How behavioral constraints may determine optimal sensory representations

Neural circuit models for evidence accumulation through choice-selective sequences

Decision making is traditionally thought to be mediated by populations of neurons whose firing rates persistently accumulate evidence across time. However, recent decision-making experiments in rodents have observed neurons across the brain that fire sequentially as a function of spatial position or time, rather than persistently, with the subset of neurons in the sequence depending on the animal's choice. We develop two new candidate circuit models, in which evidence is encoded either in the relative firing rates of two competing chains of neurons or in the network location of a stereotyped pattern ("bump") of neural activity. Encoded evidence is then faithfully transferred between neuronal populations representing different positions or times. Neural recordings from four different brain regions during a decision-making task showed that, during the evidence accumulation period, different brain regions displayed tuning curves consistent with different candidate models for evidence accumulation. This work provides mechanistic models and potential neural substrates for how graded-value information may be precisely accumulated within and transferred between neural populations, a set of computations fundamental to many cognitive operations.

Sensory perception relies on fitness-maximizing codes

Sensory information encoded by humans and other organisms is generally presumed to be as accurate as their biological limitations allow. However, perhaps counterintuitively, accurate sensory representations may not necessarily maximize the organism's chances of survival. To test this hypothesis, we developed a unified normative framework for fitness-maximizing encoding by combining theoretical insights from neuroscience, computer science, and economics. Behavioural experiments in humans revealed that sensory encoding strategies are flexibly adapted to promote fitness maximization, a result confirmed by deep neural networks with information capacity constraints trained to solve the same task as humans. Moreover, human functional MRI data revealed that novel behavioural goals that rely on object perception induce efficient stimulus representations in early sensory structures. These results suggest that fitness-maximizing rules imposed by the environment are applied at early stages of sensory processing in humans and machines.

Efficient coding of cognitive variables underlies dopamine response and choice behavior

Reward expectations based on internal knowledge of the external environment are a core component of adaptive behavior. However, internal knowledge may be inaccurate or incomplete due to errors in sensory measurements. Some features of the environment may also be encoded inaccurately to minimize representational costs associated with their processing. In this study, we investigated how reward expectations are affected by features of internal representations by studying behavior and dopaminergic activity while mice make time-based decisions. We show that several possible representations allow a reinforcement learning agent to model animals' overall performance during the task. However, only a small subset of highly compressed representations simultaneously reproduced the co-variability in animals' choice behavior and dopaminergic activity. Strikingly, these representations predict an unusual distribution of response times that closely match animals' behavior. These results inform how constraints of representational efficiency may be expressed in encoding representations of dynamic cognitive variables used for reward-based computations.

Rational inattention in mice

Behavior exhibited by humans and other organisms is generally inconsistent and biased and, thus, is often labeled irrational. However, the origins of this seemingly suboptimal behavior remain elusive. We developed a behavioral task and normative framework to reveal how organisms should allocate their limited processing resources such that sensory precision and its related metabolic investment are balanced to guarantee maximal utility. We found that mice act as rational inattentive agents by adaptively allocating their sensory resources in a way that maximizes reward consumption in previously unexperienced stimulus-reward association environments. Unexpectedly, perception of commonly occurring stimuli was relatively imprecise; however, this apparent statistical fallacy implies "awareness" and efficient adaptation to their neurocognitive limitations. Arousal systems carry reward distribution information of sensory signals, and distributional reinforcement learning mechanisms regulate sensory precision via top-down normalization. These findings reveal how organisms efficiently perceive and adapt to previously unexperienced environmental contexts within the constraints imposed by neurobiology.

Spontaneous quantitative processing in Chinese singular and plural picture naming: An event-related potentials analysis

Chinese nouns lack inflection and cannot reflect the quantitative relationship between singular and plural numbers. However, neural processes of picture naming are different from those of words. We assume that Chinese single and plural picture naming is different, and they may involve quantitative processing. Therefore, Experiment 1 was designed by picking picture naming as the task and Chinese as the target language and compared the accuracy, reaction time, and event-related potentials (ERPs) between single and plural picture naming, where two types of pictures were mixed. Although the T-test showed no significant differences in behavioral data, there were differences in ERPs. ERP differences involved two effects: P1 of 160-180 ms and P2 of 220-260 ms in the parietal-occipital lobe. These differences are suggested to reflect the neural differences in quantitative processing. Therefore, Chinese singular and plural picture naming consists of word production and implicit quantitative processing simultaneously. To explore the relationship between the two processings, we added a semantic factor (inanimate vs. animate items) to the quantity factor of Experiment 1 and carried out Experiment 2, with the observation indexes unchanged. There were no significant differences in behavioral data among the four conditions. After variance analysis, ERPs results indicated an interaction between semantic and quantitative factors in the central area at 180-280 ms. In summary, we suggest that Chinese singular and plural picture naming includes two simultaneous neural processing tasks: word production and quantitative processing, which interact in the central area at 180-280 ms.

Efficient sampling and noisy decisions

Human decisions are based on finite information, which makes them inherently imprecise. But what determines the degree of such imprecision? Here, we develop an efficient coding framework for higher-level cognitive processes in which information is represented by a finite number of discrete samples. We characterize the sampling process that maximizes perceptual accuracy or fitness under the often-adopted assumption that full adaptation to an environmental distribution is possible, and show how the optimal process differs when detailed information about the current contextual distribution is costly. We tested this theory on a numerosity discrimination task, and found that humans efficiently adapt to contextual distributions, but in the way predicted by the model in which people must economize on environmental information. Thus, understanding decision behavior requires that we account for biological restrictions on information coding, challenging the often-adopted assumption of precise prior knowledge in higher-level decision systems.

