE) is the generator of spindle oscillation. Each model neuron exhibits post-inhibitory rebound and is coupled to all the other neurons via GABAA and GABAg inhibitory synapses. An isolated network of identical RE cells typically exhibits partial synchrony as it segregates itself into clusters; each neuron tends to burst only every two or more cycles of the population voltage. Blocking the GABAA synapses or increasing their reversal potential can produce a fully synchronized network. Excitation from thalamocortical cells results in synchronized oscillations that are robust to heterogeneity and noise.
Author: Norberto M. Grzywacz 1 and Pierre-Yves Burgi 2

The Smith-Kettlewell Eye Research Institute
2232 Webster Street, San Francisco, CA 94115

Title: A HEBBIAN RULE CONSISTENT WITH PHYSIOLOGY AND PRODUCING SHARP REFINEMENT OF ORGANIZATION

Abstract: Although the popular quadratic Hebbian rules lead to successful models of plasticity and learning, they are inconsistent with physiology. We developed a new rule based on the increase of presynaptic release following messenger feedback from the postsynaptic site. This rule was analyzed mathematically for the case of one and two synapses, and computationally for larger numbers of synapses. The results indicate its consistency with long-term potentiation and depression phenomenology, and better refinements of organization than obtained with quadratic rules. Hence, although mathematically less treatable, physiological rules might produce better organization and learning than quadratic rules. Supported by the NEI (EY08921) and SNFSR
Author: Boris S. Gutkin* 1 and Charles E. Smith 2

1 Neural Processes in Cognition Program
Department of Mathematics
301 Thackeray Hall University of Pittsburgh
Pittsburgh, PA 15209

2 Biomathematics Program Department of Statistics
Campus Box 8203
North Carolina State University
Raleigh, NC 27695

Title: ANALYSIS OF A STOCHASTIC MODEL OF SOMATOSENSORY INFORMATION PROCESSING.

Abstract: This work analyzes a neural network model of somatosensory information processing with additive noise. The response, as measured by metabolic activity of cortical tissue(2-DG), is examined. The model is a network of aggregated neural units, governed by a system of coupled p.d.e's. The coupling coefficients are identical for all units in the network, each with a self-excitation, lateral inhibition pattern. Both the linear and non-linear models are investigated. The results of computer experiments on a difference equation version of the model are presented. The simulations agree well with previous work, showing the formation of a steady state, spatially inhomogeneous pattern in response to constant stimulus for weak lateral inhibition. The dynamics of steady state formation and its qualitative characteristics are not affected by the addition of noise. Strong inhibition leads to unstable behavior of the network. Results of computer experiments serve as a basis for the theoretical analysis of the model dynamics. A stability criterion for the connection patterns in the linear and sigmoidal network is developed, thereby extending previous work.
Author: Anthony Harris* 1 and Steven Small 2

1 Intelligent Systems Program
University of Pittsburgh
325 Scaife Hall
Pittsburgh, Pa 15261

2 Dept. of Neurology
325 Scaife Hall
Pittsburgh, Pa 15261

Title: DEVELOPMENT OF CORTICAL SEGREGATES BY THE COMPETITIVE DISTRIBUTION OF ACTIVATION

Abstract: Recently, an alternative to the continuous, somatotopic representation of body surface has been proposed. In this proposal, cortical organization is described as a mosaic of discrete segregates which tile the cortical surface and encode spatiotemporal properties of the incoming sensory data. A segregate is defined by the following characteristics: all the maximal receptive fields (RFs) of the neurons comprising the segregate have a single skin region in common, termed the scgrcgatc RF ccntcr, and this center is non-overlapping with other segregates' centers. This paper presents a model of development of this organization via an adaptation of the competitive distribution of activation model. We propose a model to simulate segregate development via two learning mechanisms; the first being a Hebbian learning rule to facilitate cooperation of units within a segregate, and the second being a competitive learning rule which alters RF morphology eo discretize cortical organization.
Author: Rowshanak Hashemiyoon* and John K. Chapin

Department of Physiology and Biophysics
Hahnemann University
Phila., PA 19102

Title: CODING FOR GLOBAL PARAMETERS MAY BE CARRIED BY SYNCHRONISED SUBCORTICAL OSCILLATIONS

Abstract: Much attention has recently been given to fast oscillations in the visual system. We believe these dynamical states may carry sensory information. In order to investigate this possibility, we performed multi-site single neuron and unit cluster recordings in the subcortical visual system, specifically: the optic tract (OT), pretectal area, superficial gray layer of the superior colliculus (SGS), and the lateral geniculate (LGN) thalamic nucleus in awake and anaesthetised rats. Spontaneous, highly synchronised oscillations from single unit and unit cluster activity approximating 10, 20, and 40Hz were recorded in all of these areas, lower frequencies predominating in deeper anaesthesia. Autocorrelation histogram studies reflect very high coherence of the oscillations, in one case showing a temporal coherence for approximately 8 secs.

