Sensory processing, learning and cognition

Liste des participants

Mariana Alonso, Richard Axel (organiser), Alan Carleton, Jean-Pierre Changeux, Antoine de Chevigny, Tanguy Chouard, Jean-Marc Edeline, Stuart Firestein, Rainer Friedrich, Marie-Madeleine Gabellec, Gilles Gheusi, Charles Gilbert, Samuel Lagier, John Lisman, Pierre-Marie Lledo (Organiser), Zachary F. Mainen, Pascal Martin, Kensaku Mori, Kerren Murray, Carl Petersen, Ivan Rodriguez, Edmund Rolls, Armen Saghatelyan, Gordon Shepherd (Organiser), Jean-Didier Vincent, Cécile Viollet


Sensory processing, learning and cognition
Pierre-Marie Lledo
20-26 Sept 2004

Two issues in neural coding : multiplexing by theta/gamma oscillations and burst duration codes
John Lisman

In my talk, I will discuss two major ideas regarding neural coding. The first relates to concepts introduced by Von der Malsburg and Singer. They proposed that an item is represented by a group of cells that fire synchronously and that groups representing different items fire asynchronously. However, the definition of synchrony and the temporal separation of different groups have been unclear. Based on psychophysical findings, most notably from the Sternberg task, Lisman and Idiart (Lisman and Idiart, 1995) made the following specific proposals about the how multiple items are coded in working memory:

  1. Working memory in cortex is oscillatory, repeating on each cycle of a theta frequency oscillation (5-8 Hz).
  2. A theta cycle is divided into discrete subcycles by a nested gamma frequency oscillation (~30-80 Hz).
  3. The “synchronous” activity that represents a given memory is defined as cells firing within the same gamma cycle (i.e. within ~15msec).
  4. Sequential stored memories have an ordered representation in sequential gamma subcycles, thereby implementing a discrete phase code with an absolute reference.

Initial evidence for this type of coding came from the study of the phase precession of rat hippocampal place cells. More recently, intracranial recording has been used to study theta oscillations in human cortex during working memory tasks. These experiments reveal sites at which theta oscillations are gated on and off at the onset and offset of the Sternberg task (Raghavachari et al., 2001). Interestingly these oscillations are most frequently found in occipital cortex. This may be related to the observation of theta oscillations in many regions of sensory cortex in the rat (Macrides et al., 1982; Kleinfeld et al., 2002; McLin et al., 2003). Finally, the recent results of Rainer (Lee et al., 2003) in monkey V4 provide evidence for single unit activity that oscillates at theta frequency during a memory task. A critical remaining test of phase coding is to determine whether different memories are active at different phases of theta. Taken together, the existing findings support the idea that theta oscillations are a central organizing rhythm for cortical function, perhaps acting as a temporal organizer for separating and ordering different units of information.
A second aspect of neural codes relates to the importance of the bursts of high frequency (100-500Hz) spikes. These bursts are generated by intrinisic conductances in many types of neurons. In evaluating single unit data, it remains common practice to count all spikes in a period of ~100msec and determine the “rate”, ignoring whether spikes occur in bursts. Our analysis suggests that bursts may be very important and, through variation in burst duration, form a special type of neural code. In support of such a burst duration code:

  1. Bursts are triggered by special features of the synaptic input (e.g .by slope or preceding hyperpolarization). The particular feature depends on the biophysical mechanism of burst generation in a given cell type. (Kepecs et al., 2002)
  2. The number of spikes per burst is graded with the input feature. The strongest evidence for a graded code comes from the analysis of LGN cells during white noise stimulation. We find that the number of spikes per burst is systematically related to input variables. (Kepecs and Lisman, 2003)
  3. Burst duration can be decoded. For instance, facilitating synapses can be considered as burst detectors. The output of such synapses is dependent on the number of spikes in the burst.

These results suggest that bursts enable single neurons to communicate graded signals in a short time (~20msec), as required for high-speed brain computations.

Kepecs A, Lisman J (2003) Information encoding and computation with spikes and bursts. Network 14:103-118.
Kepecs A, Wang XJ, Lisman J (2002) Bursting neurons signal input slope. J Neurosci 22:9053-9062.
Kleinfeld D, Sachdev RN, Merchant LM, Jarvis MR, Ebner FF (2002) Adaptive filtering of vibrissa input in motor cortex of rat. Neuron 34:1021-1034.
Lee H, Simpson GV, Logothetis NK, Rainer G (2003) Phase Locking of Single Neuron Activity to Theta Oscillations During Working Memory in Monkey Extrastriate Visual Cortex. Society for Neuroscience Abstract.
Lisman JE, Idiart MA (1995) Storage of 7 +/- 2 short-term memories in oscillatory subcycles. Science 267:1512-1515.
Macrides F, Eichenbaum HB, Forbes WB (1982) Temporal relationship between sniffing and the limbic theta rhythm during odor discrimination reversal learning. J Neurosci 2:1705-1717.
McLin DE, 3rd, Miasnikov AA, Weinberger NM (2003) CS-specific gamma, theta, and alpha EEG activity detected in stimulus generalization following induction of behavioral memory by stimulation of the nucleus basalis. Neurobiol Learn Mem 79:152-176.
Raghavachari S, Kahana MJ, Rizzuto DS, Caplan JB, Kirschen MP, Bourgeois B, Madsen JR, Lisman JE (2001) Gating of human theta oscillations by a working memory task. J Neurosci 21:3175-3183.