Emergence of an Adaptive Command for Orienting Behavior in Premotor Brainstem Neurons of Barn Owls

The midbrain map of auditory space commands sound-orienting responses in barn owls. Owls precisely localize sounds in frontal space but underestimate the direction of peripheral sound sources. This bias for central locations was proposed to be adaptive to the decreased reliability in the periphery of sensory cues used for sound localization by the owl. Understanding the neural pathway supporting this biased behavior provides a means to address how adaptive motor commands are implemented by neurons. Here we find that the sensory input for sound direction is weighted by its reliability in premotor neurons of the midbrain tegmentum of owls (male and female), such that the mean population firing rate approximates the head-orienting behavior. We provide evidence that this coding may emerge through convergence of upstream projections from the midbrain map of auditory space. We further show that manipulating the sensory input yields changes predicted by the convergent network in both premotor neural responses and behavior. This work demonstrates how a topographic sensory representation can be linearly read out to adjust behavioral responses by the reliability of the sensory input.SIGNIFICANCE STATEMENT This research shows how statistics of the sensory input can be integrated into a behavioral command by readout of a sensory representation. The firing rate of midbrain premotor neurons receiving sensory information from a topographic representation of auditory space is weighted by the reliability of sensory cues. We show that these premotor responses are consistent with a weighted convergence from the topographic sensory representation. This convergence was also tested behaviorally, where manipulation of stimulus properties led to bidirectional changes in sound localization errors. Thus a topographic representation of auditory space is translated into a premotor command for sound localization that is modulated by sensory reliability.

Independent processing of increments and decrements in odorant concentration by ON and OFF olfactory receptor neurons

A salient feature of the insect olfactory system is its ability to detect and interpret simultaneously the identity and concentration of an odorant signal along with the temporal stimulus cues that are essential for accurate odorant tracking. The olfactory system of the cockroach utilizes two parallel pathways for encoding of odorant identity and the moment-to-moment succession of odorant concentrations as well as the rate at which concentration changes. This separation originates at the peripheral level of the ORNs (olfactory receptor neurons) which are localized in basiconic and trichoid sensilla. The graded activity of ORNs in the basiconic sensilla provides the variable for the combinatorial representation of odorant identity. The antagonistically responding ON and OFF ORNs in the trichoid sensilla transmit information about concentration increments and decrements with excitatory signals. Each ON and OFF ORN adjusts its gain for odorant concentration and its rate of change to the temporal dynamics of the odorant signal: as the rate of change diminishes, both ORNs improve their sensitivity for the rate of change at the expense of the sensitivity for the instantaneous concentration. This suggests that the ON and OFF ORNs are optimized to detect minute fluctuations or even creeping changes in odorant concentration.

Decoding sound level in the marmoset primary auditory cortex

Neurons that respond favorably to a particular sound level have been observed throughout the central auditory system, becoming steadily more common at higher processing areas. One theory about the role of these level-tuned or nonmonotonic neurons is the level-invariant encoding of sounds. To investigate this theory, we simulated various subpopulations of neurons by drawing from real primary auditory cortex (A1) neuron responses and surveyed their performance in forming different sound level representations. Pure nonmonotonic subpopulations did not provide the best level-invariant decoding; instead, mixtures of monotonic and nonmonotonic neurons provided the most accurate decoding. For level-fidelity decoding, the inclusion of nonmonotonic neurons slightly improved or did not change decoding accuracy until they constituted a high proportion. These results indicate that nonmonotonic neurons fill an encoding role complementary to, rather than alternate to, monotonic neurons.NEW & NOTEWORTHY Neurons with nonmonotonic rate-level functions are unique to the central auditory system. These level-tuned neurons have been proposed to account for invariant sound perception across sound levels. Through systematic simulations based on real neuron responses, this study shows that neuron populations perform sound encoding optimally when containing both monotonic and nonmonotonic neurons. The results indicate that instead of working independently, nonmonotonic neurons complement the function of monotonic neurons in different sound-encoding contexts.

Magnitude Codes for Cross-Modal Working Memory in the Primate Frontal Association Cortex

Quantitative features of stimuli may be ordered along a magnitude continuum, or line. Magnitude refers to parameters of different types of stimulus properties. For instance, the frequency of a sound relates to sensory and continuous stimulus properties, whereas the number of items in a set is an abstract and discrete property. In addition, within a stimulus property, magnitudes need to be processed not only in one modality, but across multiple modalities. In the sensory domain, for example, magnitude applies to both to the frequency of auditory sounds and tactile vibrations. Similarly, both the number of visual items and acoustic events constitute numerical quantity, or numerosity. To support goal-directed behavior and executive functions across time, magnitudes need to be held in working memory, the ability to briefly retain and manipulate information in mind. How different types of magnitudes across multiple modalities are represented in working memory by single neurons has only recently been explored in primates. These studies show that neurons in the frontal lobe can encode the same magnitude type across sensory modalities. However, while multimodal sensory magnitude in relative comparison tasks is represented by monotonically increasing or decreasing response functions ("summation code"), multimodal numerical quantity in absolute matching tasks is encoded by neurons tuned to preferred numerosities ("labeled-line code"). These findings indicate that most likely there is not a single type of cross-modal working-memory code for magnitudes, but rather a flexible code that depends on the stimulus dimension as well as on the task requirements.