A frequency study was performed on awake vs. anaesthetised rats via chronic implantation of a multi-wire array in the SGS of the rat. Cross correlation histograms from the multi-unit, multisite recordings in the anaesthetised paradigm reveal synchronous oscillations approximating 20 Hz. Cross correlation histograms from the same multi-site recordings in the same rat under awake conditions show oscillatory frequency approximating 25 Hz. In general we noticed that while the animal was deeply anaesthetised, approximately 10 Hz frequencies were prevalent in younger rats, while 20 Hz frequencies prevailed in older ones. In all cases one, sometimes two, frequency bifurcations occurred as the effects of anaesthesia wore off. A spectrum analysis of the data shows that all of these frequencies (ie. 10, 20, and 40 Hz) may coexist within the same data set, exhibiting spontaneous bifurcations in upper harmonics of 10 Hz. Such frequency transitions are reminiscent of nonlinear dynamical systems which exhibit similar sudden bifurcations.

These oscillations were shown to be retinally derived by the following: 1) they were abolished by contralateral enucleation, 2) simultaneous recordings from similar locations across opposite colliculi revealed asynchronous activity between the two SGS, 3) multi-fiber recordings from the OT revealed oscillations
Author: Michael E. Hasselmo* 1 and Eric Schnell 2

1 Dept. of Psycholog
Harvard University
33 Kirkland St.,
Cambridge, MA 02138

2 Eric Schnell
Emmanuel College
Carnbridge University, Carnbridge,
CB2 3AP, England

Title: FEEDBACK REGULATION OF CHOLINERGIC MODULATION AND HIPPOCAMPAL MEMORY FUNCTION

Abstract: Acetylcholine may set the appropriate dynamics for learning within the hippocampal formation. A model of the feedback regulation of cholinergic modulation in hippocampal regions CA 1 and CA3 was developed, focusing on the putative autoassociative and heteroassociative memory function of synapses arising from CA3 pyramidal cells. Feedback regulation of cholinergic modulation allowed the network to respond to novel patterns with strong cholinergic modulation, allowing accurate learning, and to respond to familiar patterns with a decrease in cholinergic modulation, allowing recall. This function required that suppression of synaptic transmission in region CA3 and CA1 be stronger for synapses in stratum radiatum (arising from region CA3), in contrast to synapses in stratum lacunosum-moleculare (arising from the entorhinal cortex). Experiments in brain slice preparations of the hippocampus demonstrate this laminar selectivity.
Author: Hill* 1, Edwards 2, and Murphey

1 Neuroscience and Behavior Program
Department of Biology
Morrill Science Center (South)
University of Massachusetts at Amherst, MA 01003-35825

2 Department of Biology
Georgia State University
Atlanta, GA 30302-4010

Title: THE EFFECT OF NEURONAL GROWTH ON SYNAPTIC INTEGRATION

Abstract: The way in which the dimensions of neurons change during postembryonic development has important effects on their electrotonic structures. Theoretically, only one mode of growth can conserve the electrotonic structures of growing neurons without employing changes in membrane electrical properffes. If the dendritic diameters of a neuron increase as the square of the increase in dendritic lengths, then the neuron's electrotonic structure is conserved. We call this " isoelectrotonic growth. " In this study we compared the developmental changes in morphology of two identiaed invertebrate neurons with theoreticsl growth curves. We found that a cricket neuron, MGI, grows isoelectrotonically and thereby preserves its electrotonic properties. The crayfish neuron, LG, grows in a nearly isometric manner that results in an increase in its electrotonic length.
Author: E.C. Hymel*, C.R. Murphey, and B.N. Christensen

Department of Physiology & Biophysics
UTMB
Galveston, TX 77555-0641

Title: A MATHEMATICAL MODEL OF RETINAL PHOTORECEPTORS.

Abstract: Retinal photoreceptors transmit light stimuli via electrical signals generated by a change in a current, referred to here as the photocurrent. This voltage-sensitive photocurrent is gated by cGMP, and carries sodium and calcium ln a 3:1 ratio. cGMP levels are controlled by the opposing actions of guanylate cyclase and phosphodiesterase, which are in turn controlled by buffered calcium levels and photon absorption, respectively. We have used Mathematica on a SUN workstation to generate a system of nine coupled ordinary differential equations which reproduce changes in photocurrent in response to a light stimulus. In addition to the processes noted above, the actions of the Na+/K+ exchanger and the voltage- and concentration-dependent Na+/Ca++-K+ exchanger present in the photoreceptor inner and outer segments are considered. Previously published models in the literature have not incorporated all of these processes in a biologically relevant manner. The model is able to reproduce many figures available in the literature, such as photocurrents measured in response to or steady step changes in illumination and steady state levels of internal messengers. Supported by grant NEI01897 from NIH.
Author: M. A. P. Idiart*, J. Lisman, and L.F. Abbott

Department of Biology and Center for Complex Systems
Brandeis University
Waltham, MA 02254

Title: SHORT-TERM MEMORY AS A SINGLE CELL PHENOMENON.

