Search

According to the efficient coding hypothesis, sensory systems are adapted to maximize their ability to encode information about the environment. Sensory neurons play a key role in encoding by selectively modulating their firing rate for a subset of all possible stimuli. This pattern of modulation is often summarized via a tuning curve. The optimally efficient distribution of tuning curves has been calculated in variety of ways for one-dimensional (1-D) stimuli. However, many sensory neurons encode multiple stimulus dimensions simultaneously. It remains unclear how applicable existing models of 1-D tuning curves are for neurons tuned across multiple dimensions. We describe a mathematical generalization that builds on prior work in 1-D to predict optimally efficient multidimensional tuning curves. Our results have implications for interpreting observed properties of neuronal populations. For example, our results suggest that not all tuning curve attributes (such as gain and bandwidth) are equally useful for evaluating the encoding efficiency of a population., Author summary Our brains are tasked with processing a wide range of sensory inputs from the world around us. Natural sensory inputs are often complex and composed of multiple distinctive features (for example, an object may be characterized by its size, shape, color, and weight). Many neurons in the brain play a role in encoding multiple features, or dimensions, of sensory stimuli. Here, we employ the computational technique of population modeling to examine how groups of neurons in the brain can optimally encode multiple dimensions of sensory stimuli. This work provides predictions for theory-driven experiments that can leverage emerging high-throughput neural recording tools to characterize the properties of neuronal populations in response to complex natural stimuli.

Neurons recorded in behaving animals often do not discernibly respond to sensory input and are not overtly task-modulated. These non-classically responsive neurons are difficult to interpret and are typically neglected from analysis, confounding attempts to connect neural activity to perception and behavior. Here, we describe a trial-by-trial, spike-timing-based algorithm to reveal the coding capacities of these neurons in auditory and frontal cortex of behaving rats. Classically responsive and non-classically responsive cells contained significant information about sensory stimuli and behavioral decisions. Stimulus category was more accurately represented in frontal cortex than auditory cortex, via ensembles of non-classically responsive cells coordinating the behavioral meaning of spike timings on correct but not error trials. This unbiased approach allows the contribution of all recorded neurons – particularly those without obvious task-related, trial-averaged firing rate modulation – to be assessed for behavioral relevance on single trials., eLife digest Neurons encode information in the form of electrical signals called spikes. Certain neurons increase the rate at which they produce spikes under specific circumstances, e.g., whenever an animal hears a particular sound. These neurons are said to be 'classically responsive'. But not all neurons behave in this way. Others produce spikes at a variable rate that does not obviously relate to the animal's behavior. These neurons are said to be 'non-classically responsive'. They are often omitted from analyses, despite typically outnumbering their classically responsive counterparts. So, what are these neurons doing? To find out, Insanally et al. trained rats to respond to sounds. The animals learned to poke their nose into a window whenever they heard a specific tone, and to avoid responding whenever they heard any other tone. As the rats performed the task, Insanally et al. recorded from neurons in two areas of the brain, the frontal cortex and the auditory cortex. A computer then analyzed the activity of individual neurons during each trial. As expected, the firing rate of non-classically responsive cells did not relate to the animals' behavior. But the timing of this firing did. The interval between spikes contained information about which tone the animals had heard and/or how they had responded. The cells worked together in groups to encode this information. Over the course of each trial, every neuron in the group varied the interval between its spikes. Eventually, the group reached a consensus, with all neurons using the same interval to represent information relevant to the task. Groups of neurons in the frontal cortex encoded more information about the category of the tone than those in the auditory cortex. By including all neurons – both classically and non-classically responsive – this model offers a more comprehensive view of how neural activity relates to behavior. This may in turn help us understand the variable and complex neural activity seen in people with sensory and cognitive disorders.