Optimal compensation for neuron loss

The brain has an impressive ability to withstand neural damage. Diseases that kill neurons can go unnoticed for years, and incomplete brain lesions or silencing of neurons often fail to produce any behavioral effect. How does the brain compensate for such damage, and what are the limits of this compensation? We propose that neural circuits instantly compensate for neuron loss, thereby preserving their function as much as possible. We show that this compensation can explain changes in tuning curves induced by neuron silencing across a variety of systems, including the primary visual cortex. We find that compensatory mechanisms can be implemented through the dynamics of networks with a tight balance of excitation and inhibition, without requiring synaptic plasticity. The limits of this compensatory mechanism are reached when excitation and inhibition become unbalanced, thereby demarcating a recovery boundary, where signal representation fails and where diseases may become symptomatic.

Efficient Neural Codes That Minimize Lp Reconstruction Error

The efficient coding hypothesis assumes that biological sensory systems use neural codes that are optimized to best possibly represent the stimuli that occur in their environment. Most common models use information-theoretic measures, whereas alternative formulations propose incorporating downstream decoding performance. Here we provide a systematic evaluation of different optimality criteria using a parametric formulation of the efficient coding problem based on the [Formula: see text] reconstruction error of the maximum likelihood decoder. This parametric family includes both the information maximization criterion and squared decoding error as special cases. We analytically derived the optimal tuning curve of a single neuron encoding a one-dimensional stimulus with an arbitrary input distribution. We show how the result can be generalized to a class of neural populations by introducing the concept of a meta-tuning curve. The predictions of our framework are tested against previously measured characteristics of some early visual systems found in biology. We find solutions that correspond to low values of [Formula: see text], suggesting that across different animal models, neural representations in the early visual pathways optimize similar criteria about natural stimuli that are relatively close to the information maximization criterion.

Efficient sensory cortical coding optimizes pursuit eye movements

In the natural world, the statistics of sensory stimuli fluctuate across a wide range. In theory, the brain could maximize information recovery if sensory neurons adaptively rescale their sensitivity to the current range of inputs. Such adaptive coding has been observed in a variety of systems, but the premise that adaptation optimizes behaviour has not been tested. Here we show that adaptation in cortical sensory neurons maximizes information about visual motion in pursuit eye movements guided by that cortical activity. We find that gain adaptation drives a rapid (<100 ms) recovery of information after shifts in motion variance, because the neurons and behaviour rescale their sensitivity to motion fluctuations. Both neurons and pursuit rapidly adopt a response gain that maximizes motion information and minimizes tracking errors. Thus, efficient sensory coding is not simply an ideal standard but a description of real sensory computation that manifests in improved behavioural performance.

Connecting Neural Codes with Behavior in the Auditory System of Drosophila

Brains are optimized for processing ethologically relevant sensory signals. However, few studies have characterized the neural coding mechanisms that underlie the transformation from natural sensory information to behavior. Here, we focus on acoustic communication in Drosophila melanogaster and use computational modeling to link natural courtship song, neuronal codes, and female behavioral responses to song. We show that melanogaster females are sensitive to long timescale song structure (on the order of tens of seconds). From intracellular recordings, we generate models that recapitulate neural responses to acoustic stimuli. We link these neural codes with female behavior by generating model neural responses to natural courtship song. Using a simple decoder, we predict female behavioral responses to the same song stimuli with high accuracy. Our modeling approach reveals how long timescale song features are represented by the Drosophila brain and how neural representations can be decoded to generate behavioral selectivity for acoustic communication signals.

A Bayesian observer model constrained by efficient coding can explain 'anti-Bayesian' percepts

Bayesian observer models provide a principled account of the fact that our perception of the world rarely matches physical reality. The standard explanation is that our percepts are biased toward our prior beliefs. However, reported psychophysical data suggest that this view may be simplistic. We propose a new model formulation based on efficient coding that is fully specified for any given natural stimulus distribution. The model makes two new and seemingly anti-Bayesian predictions. First, it predicts that perception is often biased away from an observer's prior beliefs. Second, it predicts that stimulus uncertainty differentially affects perceptual bias depending on whether the uncertainty is induced by internal or external noise. We found that both model predictions match reported perceptual biases in perceived visual orientation and spatial frequency, and were able to explain data that have not been explained before. The model is general and should prove applicable to other perceptual variables and tasks.

Neural population coding of multiple stimuli

In natural scenes, objects generally appear together with other objects. Yet, theoretical studies of neural population coding typically focus on the encoding of single objects in isolation. Experimental studies suggest that neural responses to multiple objects are well described by linear or nonlinear combinations of the responses to constituent objects, a phenomenon we call stimulus mixing. Here, we present a theoretical analysis of the consequences of common forms of stimulus mixing observed in cortical responses. We show that some of these mixing rules can severely compromise the brain's ability to decode the individual objects. This cost is usually greater than the cost incurred by even large reductions in the gain or large increases in neural variability, explaining why the benefits of attention can be understood primarily in terms of a stimulus selection, or demixing, mechanism rather than purely as a gain increase or noise reduction mechanism. The cost of stimulus mixing becomes even higher when the number of encoded objects increases, suggesting a novel mechanism that might contribute to set size effects observed in myriad psychophysical tasks. We further show that a specific form of neural correlation and heterogeneity in stimulus mixing among the neurons can partially alleviate the harmful effects of stimulus mixing. Finally, we derive simple conditions that must be satisfied for unharmful mixing of stimuli.