Abstract: We present a model of short-term memory based on slowly-varying intrinsic properties of single neurons. Cholinergic modulation of hippocampal and cortical neurons can induce after-depolarization following spiking. When included in a model neuron, this effect leads to memory storage and a sample-and-hold function. In a network model, memories can be retained during periods of oscillation and rapidly reset through recurrent ihnibition.
Author: Dieter Jaeger, Erik De Schutter, and James M. Bower

California Institute of Technology
Div. of Biology 216-76
Pasadena CA 91125

Title: HOW ACTIVE DENDRITIC CONDUCTANCES AND SYNAPTIC INPUT CONTROL SOMATIC SPIKING IN THE PURKINJE CELL: A MODELING STUDY

Abstract: The level of depolarization at the soma of a Purkinje cell is strongly dependent on the inward and outward currents distributed over the large area of its dendritic tree. Here we analyze the relative contribution of synaptic and voltage-gated currents on the spiking behavior of the soma, using a previously constructed realistic Purkinje cell model. We find that dendritic Ca++ currents tend to depolarize the cell enough such that a net outward synaptic current is necessary to produce slow somatic spiking rates. Fast somatic firing is obtained with a net inward synaptic current, however, inhibition still needs to be present to prevent bursting behavior. Spiking variability is determined by the temporal structure of both inhibitory and excitatory input.
Author: R.V. Jensen*, G.M. Shepherd, and F. Zufall

*Department of Physics
Wesleyan Universtiy
Middletown, CT 06457

Department of Neurobiology
Yale University School of Medicine
New Haven, CT 06510

Title: KINETIC MODELS OF TRANSIENTION CHANNEL CURRENTS IN RESPONSE TO BRIEF PULSES OF AGONIST

Abstract: Patch clamp measurements of the response of cyclic-nucleotide-gated olfactory channels to brief (10-100 ms) pulses of cAMP provide new insights into the kinetics of sensory transduction.l The measured time-course of both the current onset and decay in response to a step-pulse oi agonist imposes new challenges for kinetic models. ln particular, our detailed analysis of the slow time-dependence of the transient olfactory channel currents induced by pulses of cAMP over a wide range of concentrations ( from 10-2 MÑ 10-6 M) suggests a novel, Monod-Wyman-Changeux type2, kinetic model in which the unbound closed state of the channel protein undergoes first a slow, k+l = 4 x 105M-1 s-1, binding step, then a slow conformational step, B=20s-1, to an open state, followed by a series of more rapid binding steps in the open state. A simplified numerical model of this kinetic scheme provides a good description of the onset, decay and saturated values of the olfactory channel currents and provides the basis for more detailed models of the full G-protein cascade describing olfactory transduction, from the receptor activation by odor molecules to channel opening. 3 A similar analysis is also applicable to recent measurements of the activation of slow NMDA currents by brief pulses of glutamate. 4 These empirically derived kinetic models serve in turn to motivate new experimental measurements to further elucidate the the non-equilibrium dynamics of ligand-gated ion channels.

1. F. Zufall, H.Natt, and S. Firestein, Proc. Natl. Acad. Sci. 90, 9335 (1993).
2. J. Monod, J. Wyman, and J.-P. Changeux, J. Mol. Biol. 12, 88 (1965).
3. S. Firestein, B. Darrow, G.M. Shepherd, Neuron 6, 825 (1991).
4. J.D. Clements and G.L. Westbrook, Neuron 7, 605 (1991).
Author: Leslie Kay

Graduate Group in Biophysics
University of California, Berkeley
129 LSA, UCB, Berkeley, Calif. 94720

Title: CORRELATED GAMMA, BETA, AND THETA OSCILLATIONS IN OLFACTORY AND LIMBIC BRAIN STRUCTURES DURING ODOR RECOGNITION

Abstract: Local field potential recordings from olfactory bulb (OB), prepyriform cortex (PPC), entorhinal cortex (EC), and Dentate Gyrus (DG) of awake, behaving rats show that gamma bursts are passed from OB through PPC, EC, and DG. Lower frequency gamma activity is also passed back from EC to OB as a possible handshaking signal. Beta activity, arising in the EC and passed to the PPC and OB and to the DG, precedes bursts occuring during odor recognition and may be a reafferent signal used to tune the OB to receive an expected stimulus. Theta activity is also correlated among these structures associated with a recognition task and during the interburst period.
Author: Rolf Kötter

