Odors that are essential to the survival and reproduction of a species take the form of complex mixtures of volatiles. Often, an odor source such as food or a potential mate releases a mixture with characteristic ratios between the components. Here, the encoding of the characteristic ratio between components of the pheromone released by a female moth is investigated in the antennal lobe (AL) of a male moth (Manduca sexta). The mechanisms by which olfactory systems of diverse insect species process odors are adapted to the particular environment and olfactory behavior of the animal. In the moth, innately attractive odors produce patterns of synchrony in the output of the AL, the projection neurons (PNs). Male moths exhibited attraction to synthetic mixtures of pheromone components that was selective for ratios at or near the natural ratio released by females. Selectivity increased as the moth neared the odor source and initiated mating behaviors. PNs in the macroglomerular complex (MGC) did not exhibit an effect of component ratio on their firing rate responses. However, pairs of PNs exhibited increased synchrony in response to the behaviorally effective ratios of pheromone components. Individual pairs exhibited selectivity for ratios within 1 order of magnitude from the natural ratio. Synchrony in PN spiking was not phase-locked to the network oscillations in the AL. A model for ratio-selective enhancement of synchronous PN output in the AL is proposed.

We analyze the spike trains of multiple simultaneously recorded grid cells obtained in di erent conditions, to help determine the role of recurrent network feedback in generating grid responses. An important class of models of grid cell activity is based on low dimensional continuous attractor dynamics arising from recurrent connections within the grid system. A necessary prediction of these models is that the strong recurrent connections force the grid responses of di erent cells to maintain fi xed relative spatial phases over long periods of time, even if the response patterns of each neuron change. The observation that grid cells maintain their relative spatial phase relationships across di erent familiar environments supports the presence of recurrent connections, but does not rule out the possibility that these relationships persist due to feed-forward input. We analyze the stability of pairwise neural correlations for experiments in which the spatial responses of single neurons change over time. The first such experiment involves resizing of a familiar enclosure, with the result that spatial grid responses rescale along the resized dimension. We show that the relative spatial phase of ring between pairs of cells remains stable over time even as the absolute spatial phase of ring in these same cells changes greatly through rescaling. This result is again consistent with recurrent connectivity, but it remains possible that common external sensory cues (e.g. border information arriving from boundary cells) somehow register the rescaled grids of all cells to display the same relative phases as before rescaling. In an attempt to address this, we analyze responses from animals first exposure to novel environments. Grid ring becomes more noisy and the spatial ring pattern expands, then relaxes back to the periodicity seen in familiar enclosures. During the relaxation, external sensory cues are static and thus likely not responsible for the changing grid responses. We show that the constant phase relationships seen across familiar environments are present from first exposure as well. Finally, we illustrate a generative model to predict grid cell spikes. The aim is to obtain the key determinants of grid cell ring, including animal location, velocity modulation, neural adaptation, and recurrent feedback in a Bayesian framework, and thus assess network contributions to grid cell activity. / text

A major challenge in neuroscience is to understand neural coding - how sensory stimuli, or motor actions controlling the behaviour of an animal, are represented by the activity of neurons in the nervous system. Information theory provides a powerful tool for investigating such codes; a rigorous non-parametric theoretical framework for quantifying the properties of noisy communication channels. By viewing various stages of the nervous system as such channels, we can use information theory to obtain meaningful quantitative results about the information capacity of the system, the sensory stimuli or features that are represented, as well as the performance of different candidate codes. In this thesis, we first develop an open source neuroinformatics toolbox implementing robust estimators and bias-corrections for a range of information theoretic quantities. This also includes a novel implementation of an algorithm for numerically obtaining distributions with maximum entropy over discrete probability spaces subject to marginal equality constraints. These maximum entropy distributions provide a powerful tool for investigating the effect of interactions on information transmission by neural population codes. These neuroinformatics tools are then used to explore population coding of the velocity of sinusoidal whisker stimulation in the cortex of anaesthetised rats. We show that both with a pooled model (assuming the neural population is homogenous) and with a labelled line model, interactions are present and affect the information transmission in the system. We show that interactions of order higher than two have a measurable but minor effect on the information capacity of the neural population. This is the first direct quantification of the effect of high order interactions on information transmission in a neural system, and is one of the first studies for which the data lie outside of the perturbative regime in which pairwise models are guaranteed to perform well. We then consider results from a novel experimental preparation, recording from populations in VPm thalamus under both white noise and naturalistic whisker stimulation. We show that sub-millisecond precise spike timing, previously observed with white noise stimuli, is also present with naturalistic stimuli, and that the diverse feature selectivity previously observed across different single unit recording sessions is also present with neurons simultaneously recorded within a single barrelloid. We use a novel information theoretic approach to probe the kinetic selectivity of the recorded cells, showing that they encode combinations of position, velocity and acceleration and that of these, velocity is the best encoded feature. We also quantify the information available to cortex under both a count code and a labelled line code, showing that a simple pooling of the population by a downstream decoder results in a large loss of information, but that this loss may be ameliorated by choosing more carefully the subpopulations over which to pool activity. Finally, we again apply the maximum entropy tools to quantify the effect of interactions, including a novel calculation of the maximal information available to a downstream decoder neglecting correlations of different orders, and find that, similar to the results in cortex, high order correlations do have a measurable effect on information transmission.