The influence of population size, noise strength and behavioral task on best-encoded stimulus for neurons with unimodal or monotonic tuning curves

Tuning curves and receptive fields are widely used to describe the selectivity of sensory neurons, but the relationship between firing rates and information is not always intuitive. Neither high firing rates nor high tuning curve gradients necessarily mean that stimuli at that part of the tuning curve are well represented by a neuron. Recent research has shown that trial-to-trial variability (noise) and population size can strongly affect which stimuli are most precisely represented by a neuron in the context of a population code (the best-encoded stimulus), and that different measures of information can give conflicting indications. Specifically, the Fisher information is greatest where the tuning curve gradient is greatest, such as on the flanks of peaked tuning curves, but the stimulus-specific information (SSI) is greatest at the tuning curve peak for small populations with high trial-to-trial variability. Previous research in this area has focussed upon unimodal (peaked) tuning curves, and in this article we extend these analyses to monotonic tuning curves. In addition, we examine how stimulus spacing in forced choice tasks affects the best-encoded stimulus. Our results show that, regardless of the tuning curve, Fisher information correctly predicts the best-encoded stimulus for large populations and where the stimuli are closely spaced in forced choice tasks. In smaller populations with high variability, or in forced choice tasks with widely-spaced choices, the best-encoded stimulus falls at the peak of unimodal tuning curves, but is more variable for monotonic tuning curves. Task, population size and variability all need to be considered when assessing which stimuli a neuron represents, but the best-encoded stimulus can be estimated on a case-by case basis using commonly available computing facilities.

Toward functional classification of neuronal types

How many types of neurons are there in the brain? This basic neuroscience question remains unsettled despite many decades of research. Classification schemes have been proposed based on anatomical, electrophysiological, or molecular properties. However, different schemes do not always agree with each other. This raises the question of whether one can classify neurons based on their function directly. For example, among sensory neurons, can a classification scheme be devised that is based on their role in encoding sensory stimuli? Here, theoretical arguments are outlined for how this can be achieved using information theory by looking at optimal numbers of cell types and paying attention to two key properties: correlations between inputs and noise in neural responses. This theoretical framework could help to map the hierarchical tree relating different neuronal classes within and across species.

Challenges for the understanding of the dynamics of social coordination

The way people interact can be examined by looking at the way they move relative to each other. Seeking the principles behind those interactions have consequences potentially related to any type of interpersonal function, far beyond the so-called "motor" processes typically associated with the study of movements, be it perceptive, cognitive, affective, pragmatic, or epistemic. Here, we present the way the framework of coordination dynamics define and addresses the interactive actions in a dyad. We first introduce the basics of pattern formation as the roots of the theoretical approach of coordination dynamics, and then the way this framework may contribute to establish a solution to classify behaviors. Thereafter we review promising empirical results on the dynamics of interpersonal coordination, and finally discuss were to go next to decipher the way the coordination between two people and the way each individual contribute may be disentangled.

Transformation of the neural code for tactile detection from thalamus to cortex

To understand how sensory-driven neural activity gives rise to perception, it is essential to characterize how various relay stations in the brain encode stimulus presence. Neurons in the ventral posterior lateral (VPL) nucleus of the somatosensory thalamus and in primary somatosensory cortex (S1) respond to vibrotactile stimulation with relatively slow modulations (∼100 ms) of their firing rate. In addition, faster modulations (∼10 ms) time-locked to the stimulus waveform are observed in both areas, but their contribution to stimulus detection is unknown. Furthermore, it is unclear whether VPL and S1 neurons encode stimulus presence with similar accuracy and via the same response features. To address these questions, we recorded single neurons while trained monkeys judged the presence or absence of a vibrotactile stimulus of variable amplitude, and their activity was analyzed with a unique decoding method that is sensitive to the time scale of the firing rate fluctuations. We found that the maximum detection accuracy of single neurons is similar in VPL and S1. However, VPL relies more heavily on fast rate modulations than S1, and as a consequence, the neural code in S1 is more tolerant: its performance degrades less when the readout method or the time scale of integration is suboptimal. Therefore, S1 neurons implement a more robust code, one less sensitive to the temporal integration window used to infer stimulus presence downstream. The differences between VPL and S1 responses signaling the appearance of a stimulus suggest a transformation of the neural code from thalamus to cortex.

Transformation from a pure time delay to a mixed time and phase delay representation in the auditory forebrain pathway

Birds and mammals exploit interaural time differences (ITDs) for sound localization. Subsequent to ITD detection by brainstem neurons, ITD processing continues in parallel midbrain and forebrain pathways. In the barn owl, both ITD detection and processing in the midbrain are specialized to extract ITDs independent of frequency, which amounts to a pure time delay representation. Recent results have elucidated different mechanisms of ITD detection in mammals, which lead to a representation of small ITDs in high-frequency channels and large ITDs in low-frequency channels, resembling a phase delay representation. However, the detection mechanism does not prevent a change in ITD representation at higher processing stages. Here we analyze ITD tuning across frequency channels with pure tone and noise stimuli in neurons of the barn owl's auditory arcopallium, a nucleus at the endpoint of the forebrain pathway. To extend the analysis of ITD representation across frequency bands to a large neural population, we employed Fourier analysis for the spectral decomposition of ITD curves recorded with noise stimuli. This method was validated using physiological as well as model data. We found that low frequencies convey sensitivity to large ITDs, whereas high frequencies convey sensitivity to small ITDs. Moreover, different linear phase frequency regimes in the high-frequency and low-frequency ranges suggested an independent convergence of inputs from these frequency channels. Our results are consistent with ITD being remodeled toward a phase delay representation along the forebrain pathway. This indicates that sensory representations may undergo substantial reorganization, presumably in relation to specific behavioral output.