Instutute of Otago
Dept. of Anatomy & Structural Biology
P. O. Box 913
Dunedin, New Zealand

Title: A MODEL OF BIOCHEMICAL INTERACTIONS BETWEEN GLUTAMATERGIC AND DOPAMINERGIC SIGNALS IN STRIATAL NEURONS

Abstract: An equilibrium model of glutamatergic and dopaminergic biochemical pathways in striatal neurons shows that low intracellular free calcium concentrations (below 1 u M) act synergistically with dopaminergic stimulation of adenylate cyclase. Concentrations exceeding 1u M lead to inhibition of adenylate cyclase, which is not overcome by dopaminergic stimulation. Further enzymatic pathways can be characterized as either calcium-dependent (PP2B), or as sensitive to calcium and dopamine (PKA, PPl). The regulation of phosphoproteins (DARPP, MAP2) demonstrates the complexity of crosstalk between transmitter-sensitive pathways. Such mechanisms may play a role in synaptic plasticity, corhcostriatal interplay, and reinforcement learning.
Author: Anders Lansner* and Erik Fransen

SANS - Studies of Artificial Neural Systems
Dept. of Numerical Analysis and Computing Science
Royal Institute of Technology
S-100 44 Stockholm, Sweden

Title: IMPROVING THE REALISM OF ATTRACTOR MODELS BY USING CORTICAL COLUMNS AS FUNCTIONAL UNITS

Abstract: Attractor network models of cortical associative memory functions have developed considerably over the past few years. Here we show that we can improve them further by using cortical minicolumns instead of single neurons as functional units. The connectivity of the model then becomes more realistic, since the dense and symmetric connectivity now may be sparse and strongly asymmetric at the cell-to-cell level, thus resembling more the real cortex. Simulations show that such a network, with model neurons of the Hodgkin-Huxley type, can operate as an associative memory in much the same way as previous, simpler network models.
Author: Christopher W. Lee 1* and Bruno A. Olshausen 2

1, 2 Washington University School of Medicine
Dept. of Anatomy and Neurobiology, Box 8108
660 S. Euclid Avenue
St. Louis, Missouri 63110

Title: A NONLINEAR HEBBIAN NETWORK THAT LEARNS TO DETECT DISPARITY IN RANDOM-DOT STEREOGRAMS

Abstract: We explore the use of a Hebb rule in a network of nonlinear neurons in order to evaluate their ability to extract higher-order statistics from an input stream. As a test of our methodology, we apply it to the problem of learning disparity in random-dot stereograms. Our model network consists of two layers, with nonlinearities in the first-layer units only. These non-linearities, inspired by those seen in neurobiology, allow the first layer to learn to transform the input layer's pixel-based representation into a representation based upon coupled pairs of left-right inputs. The second layer of the network then learns the disparities present by extracting patterns within this new representation via a standard competitive learning rule. The model demonstrates that a simple local learning rule can extract higher-order regularities which cannot be extracted by a purely linear model.
Author: David A. August and David B. Levy*

Department of Neurological Surgery
University of Virginia Health Sciences Center
Charlottesville, VA 22908

Title: INFORMATION MAINTENANCE BY RETINAL GANGLION CELL SPIKES

Abstract: A neuron may communicate its time-varying analog states with an interspike interval (ISI) code. Using a biophysical model of a retinal ganglion cell spike generator, we studied the performance of such a code. Our main result - a relatively small difference between an injected stimulus current and an estimated signal reconstructed from the resulting spike train - indicates tl1at the spike generator maintains information in individual ISIs. We conceptualize spikes as instantaneous "samples" of the ncuron's analog state. Specifically, ISIs are translated into samples of the analog somatic current by a monotonic function, f (ISI).
Author: K. L. Levy*1 and D. R. Kipke 2

1 Arizona State University
Bioengineering Program
Tempe, AZ 85287-6006

2 Daryl R. Kipke
Arizina State University
Bioengineering Program
Tempe, AZ 85287-6006

Title: USING MODELS TO DETERMINE THE RELATIVE IMPORTANCE OF INTRINSIC MEMBRANE PROPERTIES VERSUS SYNAPTIC ORGANIZATION IN OCTOPUS CELLS

