"23 Problems in Systems Neuroscience"
Leo van Hemmen and Terrence Sejnowski (Eds)
Oxford University Press
Preface
J. Leo van Hemmen
Physik Department
TU Munchen
85747 Garching bei Munchen
lvh@ph.tum.de
Terrence J. Sejnowski
Howard Hughes Medical Institute
Computational Neurobiology Laboratory
Salk Institute for Biological Studies
La Jolla, CA 92037
and
Division of Biological Science
University of California, San Diego
La Jolla, CA 92093
terry@salk.edu
How have brains evolved?
1. "Shall we even understand the fly's brain?"
Gilles Laurent
Division of Biology, 139-74
Caltech
1200 E. California Blvd.
Pasadena, CA 91125
laurentg@caltech.edu
I hope to illustrate two main things: the first is that small systems, and particularly small olfactory
systems, seem to use mechanisms and strategies that are not unique to them. The second is that
small systems are not at all that “simple”; this reinforces my view that we may be better off starting
with the modest goal of understanding flies first.
2. "Can we understand the action of brain in natural environments?"
Hermann Wagner
Bernhard Gaese
RWTH Aachen
Institut fur Biologie II
Lehrstuhl fur Zoologie/Tierphysiologie
Kopernikusstr. 16
52074 Aachen
wagner@bio2.rwth-aachen.de
We work mainly on reduced systems, but evolution has shaped
brains in a different way. To really understand brain function we
have to analyze it in the same environment in which brains evolved.
3. Hemisphere dominance of brain function - which functions are
lateralized and why?
Gunther Ehret
Abt. Neurobiologie
Universitat Ulm
89069 Ulm
guenter.ehret@biologie.uni-ulm.de
There are two main perspectives, a) an evolutionary one asking for
common origins and advantages of hemisphere specializations of
vertebrate, mainly mammalian, brains, and b) a proximate one
asking for genetic and physiological mechanisms responsible for the
realization of hemisphere specializations.
How is the cerebral cortex organized?
4. What is the function of the thalamus?
S. Murray Sherman
Dept. of Neurobiology
State University of New York
Stony Brook, NY 11794-5230
USA
ssherman@neurobio.sunysb.edu
The thalamus had long been thought to perform a boring,
machine-like relay of information to cortex, but recent evidence
suggests that it dynamically gates information flow and controls the
nature of what cortex receives in a state-dependent manner. Furthermore,
many areas of thalamus seem to perform a "higher-order" relay from one
cortical area to another, and indeed this trans-thalamic route may be
critical for much, perhaps all, cortico-cortical communication.
5. What is a neuronal map, how does it arise, and what is it good for?
J. Leo van Hemmen
Physik Department
TU Munchen
85747 Garching bei Munchen
lvh@ph.tum.de
Defining a `map' to be a neuronal representation of the outside
world, we are facing three closely related problems: what does
representation mean, how does it arise, and what is it good for?
The solution to this circle of problems is fundamental to
understanding how animals (and men) relate their own position
to that of a stimulus in the world surrounding them.
6. What is the role of top-down connections ?
Jean Bullier
Centre de Recherche Cerveau et Cognition
CNRS-UPS UMR 5549
Universite Paul Sabatier
133, route de Narbonne
F-31062 Toulouse Cedex
bullier@cerco.ups-tlse.fr
Until recently, all models of processing of sensory information by the
brain have exclusively considered a feedforward (or bottom-up) direction of
information processing, with a succession of filters operating in cascade
from the periphery to more "cognitive" regions of the brain.
This contrasts with the presence of an enormous network of feedback
connections that often outnumbers the feedforward connections and for which
little is known.