Many models of cortical dynamics have focused on the high-firing regime, in which neurons are driven near their maximal rate. Here we consider the responses of neurons in auditory cortex under typical low-firing rate conditions, when stimuli have not been optimized to drive neurons maximally. We used whole-cell patch-clamp recordingin vivoto measure subthreshold membrane potential fluctuations in rat primary auditory cortex in both the anesthetized and awake preparations. By analyzing the subthreshold membrane potential dynamics on single trials, we made inferences about the underlying population activity. We found that, during both spontaneous and evoked responses, membrane potential was highly non-Gaussian, with dynamics consisting of occasional large excursions (sometimes tens of millivolts), much larger than the small fluctuations predicted by most random walk models that predict a Gaussian distribution of membrane potential. Thus, presynaptic inputs under these conditions are organized into quiescent periods punctuated by brief highly synchronous volleys, or “bumps.” These bumps were typically so brief that they could not be well characterized as “up states” or “down states.” We estimate that hundreds, perhaps thousands, of presynaptic neurons participate in the largest volleys. These dynamics suggest a computational scheme in which spike timing is controlled by concerted firing among input neurons rather than by small fluctuations in a sea of background activity.

SummaryAnimals can selectively respond to a target sound despite simultaneous distractors, just as humans can respond to one voice at a crowded cocktail party. To investigate the underlying neural mechanisms, we recorded single-unit activity in primary auditory cortex (A1) and medial prefrontal cortex (mPFC) of rats selectively responding to a target sound from a mixture. We found that prestimulus activity in mPFC encoded the selection rule—which sound from the mixture the rat should select. Moreover, electrically disrupting mPFC significantly impaired performance. Surprisingly, prestimulus activity in A1 also encoded selection rule, a cognitive variable typically considered the domain of prefrontal regions. Prestimulus changes correlated with stimulus-evoked changes, but stimulus tuning was not strongly affected. We suggest a model in which anticipatory activation of a specific network of neurons underlies the selection of a sound from a mixture, giving rise to robust and widespread rule encoding in both brain regions.Video Abstract

The dynamics of subthreshold membrane potential provide insight into the organization of activity in neural circuits. In many brain areas, membrane potential is bistable, transiting between a relatively hyperpolarized down state and a depolarized up state. These up and down states, which have been proposed to play a number of computational roles, have mainly been studied in anesthetized and in vitro preparations. Here, we have used intracellular recordings to characterize the dynamics of membrane potential in the auditory cortex of awake rats. We find that long up states are rare in the awake auditory cortex, with only 0.4% of up states >500 ms. Most neurons displayed only brief up states (bumps) and spent on average ∼1% of recording time in up states >500 ms. We suggest that the near absence of long up states in awake auditory cortex may reflect an adaptation to the rapid processing of auditory stimuli.

We have developed a sparse mathematical representation of speech that minimizes the number of active model neurons needed to represent typical speech sounds. The model learns several well-known acoustic features of speech such as harmonic stacks, formants, onsets and terminations, but we also find more exotic structures in the spectrogram representation of sound such as localized checkerboard patterns and frequency-modulated excitatory subregions flanked by suppressive sidebands. Moreover, several of these novel features resemble neuronal receptive fields reported in the Inferior Colliculus (IC), as well as auditory thalamus and cortex, and our model neurons exhibit the same tradeoff in spectrotemporal resolution as has been observed in IC. To our knowledge, this is the first demonstration that receptive fields of neurons in the ascending mammalian auditory pathway beyond the auditory nerve can be predicted based on coding principles and the statistical properties of recorded sounds., Author Summary The receptive field of a neuron can be thought of as the stimulus that most strongly causes it to be active. Scientists have long been interested in discovering the underlying principles that determine the structure of receptive fields of cells in the auditory pathway to better understand how our brains process sound. One possible way of predicting these receptive fields is by using a theoretical model such as a sparse coding model. In such a model, each sound is represented by the smallest possible number of active model neurons chosen from a much larger group. A primary question addressed in this study is whether the receptive fields of model neurons optimized for natural sounds will predict receptive fields of actual neurons. Here, we use a sparse coding model on speech data. We find that our model neurons do predict receptive fields of auditory neurons, specifically in the Inferior Colliculus (midbrain) as well as the thalamus and cortex. To our knowledge, this is the first time any theoretical model has been able to predict so many of the diverse receptive fields of the various cell-types in those areas.