Maximization of learning speed in the motor cortex due to neuronal redundancy

Many redundancies play functional roles in motor control and motor learning. For example, kinematic and muscle redundancies contribute to stabilizing posture and impedance control, respectively. Another redundancy is the number of neurons themselves; there are overwhelmingly more neurons than muscles, and many combinations of neural activation can generate identical muscle activity. The functional roles of this neuronal redundancy remains unknown. Analysis of a redundant neural network model makes it possible to investigate these functional roles while varying the number of model neurons and holding constant the number of output units. Our analysis reveals that learning speed reaches its maximum value if and only if the model includes sufficient neuronal redundancy. This analytical result does not depend on whether the distribution of the preferred direction is uniform or a skewed bimodal, both of which have been reported in neurophysiological studies. Neuronal redundancy maximizes learning speed, even if the neural network model includes recurrent connections, a nonlinear activation function, or nonlinear muscle units. Furthermore, our results do not rely on the shape of the generalization function. The results of this study suggest that one of the functional roles of neuronal redundancy is to maximize learning speed.

There are overwhelmingly more neurons than muscles in the motor system. The functional roles of this neuronal redundancy remains unknown. Our analysis, which uses a redundant neural network model, reveals that learning speed reaches its maximum value if and only if the model includes sufficient neuronal redundancy. This result does not depend on whether the distribution of the preferred direction is uniform or a skewed bimodal, both of which have been reported in neurophysiological studies. We have confirmed that our results are consistent, regardless of whether the model includes recurrent connections, a nonlinear activation function, or nonlinear muscle units. Additionally, our results are the same when using either a broad or a narrow generalization function. These results suggest that one of the functional roles of neuronal redundancy is to maximize learning speed.

Olfactory maps in the brain

The responses of neural elements in many sensory areas of the brain vary systematically with their physical position, leading to a topographic representation of the outside world. Sensory representation in the olfactory system has been harder to decipher, in part because it is difficult to find appropriate metrics to characterize odor space and to sample this space densely. Recent experiments have shown that the arrangement of glomeruli, the elementary units of processing, is relatively invariant across individuals in a species, yet it is flexible enough to accommodate new sensors that might be added. Evidence supports the existence of coarse spatial domains carved out on a genetic or functional basis, but a systematic organization of odor responses or neural circuits on a local scale is not evident. Experiments and theory that relate the properties of odorant receptors to the detailed wiring diagram of the downstream olfactory circuits and to behaviors they trigger may reveal the design principles that have emerged during evolution.

The neurobiology of insect olfaction: sensory processing in a comparative context

The simplicity and accessibility of the olfactory systems of insects underlie a body of research essential to understanding not only olfactory function but also general principles of sensory processing. As insect olfactory neurobiology takes advantage of a variety of species separated by millions of years of evolution, the field naturally has yielded some conflicting results. Far from impeding progress, the varieties of insect olfactory systems reflect the various natural histories, adaptations to specific environments, and the roles olfaction plays in the life of the species studied. We review current findings in insect olfactory neurobiology, with special attention to differences among species. We begin by describing the olfactory environments and olfactory-based behaviors of insects, as these form the context in which neurobiological findings are interpreted. Next, we review recent work describing changes in olfactory systems as adaptations to new environments or behaviors promoting speciation. We proceed to discuss variations on the basic anatomy of the antennal (olfactory) lobe of the brain and higher-order olfactory centers. Finally, we describe features of olfactory information processing including gain control, transformation between input and output by operations such as broadening and sharpening of tuning curves, the role of spiking synchrony in the antennal lobe, and the encoding of temporal features of encounters with an odor plume. In each section, we draw connections between particular features of the olfactory neurobiology of a species and the animal's life history. We propose that this perspective is beneficial for insect olfactory neurobiology in particular and sensory neurobiology in general.

Stages of nonsymbolic number processing in occipitoparietal cortex disentangled by fMRI adaptation

The neurobiological mechanisms of nonsymbolic number processing in humans are still unclear. Computational modeling proposed three successive stages: first, the spatial location of objects is stored in an object location map; second, this information is transformed into a numerical summation code; third, this summation code is transformed to a number-selective code. Here, we used fMRI-adaptation to identify these three stages and their relative anatomical location. By presenting the same number of dots on the same locations in the visual field, we adapted neurons of human volunteers. Occasionally, deviants with the same number of dots at different locations or different numbers of dots at the same location were shown. By orthogonal number and location factors in the deviants, we were able to calculate three independent contrasts, each sensitive to one of the stages. We found an occipitoparietal gradient for nonsymbolic number processing: the activation of the object location map was found in the inferior occipital gyrus. The summation coding map exhibited a nonlinear pattern of activation, with first increasing and then decreasing activation, and most activity in the middle occipital gyrus. Finally, the number-selective code became more pronounced in the superior parietal lobe. In summary, we disentangled the three stages of nonsymbolic number processing predicted by computational modeling and demonstrated that they constitute a pathway along the occipitoparietal processing stream.

Cues to move increased information in superior colliculus tuning curves

Shifts in the location of spatial attention produce increases in the gain and sensitivity of neuronal responses to sensory stimuli. Cues to shift the line of sight have the same effect on sensory responses in a motor area involved in the control of eye movements, the superior colliculus. Evidence has shown that shifts of gaze and shifts of attention are linked, suggesting there may be similar underlying mechanisms. Here, we report on a novel way in which cues to move the eyes (top-down signals) can influence sensory responses of neurons by altering the variability of their discharge rate. We measured the spatial tuning of superior colliculus neuronal activity in trials with cues to either make or withhold saccadic eye movements. We found that tuning curve widths both increased and decreased, but that the information conveyed by the neuronal discharge about the stimulus increased with a cue to make a saccade. The increase in information resulted partly from a decrease in trial-to-trial variability of neuronal discharges for stimuli located at the flanks of the tuning curves rather than from increases in the discharge rate for stimuli located at the peak of the tuning curves. This result is consistent with theoretical work and provides a novel way for cognitive signals to influence sensory responses within motor regions of the brain.