Odor maps in the brain
Kensaku Mori,
Dept. of Physiol., Grad. Sch. of Med., Univ. of Tokyo, Tokyo 113-0033, Japan

The glomerular sheet of the mammalian olfactory bulb (OB) forms two symmetric maps of odorant receptors. We used the method of intrinsic signal imaging and systematic panels of stimulus odorants to examine the molecular receptive range (MRR) property of individual glomeruli in a substantial part of the dorsal and lateral surfaces of the rat OB. The results showed that in both dorsal and lateral surfaces, glomeruli with similar MRR property gathered in close proximity and formed molecular-feature clusters and subclusters. Analysis of the molecular-features/determinants effective in activating individual glomeruli suggests a systematic, gradual and multidimensional change in the represented molecular-features according to the position of clusters and subclusters in the odor maps. The relationship between the molecular-feature clusters of glomeruli and the subjectively perceived ‘odor’ of odorants that have the molecular-features/determinants suggests that the clusters participate in the representation of odor quality in the OB.
Although the shape of the clusters varied across different OBs, the clusters of glomeruli were spatially arranged at stereotypical positions and in relation to the zonal organization of the OB. Characteristic molecular-features of glomeruli in most clusters within zone 1 (dorsal zone) were relatively polar parts of odorants including polar functional group(s). In contrast, characteristic molecular-features of glomeruli in several clusters in zones 2-4 (ventrolateral zones) were hydrocarbon parts or relatively non-polar parts of odorants. These results imply a clear difference in the manner of recognizing odorant molecular features/determinants between ORs represented in zone 1 and ORs represented in zones 2-4.
Extracellular single-unit recordings from principal neurons in the specific cluster suggest that mitral and middle tufted cells differ in the manner of decoding the odor maps. Mitral cells may detect the contrast of activity between their own glomerulus and neighboring glomeruli, while middle tufted cells might be able to detect the presence of specific odorants.
Finally, I would like to discuss the state-dependent gating of signal flow from the odor maps of the OB to the olfactory cortex.

 Internal Representations of the Olfactory World
Jing W. Wang, Ph.D., Allan M. Wong, and Richard Axel, M.D.,

Howard Hughes Medical Institute and Department of Biochemistry and Molecular Biophysics, College of Physicians and Surgeons, Columbia University, New York, NY 10032

Olfactory perception requires the recognition of a vast repertoire of odorants in the periphery and central neural mechanisms that allow the discrimination of odors.  The organization of the peripheral olfactory system appears remarkably similar in fruit flies and mammals. The convergence of like axons into discrete glomerular structures provides an anatomic map in the antennal lobe.  How does the anatomic map translate into a functional map? We have developed a sensitive imaging system in the Drosophila brain that couples two-photon microscopy with the specific expression of the calcium-sensitive fluorescent protein, G-CaMP, to examine neural activity.  At natural odor concentrations, each odor elicits a distinct and sparse pattern of activity that is conserved in different flies.  We have combined Ca2+ imaging with electrical recordings to demonstrate the faithful propagation of the glomerular map by projection neurons that innervate the protocerebrum.  The quality of an odor may therefore be reflected by defined spatial patterns of activity, first in the antennal lobe and ultimately in higher olfactory centers.  We have identified a spatially invariant sensory map in the fly protocerebrum that is divergent and no longer exhibits the insular segregation of like axons observed in the antennal lobe.  This organization provides the opportunity for the integration of multiple glomerular inputs by hierarchical cell assemblies in the protocerebrum.

Uncommon scents: a genetic basis for pheromone perception in mammals
Ivan Rodriguez (Switzerland)

In most mammals, pheromone perception mediates intraspecies interactions related to reproduction, such as mate recognition, intermale aggressive behaviors, or exchanges between females and their offspring. Many of these interactions are mediated by the vomeronasal system, via sensory neurons expressing specific pheromone receptors. A major such receptor family is represented by the V1r gene superfamily, a rapidly evolving seven transmembrane receptor group which is highly divergent across mammalian species. Each vomeronasal sensory neuron expresses randomly a single V1r gene, from a single parental allele, providing thus very narrow responsive characteristics to each sensory neuron. Axonal projections of neurons expressing a given V1r receptor, converge to a few spatially restricted glomeruli in the accessory olfactory bulb, thus forming specific topographic maps corresponding to each V1r. This first level of organization is followed by a second step of information convergence, at the level of secondary neurons.
Neural Mechanisms of Perceptual Learning
Charles Gilbert, Aniruddha Das, Wu Li, Valentin Piech, Mariano Sigman, Dan Stettler