Abstract: Compartmental models are used to determine the relative importance of intrinsic membrane properties versus synaptic organization in producing the octopus cell's onset response Three cells are modeled: a cell with a generic membrane and optimal synaptic structure, a cell with an octopus-like membrane and randomly placed synapses, and a cell with an octopus-like membrane and optimal synaptic structure The models' responses to tones are compared to experimental data in terms of the latency and height of the onset peak of the post-stimulus time histograms The results demonstrate that both properties play a role in determining the cell's response to tones Using Models to Determine the Relative Importance of Intrinsic Membrane Properties versus Synaptic Organization in Octopus Cells.
Author: Zhaoping Li* and Joseph J. Atick

The Rockefeller University
1230 York Avenue
New York, NY 10021

Title: UNDERSTANDING OCULAR DOMINANCE DEVELOPMENT FROM BINOCULAR INPUT STATISTICS

Abstract: We recently hypothesized that striate cells are concerned with, among other things, reducing binocular redundancy due to binocular correlations. In a multiscale representation, this leads to a distribution of monocular/binocular cells and disparity selective cells resembling what is observed in experiments. This theory is applied here to explain the different ocular dominance column formations resulting from different (normal or abnormal) developmental conditions such as strabismus and monocular deprivation. It is shown that abnormal ocular dominance column formations, such as unequal column sizes for the two eyes or increased monocularity, are consequences of decorrelation strategies applied to the altered binocular input statistics under abnormal developmental conditions.
Author: Jim-Shih Liaw 1, Michel Baudry 1, Gilbert A. Chauvet 1, and Theodore W. Bergera 1

1 Neural Informational, and Behavioral Sciences
University of Southern California
Los Angeles, CA 90089-2520

2 Institut de Biologie Theorique, Universite d'Angers
49100 Angers, France

Title: SIMULATION OF THE MOLECULAR MECHANISMS REGULATING TRANSMITTER RELEASE

Abstract: Simulations of presynaptic calcium dynamics and calcium interaction with molecules that play a critical role in transmitter release (e.g., synaptotagmin), in response to repetitive depolarization, led to the following conclusions: 1) the increase of single-liganded synaptotagmin after the first stimulation is critical for the facilitation of release seen with repeated stimulations, 2) a model with 2 calcium-sites for synaptotagmin generates results that best match experimental data, 3) facilitation of release persists even when calcium influx is progressively reduced, and 4) near equilibrated single-liganded synaptotagmins is responsible for the decrease in the facilitatory effect with higher extracellular calcium concentrations.
Author: Jim-Shih Liaw* and Micahel A. Arbib

Center for Neural Engineering
University of Southern California
Los Angeles, CA 90089-2520

Title: TIMING A LOOMING OBJECT

Abstract: We present a novel approach to extracting information related to the absolute speed of a looming stimulus. The method is based on two properties of looming detectors. They respond sooner to a large approaching object with lower firing rate. The curves obtained by integrating the activity of looming detectors in response to simuli of various sizes looming at the same speed intersect within a small range. After thresholding, the onset time of these integrator neurons form clusters, each corresponds to a looming speed. This mechanism offers a flexible way for timing interaction with looming objects by neural modulation.
Author: Hans Liljenström and Xiangbao Wu

SANS - Studies of Artificial Neural Systems
Dept. of Numerical Analysis and Computing Science
Royal Institute of Technology
S-lOO 44 Stockholm, Sweden

Title: NOISE AND CHAOS IN ASSOCIATIVE MEMORY

Abstract: Using a model of the olfactory cortex we demonstrate how noise and chaos can play a role in the complex dynamics involved in learning and memory. We show that noise can speed up neural information processing and reduce recall time in associative memory tasks. Noise can also induce transitions between different memory states. Oscillations and (transient) chaos, induced by noise or by an increase in neuronal excitability, is shown to improve system performance. We discuss how these effects can be under neuromodulatory control.
Author: Christine Linster* 1 and Claudine Masson 2

1 ESPCI, Laboratorie d'Electronique
10, Rue Vauquelin, 75005 Paris

2.Neurobiologie Comparee des Invertebres
INRA, CNRS (URA 1190)
911140 Bures sur Yvette, France

Title: ODOR PROCESSING IN THE HONEY BEE'S ANTENNAL LOBE GLOMERULI: MODELING SENSORY MEMORY

Abstract: Based on precise data collected from experiments conducted at different organizational levels (anatomy, electrophysiology and behavior), we propose an investigation of the possible mechanisms of sensory memory in the bee's olfactory system. In this goal, we have investigated the modulation and control between the two levels of olfactory information processing, the antennal lobe glomeruli and the mushroom bodies. We use simplified neurons, but realistic architecture. As a first conclusion, we postulate that sensory memory may be achieved by modulation of the inhibitory synaptic activity of antennal lobe neurons. These first results are in good accordance with behavioral and pharmacological data.
Author: Alexander V. Lukashin*, George L. Wilcox and Apostos P. Georgopoulos