The morphology of the rat vibrissal array: a model for quantifying spatiotemporal patterns of whisker-object contact

In all sensory modalities, the data acquired by the nervous system is shaped by the biomechanics, material properties, and the morphology of the peripheral sensory organs. The rat vibrissal (whisker) system is one of the premier models in neuroscience to study the relationship between physical embodiment of the sensor array and the neural circuits underlying perception. To date, however, the three-dimensional morphology of the vibrissal array has not been characterized. Quantifying array morphology is important because it directly constrains the mechanosensory inputs that will be generated during behavior. These inputs in turn shape all subsequent neural processing in the vibrissal-trigeminal system, from the trigeminal ganglion to primary somatosensory ("barrel") cortex. Here we develop a set of equations for the morphology of the vibrissal array that accurately describes the location of every point on every whisker to within ±5% of the whisker length. Given only a whisker's identity (row and column location within the array), the equations establish the whisker's two-dimensional (2D) shape as well as three-dimensional (3D) position and orientation. The equations were developed via parameterization of 2D and 3D scans of six rat vibrissal arrays, and the parameters were specifically chosen to be consistent with those commonly measured in behavioral studies. The final morphological model was used to simulate the contact patterns that would be generated as a rat uses its whiskers to tactually explore objects with varying curvatures. The simulations demonstrate that altering the morphology of the array changes the relationship between the sensory signals acquired and the curvature of the object. The morphology of the vibrissal array thus directly constrains the nature of the neural computations that can be associated with extraction of a particular object feature. These results illustrate the key role that the physical embodiment of the sensor array plays in the sensing process.

Luminal administration ex vivo of a live Lactobacillus species moderates mouse jejunal motility within minutes

Gut commensals modulate host immune, endocrine, and metabolic functions. They also affect peripheral and central neural reflexes and function. We have previously shown that daily ingestion of Lactobacillus reuteri (LR) for 9 d inhibits the pseudoaffective cardiac response and spinal single-fiber discharge evoked by visceral distension, and decreases intestinal motility and myenteric AH cell slow afterhyperpolarization (sAHP) by inhibiting a Ca-activated K (IK(Ca)) channel. We tested whether luminal LR could acutely decrease motility in an ex vivo perfusion model of naive Balb/c jejunum. Live LR dose dependently decreased motor complex pressure wave amplitudes with 9- to 16-min onset latency and an IC(50) of 5 × 10(7) cells/ml Krebs. Heat-killed LR or another live commensal, Lactobacillus salivarius, were without effect. The IK(Ca) channel blocker TRAM-34, but neither the opener (DCEBIO) nor the hyperpolarization-activated cationic channel inhibitor ZD7288 (5 μM) (or TTX 1 μM), mimicked the LR effect on motility acutely ex vivo. We provide evidence for a rapid, strain-specific, dose-dependent action of a live Lactobacillus on small intestinal motility reflexes that recapitulates the long-term effects of LR ingestion. These observations may be useful as a first step to unraveling the pathways involved in bacteria to the nervous system communication.

Autonomous Development of Vergence Control Driven by Disparity Energy Neuron Populations

We present a simple optimization criterion that leads to autonomous development of a sensorimotor feedback loop driven by the neural representation of the depth in the mammalian visual cortex. Our test bed is an active stereo vision system where the vergence angle between the two eyes is controlled by the output of a population of disparity-selective neurons. By finding a policy that maximizes the total response across the neuron population, the system eventually tracks a target as it moves in depth. We characterized the tracking performance of the resulting policy using objects moving both sinusoidally and randomly in depth. Surprisingly, the system can even learn how to track based on stimuli it cannot track: even though the closed loop 3 dB tracking bandwidth of the system is 0.3 Hz, correct tracking policies are learned for input stimuli moving as fast as 0.75 Hz.

Contrast-dependence of surround suppression in Macaque V1: experimental testing of a recurrent network model

Neuronal responses in primary visual cortex (V1) to optimally oriented high-contrast stimuli in the receptive field (RF) center are suppressed by stimuli in the RF surround, but can be facilitated when the RF center is stimulated at low contrast. The neural circuits and mechanisms for surround modulation are still unknown. We previously proposed that topdown feedback connections mediate suppression from the "far" surround, while "near' surround suppression is mediated primarily by horizontal connections. We implemented this idea in a recurrent network model of V1. A model assumption needed to account for the contrast-dependent sign of surround modulation is a response asymmetry between excitation and inhibition; accordingly, inhibition, but not excitation, is silent for weak visual inputs to the RF center, and surround stimulation can evoke facilitation. A prediction stemming from this same assumption is that surround suppression is weaker for low than for high contrast stimuli in the RF center. Previous studies are inconsistent with this prediction. Using single unit recordings in macaque V1, we confirm this model's prediction. Model simulations demonstrate that our results can be reconciled with those from previous studies. We also performed a systematic comparison of the experimentally measured surround suppression strength with predictions of the model operated in different parameter regimes. We find that the original model, with strong horizontal and no feedback excitation of local inhibitory neurons, can only partially account quantitatively for the experimentally measured suppression. Strong direct feedback excitation of V1 inhibitory neurons is necessary to account for the experimentally measured surround suppression strength.