The Rockefeller University, New York, NY

The representation of form along the visual pathway is usually thought of in feedforward terms, with primary visual cortex (V1) representing simple stimuli such as oriented line segments, and higher order areas representing complex shapes by linking the simpler components. Cortical circuits enable neurons to integrate information over relatively large
parts of visual space, and to be selective for complex visual stimulus configurations.  The relationship between these circuits, the higher order receptive field properties of neurons in V1 and the geometry of natural scenes suggests that visual experience early in life encodes information about the structure of the natural world.  The plasticity of these circuits, seen in axonal sprouting and synaptogenesis accompanying functional recovery following retinal lesions, suggests a mechanism that may be general to experience dependent cortical plasticity.  Visual cortical plasticity in the normal adult brain allows one to discriminate learned shapes through a process known as perceptual learning.  Several characteristics of perceptual learning, however, suggest the involvement of early visual cortical areas in representing more complex shapes. The fact that perceptual learning is specific for location and orientation suggests the involvement of early stages in visual processing. In visual search tasks, learned shapes become recognized rapidly and in parallel with numerous distractors, a process thought to involve retinotopic maps.  Trained shapes take on a “pop-out” quality, whereas untrained shapes require a more effortful, serial search.  Following training, brain activity associated with trained shapes relative to the untrained shapes shows 1) an increased activity level and correlation with behavior in V1 and V2, 2) a decrease in activation of the lateral occipital cortex (LO), 3) a decrease in the dorsal attentional network including posterior parietal (PP), and premotor regions. These findings suggest that while before training LO integrates information from V1/V2 in a process coordinated by the attentional network, after training it becomes less involved as the shapes become more fully represented in the earlier, retinotopic visual cortical areas.  At the level of individual V1 neurons in animals trained on a shape discrimination task, we see a central role of top-down influences in the representation of trained information.  With training, V1 neurons adopt novel functional properties related to the attributes of the trained shapes.  These properties are only present, however, when the animal performs the trained task, and neurons respond very differently to an identical visual stimulus when the animal performs a different task.  The top-down influences were seen from the very beginning and throughout the entire time course of the neural responses. Our findings suggest that the output from V1 reflects both sensory and behavioural context, which may reflect an interaction between feedback pathways to V1 and local circuits within V1.  These contextual influences are subject to learning, leading to plasticity of function and circuits, even within V1, that extends throughout life.

Dynamic neural computations in the olfactory bulb studied by complementary methods
Rainer Friedrich

Neuronal circuits in the olfactory bulb process distributed, stimulus-specific input activity patterns across the input channels, the olfactory glomeruli. Glomerular activity patterns evoked by chemically related stimuli are similar. Activity patterns evoked by the same stimuli across the output neurons, the mitral cells, are initially also similar, but subsequently become decorrelated. Hence, neuronal circuits dynamically re-format odor representations so as to make their discrimination more reliable. Concurrently, stimulus-specific subset of mitral cells synchronize their action potential firing and become phase-locked to an oscillation in the local field potential. Patterns of phase-locked spikes convey information about stimulus category, while the residual spikes convey information about precise odor identity. The temporal structure of population activity arising from synaptic interactions in the olfactory bulb therefore affords the simultaneous representation of complementary stimulus information.

Plasticity in the thalamo-cortical auditory system: Rate coding vs. temporal coding?
Jean-Marc Edeline

NAMC, UMR-CNRS 8620, Orsay, France –

Over the last 15 years, physiological reorganizations have been described in the thalamo-cortical system of adult animals after behavioral training. At the single cell level, the receptive fields (RF) of cortical and thalamic neurons can display selective shifts to the frequency of a significant stimulus after a few tens of training trials, both after simple classical conditioning protocol and after discrimination training (see for reviews Edeline 1999, 2003; Weinberger 2003, 2004) . These effects were observed in the secondary and primary auditory cortical fields, as well as in auditory thalamus. At the system level, enlargements of cortical tonotopic maps were described after extensive (2-3 months) training in a perceptual learning task. The map expansion favored the frequency used during the behavioral training. Subsequently, when researchers have looked for physiological mechanisms, the role of the cholinergic system arising from the nucleus basalis magnocellularis has been particularly investigated. Repetitive pairing between a stimulation of the nucleus basalis and a particular tone frequency led to selective RF reorganizations and to selective map enlargements, thus suggesting that learning-induced plasticity mainly results from the activation of the cholinergic system (review in Weinberger 2004, 2003). Recent studies performed with the noradrenergic system (Manunta & Edeline 2004) indicated that selective effects can also be induced, but their direction (mostly decreases) does not square with the dominant effect obtained after behavioral training.
All these findings were systematically described based on rate coding, i.e., by quantification of the strength of evoked responses. This is quite surprizing given that the temporal precision of neuronal discharges has long been considered fundamental for several aspects of auditory processing. Recent attempts to study some facets of temporal coding in the thalamo-cortical auditory system of waking animals will be presented (Cotillon-Williams & Edeline 2003; Massaux et al., 2004). It will be suggested that considering the temporal aspects of neuronal discharges will greatly improve characterization of experience-induced plasticity and help elucidate its mechanisms.