Brain Sciences Center
Department of Veterans Affairs Medical Center
Minneapolis, MN 55417 and

Departments of Physiology
Pharmacology and Neurology
University of Minnesota Medical School
Minneapolis, MN 55455

Title: OVERLAPPING NEURAL NETWORKS FOR MULTIPLE MOTOR ENGRAMS

Abstract: We tested the hypothesis that learned movement trajectories of different shapes can be stored in, and generated by, largely overlapping neural networks. We trained a massively interconnected neural network to generate different shapes of internally stored, dynamically evolving movement trajectories using a general-purpose, core part, common to all networks, and a special-purpose part, specific for a particular trajectory. The percentage of trajectory-specific units needed to generate a certain trajectory was small (2- 5%). The core network alone could generate externally instructed trajectories but not internally stored ones, for which both the core and the trajectory-specific part were needed. These results suggest an efficient and effective mechanism for storing learned motor patterns.(Supported by the Office of Naval Research and the Minnesota Supercomputer Institute).
Author: Jeffrey W. McCandless* and Clifton M. Schor

Vision Science Group
School of Optometry
360 Minor Hall
University of California
Berkeley, CA 94720

Title: CROSS-COUPLING OF EYE POSITION SENSlTIVE NEURONS TO MAINTAIN BINOCULAR ALIGNMENT

Abstract: The oculomotor system maintains binocular eye alignment throughout life through an adaptive recalibration process. Empirical evidence has indicated there is a spatiotopic representation in the central nervous system to provide accurate binocular movements in situations where immediate error feedback is unavailable. We have investigated the mechanisms for this spatiotopic representation by modifying the default binocular alignment in one position of gaze, and measuring the associated eye responses in other positions. To explain the results, we have developed a computational model that cross-couples the activities of two classes of eye position sensitive neurons.
Author: K. Miller* 1 and P. Pennefather 2

AEON Software
975 Lewis Ave, Suite 8
Eugene, OR 97402

Nerve Cell and Synapse Group,
U of Toronto
Toronto, ON M5S 2S2

Title: A GENERAL SIMULATOR OF CELLULAR ELECTRO-CHEMICAL DYNAMICS

Abstract: Axon Engineer is a comprehensive, general and object-oriented simulator of cellular electrochemical dynamics underlying ionic signaling. It runs of the DOS platform The program makes the process of building a mathematical model transparent by presenting the investigator with a visual and intuitive environment for defining and linking objects critical to the generation of cellular ionic signals. The objects are: chemicals and buffers (both optionally diffusible), multi-state channels and transporters (with optional voltage and ligand dependencies), a broad range of stimuli and graphing windows. Each object has an extensive set of properties. The system of ordinary llifferential equations to describe membrane level events and partial differential equations for events mediated by buffered diffusion are solved using optimtied numerical methods (Le. Gear method for ODEs and the Crank-Nicholson method for PDE's). The strategies will be illustrated with a complex simulation of ionic signaling in neurons of bullfrog sympathetic ganglia.
Author: C.R. Murphey* 1 and L. E. Moore 2

1 Physiology & Biophysics
UTMB, Galveston TX 77555-0641

2 L.E. Moore. Physiology & Biophysics
UTMB, Galveston TX 77555-0641
and CNRS, U. of Rennes, 35042
RENNES Cedex, France

Title: A HYPERTEXT COMPARTMENTAL NEURONAL SIMULATOR

Abstract: A user interface in part determines the effectiveness of simulation software. The Hypertext Compartmental Neuronal Simulator is a software application for mathematical modeling single neurons or networks containing a small number of compartments (<30). The user interface consists of windows containing text mixed with interactive graphics with both point-and-click and command line input. The user interface and simulation protocols may be modified or extended dynamically at run-time using the application's embedded Tool Command Language (TCL) interpreter. Numerical efficiency is enhanced by use of a variable step-size, variable order integration method and compilation of model equations in C or Fortran. Use of de-facto standard software toolkits such as X windows and Tcl/Tk reduce the size and complexity of the application software itself to under 2000 lines of code. Source code for demonstration software solving the Hodgkin-Huxley equations is available on request. Requires Unix, X windows and the Tcl/Tk toolkit. Supported by DHHS-R01-MH45796.
Author: Sean D. Murphy* and Edward W. Kairiss

Yale University Department of Psychology and
the Center for Theoretical and Applied Neuroscience (CTAN)
Box 208205, New Haven CT 06520-8205