What does serotonin do? Just as the diffuse axonal arbours of midbrain serotonin neurons touch nearly every corner of the forebrain, so too is this ancient neuromodulator involved in nearly every aspect of learning and behaviour. The role of serotonin in reward processing has received increasing attention in recent years, but there is little agreement about how the perplexing responses of serotonin neurons to emotionally salient stimuli should be interpreted, and essentially nothing is known about how they arise. Here I approach these two aspects of serotonergic function in reverse order. In the first part of this thesis, I construct an experimentally-constrained spiking neural network model of the dorsal raphe nucleus (DRN), the main source of forebrain serotonergic input, and characterize its signal processing features. I show that potent spike-frequency adaptation deeply shapes DRN output while other aspects of its physiology are relatively less important. Overall, this part of my work suggests that in vivo serotonergic activity patterns arise from a temporal-derivative-like computation. But the temporal derivative of what? In the second part, I consider the possibility that the DRN is driven by an input that represents cumulative future reward, a quantity called state value in reinforcement learning theory. The resulting model reproduces established tuning features of serotonin neurons, including phasic activation by reward predicting cues and punishments, reward-specific surprise tuning, and tonic modulation by reward and punishment context. Because these features are the basis of many and varied existing serotonergic theories, these results show that my theory, which I call value prediction, provides a unifying perspective on serotonergic function. Finally, in an empirical test of the theory, I re-analyze data from an in vivo trace conditioning experiment and find that value prediction accounts for the firing rates of serotonin neurons to a precision ≪0.1 Hz, outperforming previous models by a large margin. Here I establish serotonin as a new neural substrate of prediction and reward, a significant step towards understanding the role of serotonin signalling in the brain.

The activity of cortical neurons in awake brains changes dynamically as a function of the behavioural and attentional state. The primary motor cortex (M1) plays a central role in regulating complex motor behaviors. Despite a growing knowledge on its connectivity and spiking pattern, little is known about intra-cellular mechanism and rhythms underlying motor-command generation. In the last decade, whole-cell recordings in awake animals has become a powerful tool for characterising both sub-and supra-threshold activity during behaviour. Seminal in vivo studies have shown that changes in input structure and sub-threshold regime determine spike output during behaviour (input-output transformations). In this thesis I make use of computational and experimental techniques to better understand (i) how the brain regulates the sub-threshold activity of the neurons during movement and (ii) how this reflects in their input-output transformation properties. In the first part of this work I present a novel probabilistic technique to infer input statistics from in-vivo voltage-clamp traces. This approach, based on Bayesian belief networks, outperforms current methods and allows an estimation of synaptic input (i) kinetic properties, (ii) frequency, and (iii) weight distribution. I first validate the model on simulated data, thus I apply it to voltage-clamp recordings of cerebellar interneurons in awake mice. I found that synaptic weight distributions have long tails, which on average do not change during movement. Interestingly, the increase in synaptic current observed during movement is a consequence of the increase in input frequency only. In the second part, I study how the brain regulates the activity of pyramidal neurons in the M1 of awake mice during movement. I performed whole-cell recordings of pyramidal neurons in layer 5B (L5B), which represent one of the main descending output channels from motor cortex. I found that slow large-amplitude membrane potential fluctuations, typical of quiet periods, were suppressed in all L5B pyramidal neurons during movement, which by itself reduced membrane potential (Vm) variability, input sensitivity and output firing rates. However, a sub-population of L5B neurons ( 50%) concurrently experienced an increase in excitatory drive that depolarized mean Vm, enhanced input sensitivity and elevated firing rates. Thus, movement-related bidirectional modulation in L5B neurons is mediated by two opposing mechanisms: 1) a global reduction in network driven Vm variability and 2) a coincident, targeted increase in excitatory drive to a subpopulation of L5B neurons.