Cotillon-Williams N. & Edeline J-M. (2003) Evoked Oscillations in the Thalamo-Cortical Auditory System Are Present in Anesthetized But Not in Unanesthetized Rats. Journal of Neurophysiology. 89 (4) 1968-1984.
Edeline J-M. (2003) The thalamo-cortical auditory receptive fields: Regulation by the states of vigilance, learning and the  neuromodulatory systems. Exp Brain Res. 153 (4), 554-572.
Edeline, J-M. (1999) Learning-induced physiological plasticity in the thalamo-cortical sensory system:  A critical evalutation of receptive field plasticity and maps changes and their potential mechanisms  Progress in  Neurobiolology , 57, 165-224
Manunta, Y. & Edeline, J-M.  (2004) Noradrenergic Induction of Selective Plasticity in the Frequency Tuning of Auditory Cortex Neurons. J. Neurophysiology 92, 1445-1463.
Massaux A., Dutrieux G, Cotillon-Williams, N, Manunta, Y. & Edeline J-M. (2004) Auditory thalamus bursts in anesthetized and non-anesthetized states: contribution to functional properties J. Neurophysiology. 91, 2117-2134.
Weinberger NM. (2004) Specific long-term memory traces in primary auditory cortex. Nat Rev Neurosci. 5(4):279-90.
Weinberger NM.(2003) The nucleus basalis and memory codes: auditory cortical plasticity and the induction of specific, associative behavioral memory. Neurobiol Learn Mem., 80(3):268-84.

Information encoding in the visual system, and the solution of the binding problem.
Edmund T Rolls.
University of  Oxford, Dept Exp Psych, Oxford OX1 3UD, England.

A key issue in understanding brain function is how information is encoded by the firing of neurons. It has been suggested that in addition to the encoding of information by the firing rates of neurons, information may also be encoded by synchronous firing of subpopulations of neurons, and that this temporal linking might be a solution to the binding problem.
The principled, rigorous, and quantitative way to address the issue of encoding is the use of information theory, for this enables what is learned from the firing rates, and what is learned from any stimulus-dependent synchronous which might be present, to be compared in the same metric, bits of information. We have developed two information theoretic methods to apply to the issue of neuronal encoding. The first uses a Taylor expansion of the Shannon information equation to second order so that it can measure the information present in the cross-correlations between pairs of neurons (Panzeri et al, 1999; Rolls et al, 2003b). The second uses a decoding method based on both the firing rates and the crosscorrelations between neurons to estimate from the neuronal firing which stimulus was shown, and then to calculate the mutual information between this estimated stimulus and the real stimulus that was shown on a trial (Franco et al, 2004). The advantage of this method is that it can be applied to many simultaneously recorded neurons.
These methods have been applied to simultaneously recorded inferior temporal cortex neurons, where individual neurons respond to parts of objects or faces, and others to whole objects or faces (Perrett, Rolls and Caan, 1982), reflecting the binding together of the parts in the correct spatial arrangement in that many of these neurons do not respond to faces with the parts jumbled (Rolls et al 1994). It has been found that most of the information (typically more than 95%) about which of 20 faces has been shown is present in the firing rates, and that typically less than 5% of the information is present in stimulus-dependent cross-correlations between the simultaneously recorded neurons (Franco et al, 2004, Rolls et al 2003b, 2004). Moreover, a great deal of information is available from the firing rate about which stimulus has been show, in that the information increases approximately linearly with the number of neurons (Rolls et al 1997, 2004; Abbott et al, 1996), and is available in short times, such as 20 ms and 100 ms (Tovee and Rolls, 1995).
This information type of analysis has been extended to an ecologically valid visual situation requiring binding of features and segmentation from the background in which the macaque must choose one of two objects shown against a natural visual background to obtain fruit juice. When the visual system is operating under these natural vision conditions, it is still found that most, typically more than 95%, of the information is present in the firing rates, with less than 5% in any stimulus-dependent synchrony that may be present (Aggelopoulos, Franco and Rolls, 2004).
In natural vision conditions, part of the solution used to the binding problem is to reduce the size of the receptive fields of inferior temporal cortex neurons which typically include the fovea to approximately the size of objects (Rolls et al, 2003a). In these conditions, even the position of another object relative to the fovea can be encoded, because some inferior temporal cortex neurons have asymmetric receptive fields. Another part of the solution is to make the neurons respond (with a firing rate change) to combinations of features in the correct relative spatial position, and having many neurons of this type arranged in a hierarchical feedforward cortical circuitry is an excellent solution to the binding problem (Rolls and Deco, 2002).
A computational issue with stimulus-dependent synchrony is that in any case it is only probably feasible to use it for grouping features, and not for solving the binding problem proper, for which the correct relative spatial positions of features must be encoded. To use stimulus-dependent synchronization to help solve the binding problem would appear to need a separate population of “relative spatial position binding neurons”, each one of which would have to be differently keyed to each of the features that have to be bound in the correct spatial configuration. This would increase combinatorily the total number of (feature and “relative spatial position binding”) neurons and the number of time windows required for each binding, making the whole system essentially intractable (Rolls and Deco, 2002, pp 460-461).
It is concluded that in the higher parts of the visual system, the rigorous application of information theory shows that a great deal of information is encoded by the firing rates of different neurons each with different sparse distributed representations of a set of stimuli. Further, stimulus-dependent synchronization adds very little information to that available from the firing rates, and although it could be computationally useful for grouping a set of features, would be very difficult to use to solve the binding problem in which the relative spatial position of features must be encoded. An alternative principle is that the visual system operates as a convergent hierarchy with neurons at any one layer responding to combinations of features in the correct spatial position. This latter system seems consistent with most of the evidence, and is computationally sound (Rolls and Deco, 2002; Riesenhuber and Poggio 2000).