Title: THE STEM CELL: A COMPUTATIONAL MODEL OF BIOLOGICAL NEURONS USEFUL FOR SPATIO-TEMPORAL PATTERN RECOGNITION

Abstract: A biological neuron can be viewed as a match-filter that instantiates a mapping, M, from spatiotemporal (synaptic) events to one-dimensional temporal events (action potentials). An abstraction of a biological neuron called a Spatio-Temporal Event Mapping (STEM) cell has been designed to perform this mapping in a general way. We show that the STEM-cell is capable of learning M for a biophysical model of a neuron, and that it offers advantages over a biophysical model in terms of computational efficiency and analytical tractability.
Author: Farzan Nadim* 1, Ronald L. Calabrese 2, and Erik De Schutter 3

1 Dept. of Biology
Emory University
1510 Clifton Rd.
Atlanta, GA 30322

2 Dept. of Biology
Emory University
1510 Clifton Rd.
Atlanta, GA 30322

3 The Born Bunge Foundation
University of Antwerp
B2610 Antwerp, Belgium

Title: THE INTERPLAY OF INTRINSIC AND SYNAPTIC CURRENTS IN A HALF-CENTER OSCILLATOR

Abstract: The bursting oscillations underlying the heartbeat of the medicinal leech are generated by pairs of reciprocally inhibitory heart interneurons. We provide a conductance-based model including several intrinsic ionic currents for such a pairof neurons. The model also includes a calcium-dependent graded synaptic current, in addition to spike-mediated synaptic currents. The mechanism for oscillations is a mixture of "escape" and "release." We explore the interaction of the synaptic currents with the intrinsic currents underlying the bursting oscillations. This interaction determines whether the escape or the release aspect of the oscillation is dominant.
Author: Kiyohiko Nakamura

Graduate School of Information Systems
University of Electro-Communications
1-5-1 Chofugaoka, Chofu, Tokyo 182, Japan

Title: CORTICAL COMPETITIVE PROCESSING PRODUCES INTERFERENCE IN WORD-COLOR NAMING TASKS

Abstract: Neural processing of the millesecond time range is analyzed with the Stroop interference: the word "red" printed in blue delays naming its color "blue" by about 100 msec. The model cortex is the Broca area which is an array of cortical columns. They receive projections from the Wernick(word-processing) area and colorprocessing pathways. Model analysis shows (1) lateral inhibition between the columns restricts columns' firing to the most activated ones by surpressing the others before firing, which enables the cortical processing of a few hundred milliseconds, but (2) the inhibition delays the processing if the two projections deliver the inconguent stimuli.
Author: Michael G. Paulin 1 and Mark E. Nelson* 2

1 Dept. of Zoology
Univ. of Otago
Dunedin, New Zealand

2 Beckman Institute Univ. of Illinois
405 N. Mathews, Urbana, IL 61801

Title: MODELS OF ADAPTIVE SELF-NOISE CANCELLATION IN THE ELASMOBRANCH ELECTROSENSORY SYSTEM

Abstract: Marine elasmobranchs (sharks, skates, and rays) possess a specialized sensory system that is extremely sensitive to weak electric fields, with behavioral thresholds having been reported at field strengths as low as 5 nV/cm. This system is faced with a challenging signal processing problem in that elasmobranchs themselves generate relatively large bioelectric fields of the same general frequency characteristics as those they are trying to detect. We present and analyze alternative models of adaptive filtering in the dorsal octavolateral nucleus (DON), a region of the medulla that is responsible for cancelling out this self-generated background noise.
Author: Zhimin Ding and Mark E. Nelson*

*Beckman Institute
Univ. of lllinois
405 N. Mathews, Urbana, IL 61801

Title: SENSORY FEEDBACK AND CONTROL OF LEG-SUBSTRATE INTERACTIONS FOR AGILE LOCOMOTION IN INSECTS AND ROBOTS

Abstract: Most research on the neural control of multilegged locomotion has focused on the problem of pattern generation, namely on how to produce appropriately phased patterns of bursting activity in leg motor neurons so as to generate stable walking gaits. Fixed pattern generation alone, however, is not sufficient to achieve agile locomotion over irregular terrain. Agile locomotion requires the incorporation of sensory feedback signals which convey information about leg-substrate interactions. We present a biologically-plausible model of a single-leg controller for substratefinding and compliance control, discuss its relevance to insect neurobiology, and evaluate its performance in controlling a multijoint robotic leg.
Author: Jill M. Nicolaus* and Philip S. Ulinski

Dept. Organismal Biology and Anatomy
University of Chicago
1025 E. 57th St. Chicago, IL 60637

Title: FUNCTIONAL INTERACTIONS BETWEEN INWARDLY RECTIFYING CONDUCTANCES AND GABA-MEDIATED INHIBITION.