Initiation and termination of integration in a decision process

Experimental data indicate that simple motor decisions in vertebrates are preceded by integration of evidence in certain cortical areas, and that the competition between them is resolved in the basal ganglia. While the occurrence of cortical integration is well established, it is not yet clear exactly how the integration occurs. Several models have been proposed, including the race model, the feed forward inhibition (FFI) model and the leaky competing accumulator (LCA) model. In this paper we establish qualitative and quantitative differences between the above mentioned models, with respect to how they are able to initiate the integration process without integrating noise prior to stimulus onset, as well as the models' ability to terminate the integration after a decision has been made, to ensure the possibility of subsequent decisions. Our results show that the LCA model has advantages over the race model and the FFI model in both respects, leading to shorter decision times and an effective termination process.

Neural population coding is optimized by discrete tuning curves

The sigmoidal tuning curve that maximizes the mutual information for a Poisson neuron, or population of Poisson neurons, is obtained. The optimal tuning curve is found to have a discrete structure that results in a quantization of the input signal. The number of quantization levels undergoes a hierarchy of phase transitions as the length of the coding window is varied. We postulate, using the mammalian auditory system as an example, that the presence of a subpopulation structure within a neural population is consistent with an optimal neural code.

What response properties do individual neurons need to underlie position and clutter "invariant" object recognition?

Primates can easily identify visual objects over large changes in retinal position--a property commonly referred to as position "invariance." This ability is widely assumed to depend on neurons in inferior temporal cortex (IT) that can respond selectively to isolated visual objects over similarly large ranges of retinal position. However, in the real world, objects rarely appear in isolation, and the interplay between position invariance and the representation of multiple objects (i.e., clutter) remains unresolved. At the heart of this issue is the intuition that the representations of nearby objects can interfere with one another and that the large receptive fields needed for position invariance can exacerbate this problem by increasing the range over which interference acts. Indeed, most IT neurons' responses are strongly affected by the presence of clutter. While external mechanisms (such as attention) are often invoked as a way out of the problem, we show (using recorded neuronal data and simulations) that the intrinsic properties of IT population responses, by themselves, can support object recognition in the face of limited clutter. Furthermore, we carried out extensive simulations of hypothetical neuronal populations to identify the essential individual-neuron ingredients of a good population representation. These simulations show that the crucial neuronal property to support recognition in clutter is not preservation of response magnitude, but preservation of each neuron's rank-order object preference under identity-preserving image transformations (e.g., clutter). Because IT neuronal responses often exhibit that response property, while neurons in earlier visual areas (e.g., V1) do not, we suggest that preserving the rank-order object preference regardless of clutter, rather than the response magnitude, more precisely describes the goal of individual neurons at the top of the ventral visual stream.

Neural representations that support invariant object recognition

Neural mechanisms underlying invariant behaviour such as object recognition are not well understood. For brain regions critical for object recognition, such as inferior temporal cortex (ITC), there is now ample evidence indicating that single cells code for many stimulus aspects, implying that only a moderate degree of invariance is present. However, recent theoretical and empirical work seems to suggest that integrating responses of multiple non-invariant units may produce invariant representations at population level. We provide an explicit test for the hypothesis that a linear read-out mechanism of a pool of units resembling ITC neurons may achieve invariant performance in an identification task. A linear classifier was trained to decode a particular value in a 2-D stimulus space using as input the response pattern across a population of units. Only one dimension was relevant for the task, and the stimulus location on the irrelevant dimension (ID) was kept constant during training. In a series of identification tests, the stimulus location on the relevant dimension (RD) and ID was manipulated, yielding estimates for both the level of sensitivity and tolerance reached by the network. We studied the effects of several single-cell characteristics as well as population characteristics typically considered in the literature, but found little support for the hypothesis. While the classifier averages out effects of idiosyncratic tuning properties and inter-unit variability, its invariance is very much determined by the (hypothetical) ‘average’ neuron. Consequently, even at population level there exists a fundamental trade-off between selectivity and tolerance, and invariant behaviour does not emerge spontaneously.

Maximally informative stimuli and tuning curves for sigmoidal rate-coding neurons and populations

A general method for deriving maximally informative sigmoidal tuning curves for neural systems with small normalized variability is presented. The optimal tuning curve is a nonlinear function of the cumulative distribution function of the stimulus and depends on the mean-variance relationship of the neural system. The derivation is based on a known relationship between Shannon's mutual information and Fisher information, and the optimality of Jeffrey's prior. It relies on the existence of closed-form solutions to the converse problem of optimizing the stimulus distribution for a given tuning curve. It is shown that maximum mutual information corresponds to constant Fisher information only if the stimulus is uniformly distributed. As an example, the case of sub-Poisson binomial firing statistics is analyzed in detail.

Neural response properties of primary, rostral, and rostrotemporal core fields in the auditory cortex of marmoset monkeys

The core region of primate auditory cortex contains a primary and two primary-like fields (AI, primary auditory cortex; R, rostral field; RT, rostrotemporal field). Although it is reasonable to assume that multiple core fields provide an advantage for auditory processing over a single primary field, the differential roles these fields play and whether they form a functional pathway collectively such as for the processing of spectral or temporal information are unknown. In this report we compare the response properties of neurons in the three core fields to pure tones and sinusoidally amplitude modulated tones in awake marmoset monkeys (Callithrix jacchus). The main observations are as follows. (1) All three fields are responsive to spectrally narrowband sounds and are tonotopically organized. (2) Field AI responds more strongly to pure tones than fields R and RT. (3) Field RT neurons have lower best sound levels than those of neurons in fields AI and R. In addition, rate-level functions in field RT are more commonly nonmonotonic than in fields AI and R. (4) Neurons in fields RT and R have longer minimum latencies than those of field AI neurons. (5) Fields RT and R have poorer stimulus synchronization than that of field AI to amplitude-modulated tones. (6) Between the three core fields the more rostral regions (R and RT) have narrower firing-rate-based modulation transfer functions than that of AI. This effect was seen only for the nonsynchronized neurons. Synchronized neurons showed no such trend.