Rolls, E.T. and Deco, G. (2002) Computational Neuroscience of Vision. Oxford University Press: Oxford.
Rolls, E.T. (2000)  Functions of the primate temporal lobe cortical visual areas in invariant visual object and face recognition. Neuron 27: 205-218.
Abbott, L.F., Rolls, E.T. and Tovee, M.J. (1996) Representational capacity of face coding in monkeys. Cerebral Cortex 6: 498-505.
Aggelopoulos, NC, RollsET and Franco, L. (2003) Natural scene perception: some inferior temporal cortex neurons convey information about the location of stimuli with respect to the fovea. Society for Neuroscience Abstract.
Franco, L., Rolls, E.T., Aggelopoulos, N.C. and Treves, A. (2004) The use of decoding to analyze the contribution to the information of the correlations between the firing of simultaneously recorded neurons. Experimental Brain Research 155: 370-384.
Perrett, D.I., Rolls, E.T. and Caan, W. (1982) Visual neurons responsive to faces in the monkey temporal cortex. Experimental Brain Research, 47: 329-342.
Riesenhuber and Poggio, T. (2000). Models of object recognition. Nature Neuroscience Supplement 3: 1199-2004.
Rolls, E.T., Tovee, M.J., Purcell, D.G., Stewart, A.L. and Azzopardi, P. (1994) The responses of neurons in the temporal cortex of primates, and face identification and detection. Experimental Brain Research 101: 474-484.
Rolls, E.T., Aggelopoulos, N.C., and Zheng, F. (2003a) The receptive fields of inferior temporal cortex neurons in natural scenes. Journal of Neuroscience 23: 339-348.
Rolls, E.T., Franco, L., Aggelopoulos, N.C., and Reece, S. (2003b) An information theoretic approach to the contributions of the firing rates and correlations between the firing of neurons. Journal of Neurophysiology 89: 2810-2822.
Rolls, E.T., Aggelopoulos, N.C., Franco, L., and Treves, A. (2004) Information encoding in the inferior temporal cortex: contributions of the firing rates and correlations between the firing of neurons. Biological Cybernetics 90: 19-32.
Tovee, M.J. and Rolls, E.T. (1995) Information encoding in short firing rate epochs by single neurons in the primate temporal visual cortex. Visual Cognition 2: 35-58.

Mechanical amplification by hair cells in the vertebrate ear.
Pascal Martin – Laboratoire Physico-Chimie Curie (UMR168)
Institut Curie recherche – 26, rue d’Ulm 75005 Paris, France

The vertebrate ear relies on nonlinear amplification to enhance its sensitivity and frequency selectivity to oscillatory mechanical stimuli.  In the sacculus of the bullfrog, a hair cell can display active spontaneous oscillations of its mechanosensory hair bundle.  Mechanical stimulation of a single hair bundle with a flexible fiber reveals that an oscillatory hair bundle exhibits exquisite sensitivity to small sinusoidal stimuli near the characteristic frequency of spontaneous oscillations but that the bundle’s sensitivity is greatly diminished for intense stimuli or when the stimulus frequency does not approximate that of the spontaneous movements.  This behavior resembles that of a dynamical system that operates on the brink of an oscillatory instability, a Hopf bifurcation.  A single hair bundle, however, is inevitably subjected to fluctuations that destroy the phase coherence of spontaneous oscillations and limit the bundle’s sensitivity and frequency selectivity.  Here, we combine experimental observations with theoretical simulations to study active hair-bundle motility.  An oscillatory instability is shown to result from the interplay between a region of negative stiffness in the bundle’s force-displacement relation and the Ca2+-regulated activity of molecular motors.  We calculate a state diagram which describes the possible dynamical states of the hair bundle in the absence of noise.  Taking thermal and non thermal sources of fluctuations into account, we then find conditions that yield a response function and spontaneous noisy movements of the hair bundle in quantitative agreement with experiments.  Among all possible operating points, we show that a hair bundle studied experimentally operates near the global optimum of mechanosensitivity in our state diagram.