This thesis develops a new method for time-frequency signal processing and examines the relevance of the new representation in studies of neural coding in songbirds. The method groups together associated regions of the time-frequency plane into objects defined by time-frequency contours. By combining information about structurally stable contour shapes over multiple time-scales and angles, a signal decomposition is produced that distributes resolution adaptively. As a result, distinct signal components are represented in their own most parsimonious forms.  Next, through neural recordings in singing birds, it was found that activity in song premotor cortex is significantly correlated with the objects defined by this new representation of sound. In this process, an automated way of finding sub-syllable acoustic transitions in birdsongs was first developed, and then increased spiking probability was found at the boundaries of these acoustic transitions. Finally, a new approach to study auditory cortical sequence processing more generally is proposed. In this approach, songbirds were trained to discriminate Morse-code-like sequences of clicks, and the neural correlates of this behavior were examined in primary and secondary auditory cortex. It was found that a distinct transformation of auditory responses to the sequences of clicks exists as information transferred from primary to secondary auditory areas. Neurons in secondary auditory areas respond asynchronously and selectively -- in a manner that depends on the temporal context of the click. This transformation from a temporal to a spatial representation of sound provides a possible basis for the songbird's natural ability to discriminate complex temporal sequences.

Understanding the complex brain computations leading to object recognition requires quantitatively characterizing the information represented in inferior temporal cortex (IT), the highest stage of the primate visual stream. A read-out technique based on a trainable classifier is used to characterize the neural coding of selectivity and invariance at the population level. The activity of very small populations of independently recorded IT neurons (~100 randomly selected cells) over very short time intervals (as small as 12.5 ms) contains surprisingly accurate and robust information about both object ‘identity’ and ‘category’, which is furthermore highly invariant to object position and scale. Significantly, selectivity and invariance are present even for novel objects, indicating that these properties arise from the intrinsic circuitry and do not require object-specific learning. Within the limits of the technique, there is no detectable difference in the latency or temporal resolution of the IT information supporting so-called ‘categorization’ (a.k. basic level) and ‘identification’ (a.k. subordinate level) tasks. Furthermore, where information, in particular information about stimulus location and scale, can also be read-out from the same small population of IT neurons. These results show how it is possible to decode invariant object information rapidly, accurately and robustly from a small population in IT and provide insights into the nature of the neural code for different kinds of object-related information.

Motion in the outside world forms one of the primary uses of visual information for many animals. The ability to interpret motion quickly and accurately permits interaction with and response to events in the outside world. While much is known about some aspects of motion perception, there is less agreement about how feature selectivity leading to motion perception is actually formed in the convergent and divergent pathways of the visual system. It is even less clear how these classical understandings of motion processing, often driven by artificial stimuli with little resemblance to the outside world, correspond to responses of neurons when using more natural stimuli. In this thesis, we probe these gaps, first by demonstrating that synchronization within the visual thalamus leads to efficient representations of motion (through tuning properties) in primary visual cortex, exploiting precise timing across populations in a unique manner compared to traditional models. We then create a novel “minimally-natural” stimulus with the appearance of an infinite hallway wallpapered with sinusoidal gratings, to probe how such minimally natural features modulate our predictions of neural responses based upon feature tuning properties. Through encoding and decoding models we find that measuring a restricted tuning parameter space limits our ability to capture all response properties but preserves relevant information for decoding. We finish with an exploration of ethologically relevant natural features, perspective and complex motion, and show that even moderate amounts of each feature within or near the classical V1 receptive field changes the neural response from what classical feature tuning would predict and improves stimulus classification tremendously. Together all of these results indicate that capturing information about motion in the outside world through visual stimuli requires a more advanced model of feature selectivity that incorporates parameters based on more complex spatial relationships.