Population coding with motion energy filters: the impact of correlations

The codes obtained from the responses of large populations of neurons are known as population codes. Several studies have shown that the amount of information conveyed by such codes, and the format of this information, is highly dependent on the pattern of correlations. However, very little is known about the impact of response correlations (as found in actual cortical circuits) on neural coding. To address this problem, we investigated the properties of population codes obtained from motion energy filters, which provide one of the best models for motion selectivity in early visual areas. It is therefore likely that the correlations that arise among energy filters also arise among motion-selective neurons. We adopted an ideal observer approach to analyze filter responses to three sets of images: noisy sine gratings, random dots kinematograms, and images of natural scenes. We report that in our model, the structure of the population code varies with the type of image. We also show that for all sets of images, correlations convey a large fraction of the information: 40% to 90% of the total information. Moreover, ignoring those correlations when decoding leads to considerable information loss-from 50% to 93%, depending on the image type. Finally we show that it is important to consider a large population of motion energy filters in order to see the impact of correlations. Study of pairs of neurons, as is often done experimentally, can underestimate the effect of correlations.

Neuronal population coding of continuous and discrete quantity in the primate posterior parietal cortex

Quantitative knowledge guides vital decisions in the life of animals and humans alike. The posterior parietal cortex in primates has been implicated in representing abstract quantity, both continuous (extent) and discrete (number of items), supporting the idea of a putative generalized magnitude system in this brain area. Whether or not single neurons encode different types of quantity, or how quantitative information is represented in the neuronal responses, however, is unknown. We show that length and numerosity are encoded by functionally overlapping groups of parietal neurons. Using a statistical classifier, we found that the activity of populations of quantity-selective neurons contained accurate information about continuous and discrete quantity. Unexpectedly, even neurons that were nonselective according to classical spike-count measures conveyed robust categorical information that predicted the monkeys' quantity judgments. Thus, different information-carrying processes of partly intermingled neuronal networks in the parietal lobe seem to encode various forms of abstract quantity.

Robustness of neural codes and its implication on natural image processing

In this study, based on the view of statistical inference, we investigate the robustness of neural codes, i.e., the sensitivity of neural responses to noise, and its implication on the construction of neural coding. We first identify the key factors that influence the sensitivity of neural responses, and find that the overlap between neural receptive fields plays a critical role. We then construct a robust coding scheme, which enforces the neural responses not only to encode external inputs well, but also to have small variability. Based on this scheme, we find that the optimal basis functions for encoding natural images resemble the receptive fields of simple cells in the striate cortex. We also apply this scheme to identify the important features in the representation of face images and Chinese characters.

Untangling invariant object recognition

Despite tremendous variation in the appearance of visual objects, primates can recognize a multitude of objects, each in a fraction of a second, with no apparent effort. However, the brain mechanisms that enable this fundamental ability are not understood. Drawing on ideas from neurophysiology and computation, we present a graphical perspective on the key computational challenges of object recognition, and argue that the format of neuronal population representation and a property that we term 'object tangling' are central. We use this perspective to show that the primate ventral visual processing stream achieves a particularly effective solution in which single-neuron invariance is not the goal. Finally, we speculate on the key neuronal mechanisms that could enable this solution, which, if understood, would have far-reaching implications for cognitive neuroscience.

A labeled-line code for small and large numerosities in the monkey prefrontal cortex

How single neurons represent information about the magnitude of a stimulus remains controversial. Neurons encoding purely sensory magnitude typically show monotonic response functions ("summation coding"), and summation units are usually implemented in models of numerosity representation. In contrast, cells representing numerical quantity exhibit nonmonotonic tuning functions that peak at their preferred numerosity ("labeled-line code"), but the restricted range of tested quantities in these studies did not permit a definite answer. Here, we analyzed both behavioral and neuronal representations of a broad range of numerosities from 1 to 30 in the prefrontal cortex of monkeys. Numerosity-selective neurons showed a clear and behaviorally relevant labeled-line code for all numerosities. Moreover, both the behavioral and neuronal tuning functions obeyed the Weber-Fechner Law and were best represented on a nonlinearly compressed scale. Our single-cell study is in good agreement with functional imaging data reporting peaked tuning functions in humans, demonstrating neuronal precursors for human number competence in a nonhuman primate. Our findings also emphasize that the manner in which neurons encode and maintain magnitude information may depend on the precise task at hand as well as the type of magnitude to represent and memorize.

Differential neural coding of acoustic flutter within primate auditory cortex

A sequence of acoustic events is perceived either as one continuous sound or as a stream of temporally discrete sounds (acoustic flutter), depending on the rate at which the acoustic events repeat. Acoustic flutter is perceived at repetition rates near or below the lower limit for perceiving pitch, and is akin to the discrete percepts of visual flicker and tactile flutter caused by the slow repetition of sensory stimulation. It has been shown that slowly repeating acoustic events are represented explicitly by stimulus-synchronized neuronal firing patterns in primary auditory cortex (AI). Here we show that a second neural code for acoustic flutter exists in the auditory cortex of marmoset monkeys (Callithrix jacchus), in which the firing rate of a neuron is a monotonic function of an acoustic event's repetition rate. Whereas many neurons in AI encode acoustic flutter using a dual temporal/rate representation, we find that neurons in cortical fields rostral to AI predominantly use a monotonic rate code and lack stimulus-synchronized discharges. These findings indicate that the neural representation of acoustic flutter is transformed along the caudal-to-rostral axis of auditory cortex.