 Neural mechanisms for olfactory decisions in the rat
Zachary F Mainen
Cold Spring Harbor Laboratory

My laboratory is studying how olfactory stimuli are encoded in the brain and how such information is used to guide behavioral decisions. To do so, we are using a combination of neural ensemble recordings and quantitative behavioral measurements in rats. We hope to sketch a basic outline of the neural processes underlying the execution of a sensory-guided decision. A two-alternative forced-choice paradigm (modeled on those typically used in human or monkey psychophysics) allows us to separately quantify and manipulate sensory and motivational variables related to spatial decisions.
Our behavioral experiments have placed some initial constraints on the time scale of olfactory coding and related decision processes in this task. Evolving temporal codes and repetitive sampling have been hypothesized to allow animals to make finer odor discriminations at the expense of additional processing time. To test this, we measured the relationship between the speed and accuracy of olfactory discriminations while varying stimulus similarity. Odor sampling time in this task was rapid (~200-300 ms) and largely independent of odor similarity, with about 30 ms increase from very easy to near-threshold discriminations. The data appear consistent with a simple ‘random walk’ model which has been used to explain a body of human reaction time data as well as neural data from monkeys. However, the integration time used by the rat appears remarkably short compared to that sometimes observed in primate experiments.
A plausible suggestion for olfactory sensory coding consistent with our behavioral data is a rapid temporal phase code referenced to (and perhaps reset by) the respiration cycle. Sniffing measurements show that accurate decisions can be executed within a single sniff cycle at 7-8Hz (theta frequency). Interestingly, we also find rats precisely synchronize the time of their decision (movement from the odor port) with a specific phase of the respiration cycle. These data are in line with the idea that phase locking at the theta rhythm may be a mechanism for coordination of sensorimotor activity across brain areas (a kind of “network protocol”). Using multi-site local field potential recordings we are examining this hypothesis directly.
To further probe the neural mechanisms of the decision process itself, we are recording neural ensemble activity in the two-alternative discrimination task using multiple tetrodes implanted in the ventrolateral orbitofrontal cortex (VLO) and anterior piriform cortex (APC), two areas in which olfactory information might be coupled to spatial response choice. Analysis of single unit behavioral correlates indicates that the three key components of the task (odor stimulus, choice direction and water reward) are indeed richly represented in both areas. We are also examining second order interactions between neurons (e.g. trial-by-trial fluctuations in response) in relationship to the task. This type of circuit-level analysis shows effective excitatory and inhibitory connectivity and can thereby help to test and constrain recurrent network models hypothesized to underlie sensory integration and decision-making.

Maintaining accuracy at the expense of speed: stimulus similarity defines odor discrimination time in mice
Alan Carleton1, 2, Nixon M. Abraham1, Hartwig Spors1, Troy W. Margrie3, Thomas Kuner1 and Andreas T. Schaefer1
1 WIN Group of Olfactory Dynamics – Max-Planck-Institut für medizinische Forschung – Jahnstrasse 29, D-69120 Heidelberg, Germany
2 Ecole Polytechnique Fédérale de Lausanne,   Brain and Mind Institute, 1015 Lausanne, Switzerland – e-mail:
3 The Wolfson Institute for Biomedical Research, Department of Physiology – University College London, Gower Street, London WC1E 6BT, United Kingdom

Odor discrimination times and their dependence on stimulus similarity were determined to test temporal and spatial models of odor representation in mice. In a go/no-go operant conditioning paradigm, discrimination accuracy exceeded 95% even for odor mixtures that, in intrinsic and voltage-sensitive dye imaging experiments, evoked highly overlapping spatio-temporal patterns in the olfactory bulb. Mice could discriminate two simple odors in less than 200 ms but took 70-100 ms longer for highly similar binary mixtures, revealing a trade-off between accuracy and speed: accurate discrimination of highly similar odor mixtures was maintained at the expense of speed. We propose a spatio-temporal scheme that permits rapid discrimination of dissimilar stimuli, but depends on temporal integration to achieve fine discrimination of odors.

Mapping a Multidimensional Signal;  From Genomics to Pharmacology
Stuart Firestein, Columbia University, New York

The concept of a receptive field has been critical to developing both an intellectual and experimental foundation for understanding coding and neural organization in sensory systems.  In vision, hearing and touch the physical attributes of the stimulus lend themselves to representation within the brain in two or at most three dimensions – e.g., wavelength, frequency, Cartesian coordinates.  This has given rise to the notion of a sensory map in which the outside world is represented and deciphered in brain space by some set of neural transformations from physical energy to spatial dimension.