This essay will address a number of possible roles for these feedback
connections by asking further questions : are they involved in attention?,
are they involved in memory recall?, are they involved in global to local
interactions? are they involved in recalibrating lower order areas
following plastic changes in higher order areas? what is their role during
postnatal development? It may be naive to think
that a single role will be found for these top-down connections. Different
sets of connections operate at different scales in terms of cortical space
(some connections are focused, others are diffuse) and in terms of time
(some work at the millisecond, some at the second, depending on the type of
connections and receptors involved). Therefore it is likely that several of
the questions asked concerning the role of top down connections will be
answered in a positive manner and that one will have to think in terms of
many different types of top down influences on neuronal populations.
How Do Neurons Interact?
7. “How fast is neuronal signal transmission?”
Wulfram Gerstner
Ecole Polytechnique Federale de Lausanne
Mantra-LAMI
Informatique-J
CH-1015 Lausanne
Wulfram.Gerstner@di.epfl.ch
Starting from reaction time experiments (e.g., those of Simon Thorpe),
I will argue that the standard notion that neurons or
neural systems are `slow' is misleading.
My specific question is: how fast can a neuronal population
react to a new input? I will point out that for certain sets of parameters
the response can be practically immediate.
This means that integration delays are, in some cases,
negligible so that only axonal delays remain.
Implications for potential rapid feedback will be mentioned.
8. "What is the origin and functional properties of irregular activity?"
Dr. Carl van Vreeswijk
Carl.Van-Vreeswijk@biomedicale.univ-paris5.fr
Since the work of Adrian it is known that neuronal activity varies consistently
with the input/output of the cortical circuit in which the neuron is embedded.
This has led to the idea of a rate code. However the activity of the neurons
is very irregular, the typical coefficient of variation is about 1, so that
the neuron's rate can only be inferred poorly from the number of spikes fired.
It would seem that the rate is coded inefficiently, casting doubt on the rate
code assumption.
In this chapter I will discuss how irregular activity could arise in highly
interconnected cortical circuits, show that such a state has important
funtional advantages and the the disadvantages of a high CV are much smaller
than commonly assumed. This suggests that the doubt about the rate code is
based on a misunderstanding.
9. “Are single cortical neurons independent or are they obedient members of a huge orchestra?”
Amiram Grinvald
Tal Kenet
Amos Arieli
Misha Tsodyks
Dept. of Neurobiology
The Weizmann Institute of Science
P.O. Box 26
Rehovot 76100
Israel
bngrinva@weizmann.weizmann.ac.il
10. “What is the other 85% of V1 doing?”
Bruno A. Olshausen
Redwood Neuroscience Institute
and Center for Neuroscience, UC Davis
David J. Field
Department of Psychology
Cornell University
We pose the following challenge: that despite four decades
of research characterizing the response properties of V1 neurons, we still do
not have a clear picture of how V1 really operates—i.e., how a population of
V1 neurons represent natural scenes under realistic viewing conditions. We
identify five problems with the current view that stem largely from biases in
the design and execution of experiments, in addition to the contributions of
non-linearities in the cortex that are not well understood. Our purpose is to
open the window to new theories, a number of which we describe along with
some proposals for testing them.
What can brains compute?
11. “What is the Formal Computation in Early Vision?”
Steven W. Zucker
Dept. of Computer Science
Yale University
P.O. Box 208285
New Haven, CT 06520
USA
zucker@cs.yale.edu
A model for early visual computations, based on a model for visual
cortex in primates, will serve as an introduction to the question:
"which formal computation underlies neural computation?". We will
show how linear complementarity problems, quadratic programs,
and polymatrix games arise. We will suggest spiking neuron
methods for computing them.
12. “How do neurons compute?”
Catherine Carr
D Soares
S Parameshwaran
S Kalluri
Department of Biology
University of Maryland
College Park, MD 20742-4415
USA
J Simon
Institute for Systems Research
University of Maryland
College Park MD 20742 USA
T Perney
Center for Molecular & Behavioral Neuroscience
Rutgers University, Newark NJ 07012 USA
cc117@umail.umd.edu
An important question in neurobiology remains that of how neurons
compute. Attendant on this is the question of whether neurons are
designed for specific computations. We can show that some neurons are
clearly designed for particular computations, presumably by natural
selection. Does this mean that all neurons are designed for particular
computations? Some neurons may have more general responses. Other neuron
types may change their response types under the action of some
modulator, but these neurons can be regarded as the case of a neuron
designed for several computations, rather than for some general
input-output function.