Dans cette thèse, nous abordons les problèmes de la transmission et du traitement de l'information par les assemblées de neurones, du point de vue de l'approche inter-disciplinaire des systèmes complexes en nous référant principalement aux formalismes de la théorie de l'information et de la théorie des systèmes dynamiques. Dans ce contexte, nous nous focalisons sur les mécanismes de représentation de l'information sensorielle par les activités neuronales à travers le codage neuronal. Nous explorons la structure de ce code, à plusieurs échelles grâce à l'étude de différents signaux électrophysiologiques issus de populations de neurones (signaux unitaires, LFP et EEG). Sur le plan méthodologique, nous avons implémenté différents indices permettant d'extraire objectivement l'information des activités neuronales, mais également d'en caractériser la dynamique sous-jacente à partir de séries temporelles de taille finie (le taux d'entropie). Nous avons également étudié un indicateur peu utilisé (le taux d'information mutuelle), qui permet de quantifier l'auto-organisation et les relations de couplage entre deux systèmes. Grâce à des approches théoriques et numériques, nous analysons les propriétés caractéristiques de ces indices et proposons leur utilisation dans le cadre de l'étude des systèmes neuronaux. Ce travail permet de caractériser la complexité de différentes activités neuronales associées aux dynamiques de transmission de l'information. / In this thesis, we address the problem of transmission and information processing by neuronal assemblies, in terms of the interdisciplinary approach of complex systems by referring mainly to the formalisms of information theory and dynamical systems. In this context, we focus on the mechanisms underlying sensory information representation by neuronal activity through neural coding. We explore the structure of this code under several scales through the study of different neuronal population electrophysiological signals (singel unit, LFP and EEG). We have implemented various indices in order to extract objectively information from neural activity, but also to characterize the underlying dynamics from finite size time series (the entropy rate). We also defined a new indicator (the mutual information rate), which quantifies self-organization and relations of coupling between two systems. Using theoretical and numerical approaches, we analyze some characteristic properties of these indices and propose their use in the context of the study of neural systems. This work allows us to characterize the complexity of different neuronal activity associated to information transmission dynamics.

Two fundamental questions in the investigation of any sensory system are what physical signals drive its primary sensory neurons and how such signals are encoded by the successive neural levels during natural behaviour. Due to the complexity of experiments with awake, actively sensing animals, most previous studies focused on anesthetized animals, where the motor component of sensation is abolished and therefore those questions are so far largely unanswered. The aim of this thesis is to exploit recent advance in electrophysiological, behavioural and computational techniques to address those questions in the sub-cortical whisker system of the mouse. To determine the input to the whisker system, in Chapter 2 I recorded from primary whisker afferents (PWAs) of awake, head-fixed mice as they explored a pole with their whiskers, and simultaneously measured both whisker motion and forces with high-speed videography. To predict PWA firing, I used Generalised Linear Models. I found that PWA responses were poorly predicted by whisker angle, but well predicted by rotational force (moment) acting on the whiskers. This concept of “moment encoding” could account for the activity of PWAs under diverse conditions - whisking in air, active whisker-mediated touch and passive whisker deflection. The discovery that PWAs encode moment raises the question of how mice employ moment to control their tactile behaviours. In Chapter 3 I therefore measured moment at the base of the whiskers of head-fixed mice, performing a novel behavioural task, which involved whisker-based object localisation. I then tested which features of moment during whiskerobject touch could predict mouse choice. By using probabilistic classifiers, I discovered that mouse choices could be accurately predicted from moment magnitude and direction during touch, combined with a non-sensory variable - the mouse choice in the previous trial. Finally, in Chapter 4 I asked how tactile coding generalized to whisker system sub-cortical brains regions during a natural active whisker-based behaviour. I therefore combined a naturalistic whisker-guided navigation task and extracellular recording with a novel generation of high density silicon probes (O3 Neuropixel probes) and studied how touch and locomotion were encoded by the whisker first (ventral posterior nucleus, VPM) and higher order thalamic relay (posterior complex, PO) and hypothalamic regions and (zona incerta, ZI). Using multiple linear regressions, I found that neurons in the relay nucleus VPM encoded not only touch, but also locomotion signals. Similarly, neurons in the higherorder regions PO and ZI were driven by both touch and locomotion. My study showed that in the awake animal, in the central part of the whisker system, peripheral signals were preserved, but were encoded concomitantly with motor variables, such as locomotion. In summary, in this thesis I identified the mechanical variable representing the major sensory input to the whisker system. I showed that mice are able to employ it to guide behaviour and found that correlate of this signal was encoded by central neurons of the whisker system in VPM, PO and ZI, concomitantly with locomotion.