In the olfactory system we have a stimulus which is multidimensional and does not lend itself easily to such a mapping transformation.  Olfactory stimuli encompass an enormous collection of diverse compounds that differ in atomic composition, functional group, charge, volatility/solubility, structure, and even chirality.  Some odors that appear chemically unrelated produce similar perceptions, while others that differ by as little as a single carbon atom have radically different scents.
Initially the large family of odor receptor GPCRS appeared to provide a molecular basis for this diversity, but it has now become clear that cannot be the whole story – and perhaps is none of it.  It cannot be stated, for example, that there is a critical number of receptors necessary for a sense of smell, nor that having more receptors leads to improved smell, at lest not in a simple linear relation.  Among mammals the number of receptors varies from a few hundred in humans to nearly 1500 in rats.  Although rodents may relay on their olfactory sense to greater degree than humans they are not demonstrably superior in many olfactory tasks.  Between different phyla there are even more significant variations in receptor number, but the effect on the sense of smell is hard to measure.
Nonetheless it seems that progress in understanding olfaction  requires us a more complete understanding of the receptors and their importance to overall olfactory performance.  In this regard there are two features of the receptors that stand out as deserving special attention: their large number and their diverse binding properties.  We have approached these by using tools from two different but converging techniques in modern biology – pharmacology and genomics.  Each of these techniques is at base a method of mapping certain features of the receptors – a molecular receptive range or a genetic receptive range.  Although not yet feasible we look forward to seeing these two “maps” converge, leading to a comprehensive view of olfactory coding at the initial point of contact between the environment and the nervous system.

I will present data from both of these on-going efforts – genomics and bioinformatics of the odor receptor genes, and an analysis of their molecular receptive
fields by pharmacological testing. These suggest the parameters and the limits of an olfactory map that encodes the complex multidimensional stimuli of the chemical environment.

New perspectives on mechanisms underlying the human sense of smell.
G.M. Shepherd.  Department of Neurobiology, Yale University School of Medicine, 333 Cedar Street, New Haven, CT 06517.

Accumulating evidence from both experimental and computational studies has provided for the outlines of a theory of odor discimination.  Here we briefly outline this theory and discuss its relevance to human smell perception.  The theory begins with evidence that the primitives of smell are odor molecule determinants, such as functional group, chain length, etc.  These determinants interact with sites in the receptor binding pocket to activate different receptors to differing degrees. The receptor cell subsets converge on one or a few glomeruli.  The olfactory glomerular sheet encodes the odor molecule determinants into activity patterns, called odor maps, which function as odor images in representing odor space.  Analysing the images thus involves basic operations of pattern recognition.  The odor images are processed by microcircuits in the olfactory bulb.  Interglomerular circuits carry out more specific contrast enhancement between glomerular units, whereas mitral-granule interactions mediate interactions that are more context-enriched.  The microcircuits also provide for synchronization of processing for output to olfactory cortex.  The olfactory cortex contains a canonical cortical circuit that  functions as a content-addressable memory for recognition and memory of the stimulating odors. There is a decline in numbers of olfactory receptor genes in evolution, from over 1000 in rodents to some 350 in humans, suggesting that the discrimination mechanisms are degraded in humans.  However, consideration of factors involved in odor delivery, the greater capacity of the human brain, and evidence from psychophysical testing, suggest that humans in fact have an excellent sense of smell.  The olfactory system thus provides an excellent model for a broader perspective on a systems biology approach to understanding brain function.

Approaches to studying the synaptic mechanisms of sensory perception in the mouse whisker signalling pathway.
Carl Petersen, Lausanne (Switzerland)

Mice use their whiskers to actively gather information regarding objects in the immediate vicinity of their snouts. Whisker deflections evoke cortical signals, which originate in the sensory neurons and are relayed via synapses in the brain stem and thalamus. These cortical sensory responses occur within a well-defined somatotopic map in the barrel cortex, where each whisker is represented by a so-called barrel. This allows the physiological function of cortical neurons to be studied in the context of an anatomical map, which is important for gathering quantitative data.
We are developing behavioural tests, to be combined with recordings of cortical processing to investigate the synaptic mechanisms of sensory perception and associative learning in a simple sensory signalling pathway. Our approaches focus on voltage-sensitive dye imaging and whole-cell recordings, which can be performed on both awake and anestethised mice.  These techniques give us a preliminary view of the synaptic subthreshold processing occuring in barrel cortex during behaviour.

Inhibitory Interneurons in the Olfactory Bulb: From Development to Function
Pierre-Marie Lledo, Lab of Perception and Memory, CNRS URA 2182, Pasteur Institute, Paris, France

In the CNS, networks of inhibitory interneurons play a crucial role in modulating the electrical activity patterns of the principal neurons. Inhibitory interneurons containing GABA carry out specific functions within networks. Identifying and defining the characteristic features of the inhibitory neurons is essential for achieving a cellular understanding of complex brain activities, from perception to cognition. To tackle these questions, the olfactory bulb is ideally suited because of its readily accessible position at the rostral end of the CNS and its distinct pathways of input and output. Recent evidence from our lab indicates that bulbar inhibitory interneurons govern the activity of the profusely interconnected ensembles of projecting neurons, and as such they are responsible for the precise timing of individual principal cell discharges in relation to the emergent behavior of cell assemblies. The olfactory bulb constitutes also the only brain area where local GABAergic neurons are continuously replaced. How the newborn neurons integrate into a pre-existing neural network and how basic functions are maintained when a large percentage of neurons are subjected to continuous renewal, are important questions that have attracted our attention.
We shall see how the production of GABAergic interneurons is specifically adapted to experience-dependent regulation of neural networks. In particular, we shall report the degree of sensitivity of the bulbar neurogenesis to the level of sensory inputs and, in turn, how the adult neurogenesis adjusts the neural network functioning to optimize sensory information processing. By maintaining a constitutive turnover of interneurons, we will see how the ongoing neurogenesis is associated with improved abilities. Finally, we will bring together recently described properties and emerging principles of interneuron function in the olfactory bulb that support a much more complex role for these cells than just providers of inhibition.