To make the case that neurons in the vertebrate CNS can be designed for
particular tasks, we will use the example of temporal coding cells in
the auditory system because their function is well known and thus we can
tie physiological and morphological observations to function. We will
argue that it is important to understand the functions of single
neurons.
13. "How can neural systems compute in the time domain?"
Andreas V.M. Hertz
Innovationskolleg Theoretische Biologie
Humboldt-Universit”t zu Berlin
Invalidenstr. 43
10115 Berlin
a.herz@biologie.hu-berlin.de
Given the large variety and intriguing temporal complexity of many natural pattern sequences,
sophisticated neural representations are likely to have been invented during the course of evolution.
No simple, universal answer is therefore to be expected to the question posed above. Progress in the
understanding of neural coding in the time domain may nevertheless be achieved if one
concentrates on a more specific problem: “Which types of representations best support flexible and
robust computations of temporal relations?”
14. “How common are neural codes?”
David McAlpine
Alan R. Palmer
Department of Physiology
University College London
Gower Street
London
WC1E 6BT
d.mcalpine@ucl.ac.uk
The ability to localize sounds in space is a fundamental attribute of
the way that humans and animals perceive their environment. It has
obvious survival value in enabling them to determine where their prey or
a predator is. However, unlike in vision or somato-sensation, the
sensory end-organ of audition, the cochlea, contains no explicit
representation of the spatial environment. Auditory space must therefore
be computed indirectly from non-spatial cues. This chapter deals with
the problems that the auditory system must overcome in achieving this
task, and underlines the importance of understanding the specific
limitations imposed by the stimulus sub-domain to which different
species are responsive. Seemingly identical tasks that appear to be
addressed in an identical fashion across species may require uniquely
different neural codes in order to permit the extraction of useful
sensory information.
15. “How does the hearing system perform auditory scene analysis?”
Georg Klump
Institut fur Zoologie
TU Munchen
85350 Freising-Weihenstephan
Georg.Klump@bio.tum.de
Bregman (1990) identified a number of processes that provide for auditory
object formation and auditory scene analysis. Many of these processes do not
require cognitive abilities and can be found in non-human vertebrates.
The chapter will review evidence for mechanisms supporting auditory object
formation in animals and elucidate how these can be used for effectively
analyzing natural soundscapes.
16. “How does our visual system achieve shift and size invariance?”
Laurenz Wiskott
Innovationskolleg Theoretische Biologie
Humboldt-Universit"at zu Berlin
Invalidenstrasse 43
10115 Berlin
wiskott@itb.biologie.hu-berlin.de
After a short introduction about psychophysical and neurophysiological
aspects of shift and size invariant recognition in the visual system an
overview is given over the main computational models. Then, a few open
questions and possible approaches to solving them are discussed. Shift and
size invariance are just two examples of the more fundamental question of
how the brain builds and learns to build invariant representations in
general.
Organization of cognitive systems.
17. “What is reflected in sensory neocortical activity: External stimuli or
what the cortex does with them?”
Henning Scheich
Frank W. Ohl
Holger Schulze
Andreas Hess
Andre Brechmann
Leibniz-Institut fur Neurobiologie
Brenneckestrasse 6
Postfach 1860
39008 Magdeburg
scheich@ifn-magdeburg.de
Cortical activity is manifest in spatio-temporal patterns. These
patterns are usually conceived as representations of stimuli in maps.
Various lines of evidence suggest however that these map-based patterns
also depend on the specific purpose served by the cortical processing.
Therefore a broader concept of the function of sensory neocortex seems
to be required.
18. "To what extent does perception depend upon action?"