Abstract: We are studying interactions of inwardly rectifying (IR) conductances and inhibitory inputs in pyramidal cells in turtle visual cortex, using a two-neuron model. Kinetic parameters for fast and slow IR conductances were estimated from intracellular data. Trains of G A B A A -mediated IPSPs effectively activate fast IR conductances, while GABAB-mediated IPSPs activate slow IR conductances. Fitting activation kinetics to physiological data on a specific cell type is critical to modeling conductances that activate in appropriate voltage ranges. The fast and slow IR conductances may help to control levels of cortical excitability under different stimulus conditions. Supported by PHS grant EY 68352.
Author: Miguel A. L. Nicolelis* and John K. Chapin

Department of Physiology and Eiophysics
Hahnemann University
Philadelphia, PA 19102.

Title: NEURAL NETWORK MECHANISMS OF DYNAMIC AND DISTRIBUTED SENSORY PROCESSING: CHARACTERIZATION USING SIMULTANEOUS MULTI- SINGLE NEURON RECORDINGS

Abstract: To characterize sensory information processing at the network level we have developed techniques for chronic simultaneous recording of large numbers of single neurons within multiple connected areas of the brain of awake, behaving animals. Here, 48 single neurons were simultaneously recorded in the facial whisker representations of the trigeminal ganglion (Vg), principal (PrV) and spinal (SpV) trigeminal nuclei, ventral posteromedial thalamus (VPM) and primary somatosensory (Sl) cortex. Multivariate statistical analysis and 3D graphics techniques were used to analyze responses to discrete whisker stimulation through these different levels. Whereas at lower levels of this system whisker location was directly coded by discrete short latency responses, at higher levels this code was successively more broadly distributed over the neuronal population and over post-stimulus time. In addition, spontaneous 8-30 Hz oscillatory (mu-type) discharge was ubiquitous in these recordings in awake animals. Occuring mainly during immobile motor expectancy, these were uncorrelated with whisker movements, but interacted complexly with responses to whisker stimulation. To conclude, sensory processing in this system involves spatiotemporally distributed coding and is also subject to dynamical state transitions.
Author: Marius Usher 1 and Ernst Niebur*2

1 Department of Psychology
Carnegie-Mellon University
Pittsburgh, PA 15213

2 Computation and Neural Systems, 139-74
California Institute of Technology
Pasadena, CA 91125

Title: STIMULUS-DEPENDENT SYNCHRONIZATION IN A MODEL OF ORIENTATION-TUNED SPIKING NEURONS

Abstract: We investigate a model for neural activity in the primary visual cortex. Cells are organized in a two-dimensional layer with an embedded orientation preference map and are modeled as leaky integrate-and-fire neurons. The connectivity consists of local excitation, surround inhibition and long-range clustered connections, where the latter are only between cells with similar orientation preference. Due to the centersurround connectivity, the system generates irregular spike trains with broad inter-spike interval distribution as found in cortical data (Softky and Koch, 1993). Neighboring cells show a large arnount of synchronization while more distant cells show stimulus dependent synchronization similar to that observed experimentally (Gray and Singer, 1989) the cross-correlation peaks are larger for cells stimulated with the same orientation (vs. different orientation) even if they don't have the same orientation preference.
Author: Bruno A. Olshausenl 1 and Charles H. Anderson 2

1 Dept. of Anatomy and Neurobiology, Box 8108
Washington University School of Medicine
660 S. Euclid Avenue
St. Louis, MO 63110

2 Computation and Neural Systems Program
California Institute of Technology
Pasadena, California 91125

Title: A MODEL OF THE SPATIAL-FREQUENCY ORGANIZATION IN PRIMATE STRIATE CORTEX

Abstract: Despite much existing data on the spatial-frequency tuning properties of retinal and early cortical cells, there still exists no coherent model of the spatial-frequency organization in striate cortex and how it relates to visual function. Here, we present a revised Koenderink "stack model" for cortical spatial-frequency organization that fits quantitatively with existing neurobiological data. The model makes a number of predictions that may be tested in future experiments.
Author: Ralf Opara* 1 and Florentin Wrgtter 2

1 Institute of Physiology
Dept. of Neurophysiology
Ruhr-Universität Bochum
D-44780 Bochum, FRG

2 Institute of Physiology
Dept. of Neurophysiology
Ruhr-Universität Bochum
D-44780 BOCHUM, FRG

Title: A MULTI MODULAR HARDWARE-SOFTWARE DEVELOPMENT FOR VISUAL SCENE ANALYSIS

Abstract: We present a specialized multi-modular hardware-software development for real time analysis of visual scenes, where cortical structure is taken as a guideline. The goal was not to compete with fast supercomputers but rather to focus on the aspect of cortical task splitting and communication between the tasks. With this system, complex visual scenes are segregated in parallel into different very simple low-performance tasks (orientation, endstopping, velocity, disparity) while the complete scene analysis is strongly enhanced by the interactions between the tasks.