Adult neurogenesis: A Neural Basis for Experience-induced plasticity and brain repair
Armen Saghatelyan, Laboratory of Perception and Memory, CNRS URA 2182, Pasteur Institute, 25 rue du Dr. Roux, 75015 Paris Cedex, France

During development and shortly after birth, neuronal activity contributes to the organization of the nervous system. The olfactory bulb (OB), which is the first-order sensory relay for olfactory processing, retains the ability to acquire new interneurons throughout life. It is therefore a particularly appropriate region to study the role of experience in sculpting a neuronal network postnatally.
We have found that modified levels of sensory activity affects radial migration of neuroblasts from the rostral migratory stream (RMS) to the outer layers of OB, as well as survival of newly generated cells once they reached OB, whereas proliferation of neuronal precursors, their tangential migration and differentiation remain largely unaffected. Moreover, we show that sensory activity is primordial for normal maturation processes of newborn interneurons in the OB, since block of odor-induced activity reduces the length of dendritic tree as well as spine density of newly generated interneurons. Importantly, odor deprivation-induced changes in the neurogenesis were accompanied by alterations in the inhibitory-excitatory balance in the OB.
In addition to adjust olfactory bulb functions to an ever-changing environment, new cells continuously generated in the adult brain can be used for brain repair. Repair mechanisms in the adult central nervous system have long been thought to be very limited. However, recent findings elegantly demonstrated that endogenous adult stem cells have the ability to regenerate functional neurons in non-neurogenic diseased areas. Neuronal progenitors migrate to the damaged areas from the neurogenic source localized in the subventricular zone (SVZ) suggesting that changes either in the migratory capabilities of neuroblasts and/or in the microenvironment of the target regions may recruit the newly generated neurons for repair. Here, I will summarize our recent findings about the mechanisms controlling recruitment of newborn neurons in the adult brain and discuss implications for the development of new strategies for the treatment of neurodegenerative diseases.

From the external world to the internal state of affects
J.D. Vincent, CNRS, Gif-sur-Yvette, France

The olfactory system is the first sensorial canal to become active in newborns. In mammals, the sense of smell becomes operational with the first respiratory movements of the newborn. Even in humans, during the first hours of life in the open air, the newborn child behaves like a macrosmatic animal. Meanwhile, the human being is totally dominated by affect. During the rest of the development period and all of adult life, olfaction will remain the sense that opens the most direct route to the affective sphere. Its role not only in emotions but more generally in all types of behaviours including cognitive functions has been until recently neglected. This may be due to the classical view that the affective state, expression of the subjectivity, is always subordinated to the act. I propose an opposing theory according to which the evaluation of our actions becomes accessible only when they are subordinated to the affective state. The subject (body/mind) is its own motive. All animal and human behaviours are organized around two fundamental affective states: pleasure and aversion. The emotions can be defined as basic affects that assist the power of action in terms of maximizing rewards and minimizing punishments. The olfactory system has direct anatomical input connections with the formerly so-called rhinencephalon, a part of the brain that plays a major role in the mechanisms of emotions. It is not surprising therefore that odors act as potent stimuli for emotional reactions in animals and humans. In my presentation I shall review the anatomical and physiological data which support the major roles of olfaction in affective processes throughout life with special focus on development. The role of olfactory impairment in psychological disturbance during childhood is poorly documented and anosmia considered as a minor infirmity, which is not in fact the case.

A neuronal theory of access to consciousness examined with cognitive tasks and knock-out mice
Jean-Pierre Changeux, CNRS URA2182 “Récepteurs et Cognition”, Institut Pasteur, 25 rue du Dr Roux, 75724 Paris Cedex 15, France.

The subjective experience of perceiving visual stimuli is accompanied by objective neuronal activity patterns such as sustained activity in primary visual area (V1), amplification of perceptual processing, correlation across distant regions, joint parietal, frontal and cingulate activation, -band oscillations, and P300 wave-form. We describe a neuronal network model that aims at explaining how those physiological parameters may cohere with conscious reports. The model proposes that the step of conscious perception, referred to as access awareness, is related to the entry of processed visual stimuli into a global brain state that links distant areas including the prefrontal cortex through reciprocal connections, and thus makes perceptual information reportable by multiple means. We use the model to stimulate a classical psychological paradigm: the attentional blink. In additon to reproducing the main objective and subjective features of this paradigm, the model predicts an unique property of nonlinear transition from nonconscious processing to subjective perception. This all-or-none dynamics of conscious perception was verified behaviorally in human subjects.

Dehaene S., Sergent C. & Changeux J.P. 2003, Proceedings of the National Academy of Sciences, USA, 100:8520-8525.

This entry was posted in Reviews and tagged , , , , , , , , , , , , , . Bookmark the permalink.