Giacomo Rizzolatti
Vittorio Gallese
Institute of Human Physiology
University of Parma, Italy
fisioum@synet.symbolic.pr.it
We discuss the relation between action and perception as it emerges from
neurophysiological data. We propose that both action perception and space perception
derive from a preceding motor knowledge based on self-generated actions.
19. "What are the projective fields of cortical neurons?”
Terrence J. Sejnowski
Howard Hughes Medical Institute
Computational Neurobiology Laboratory
Salk Institute for Biological Studies
La Jolla, CA 92037
and
Division of Biological Science
University of California, San Diego
La Jolla, CA 92093
terry@salk.edu
The inputs to a neuron can be explored by carefully choosing the sensory stimulus, but the
receptive field properties only provide part of the information needed to characterize the neuron. In
addition, the impact of the neuron on other neurons needs to be assessed, called the projective field
of the neuron. A broad experimental program is proposed here that could reveal the projective
fields of cortical neurons, which may provide the missing information needed to unlock the
mysteries of the cerebral cortex.
20. “To what extent is the brain reconfigurable?”
John Reynolds
The Salk Institute
10010 N. Torrey Pines Road | P.O. Box 85800
La Jolla, CA 92037 | San Diego, CA 92186-5800
USA
reynolds@salk.edu
Psychophysical and neurophysiological observations raise the
question of how the different features of an object, encoded by neurons in many different areas of
visual cortex, become linked together so that attending to and discriminating one feature of an
object causes other features of that object to be selected, while suppressing processing of the
features that make up competing objects.
21. "Where are the switches on this thing?"
Laurence Abbott
Volen Center
Brandeis University
Waltham, MA 02454-9110
http://play.ccs.brandeis.edu/abbott
abbott@brandeis.edu
We know a great deal about how information relevant to sensory
stimuli and motor responses is encoded in the brain. But how do we
channel sensory information through the brain and activate the
pathways necessary to generated appropriate responses. As well as
having banks of selective representational units, the brain must
involve complex switching circuitry based on biophysical mechanisms
we do not yet understand. Hence the questions: where are the
switches that channel information flow in the brain and how do they
operate?
22 "Do qualia, metaphor, language and abstract thought emerge from synethesia?"
VS Ramachandran
Department of Psychology
Univerersity of California, San Diego
La Jolla, CA 92093
vramacha@ucsd.edu
Edward M. Hubbard, MA
Brain and Perception Laboratory
University of California, San Diego
9500 Gilman Dr. 0109
La Jolla, CA 92093-0109
and
SNL-B
Salk Institute for Biological Studies
10010 North Torrey Pines Road
La Jolla, California 92037-1099
edhubbard@psy.ucsd.edu
In this essay, we will consider synesthesia. First, we will show that synesthesia is a genuine sensory
effect. Second, we will suggest what the underlying neural mechanisms might be. Third, we will
point out that far from being just an oddity, synesthesia might help illuminate some of the most
puzzling aspects of the mind such as the evolution of metaphor, language and even abstract thought
in humans. In addition, we will discuss the philosophical riddle of qualia and point out how
studying synesthesia and other sensory phenomena (such as the filling-in of scotomas and blind
spots) can provide some hints about the evolution, functional significance and neural correlates of
qualia.
23. “What are the neural correlates of consciousness?”
Francis Crick
The Salk Institute
10010 North Torrey Pines Road
La Jolla, CA 92037
Christof Koch
Division of Biology, 139-74
Caltech
1200 E. California Blvd.
Pasadena, CA 91125
koch@klab.caltech.edu
There are a host of hard problems in neuroscience, but consciousness is certainly the most
mysterious of them, and so in some sense the hardest. How is one to explain “qualia”—the redness
of red or the painfulness of pain—in terms of known science? So far no one has put forward any
concrete hypothesis that sounds even remotely plausible. Our strategy is to leave the core of the
problem on one side for the time being, and instead try first to discover the Neural Correlates of
Consciousness, now widely called the NCC.