Sparse coding algorithms trained on natural images can accurately predict the features that excite visual cortical neurons, but it is not known whether such codes can be learned using biologically realistic plasticity rules. We have developed a biophysically motivated spiking network, relying solely on synaptically local information, that can predict the full diversity of V1 simple cell receptive field shapes when trained on natural images. This represents the first demonstration that sparse coding principles, operating within the constraints imposed by cortical architecture, can successfully reproduce these receptive fields. We further prove, mathematically, that sparseness and decorrelation are the key ingredients that allow for synaptically local plasticity rules to optimize a cooperative, linear generative image model formed by the neural representation. Finally, we discuss several interesting emergent properties of our network, with the intent of bridging the gap between theoretical and experimental studies of visual cortex., Author Summary In a sparse coding model, individual input stimuli are represented by the activities of model neurons, the majority of which are inactive in response to any particular stimulus. For a given class of stimuli, the neurons are optimized so that the stimuli can be faithfully represented with the minimum number of co-active units. This has been proposed as a model for visual cortex. While it has previously been demonstrated that sparse coding model neurons, when trained on natural images, learn to represent the same features as do neurons in primate visual cortex, it remains to be demonstrated that this can be achieved with physiologically realistic plasticity rules. In particular, learning in cortex appears to occur by the modification of synaptic connections between neurons, which must depend only on information available locally, at the synapse, and not, for example, on the properties of large numbers of distant cells. We provide the first demonstration that synaptically local plasticity rules are sufficient to learn a sparse image code, and to account for the observed response properties of visual cortical neurons: visual cortex actually could learn a sparse image code.

In vivo whole-cell patch-clamp recording provides a means for measuring membrane currents and potentials from individual cells in the intact animal. Patch-clamp methods have largely been developed in vitro. This body of work has contributed enormously to the understanding of many important phenomena in excitable cells--including synaptic plasticity in the mammalian central nervous system, and the behavior of individual protein channels. In recent years, an increasing number of groups have applied whole-cell recording techniques in the intact animal. Such in vivo studies offer the tantalizing possibility of uncovering the underlying principles and mechanisms of neural interactions within the natural context of fully intact biological networks. This unit focuses on strategies for overcoming the specific technical challenges posed by in vivo whole-cell recording. A straightforward procedure is described for obtaining whole-cell records from the cortex of the anesthetized rat; this procedure has also been applied successfully to awake animals and other rodent species with minor modifications.

Neurons are often assumed to operate in a highly unreliable manner: a neuron can signal the same stimulus with a variable number of action potentials. However, much of the experimental evidence supporting this view was obtained in the visual cortex. We have, therefore, assessed trial-to-trial variability in the auditory cortex of the rat. To ensure single-unit isolation, we used cell-attached recording. Tone-evoked responses were usually transient, often consisting of, on average, only a single spike per stimulus. Surprisingly, the majority of responses were not just transient, but were also binary, consisting of 0 or 1 action potentials, but not more, in response to each stimulus; several dramatic examples consisted of exactly one spike on 100% of trials, with no trial-to-trial variability in spike count. The variability of such binary responses differs from comparably transient responses recorded in visual cortical areas such as area MT, and represent the lowest trial-to-trial variability mathematically possible for responses of a given firing rate. Our study thus establishes for the first time that transient responses in auditory cortex can be described as a binary process, rather than as a highly variable Poisson process. These results demonstrate that cortical architecture can support a more precise control of spike number than was previously recognized, and they suggest a re-evaluation of models of cortical processing that assume noisiness to be an inevitable feature of cortical codes.

Imagine a fly navigating through a forest at 2 m/s. In order to correct for the effects of the wind and other flight instabilities, the fly must continually estimate its heading direction if only to avoid running into a tree or inadvertently flying in circles. Given the striking prominence of eyes on a fly's body, it is not surprising that vision plays a key role in many of its behaviors, including flight (Egelhaaf and Borst 1993). In fact, when a fly is suspended from a wire inside a rotating drum so that the visual scene in front of the fly moves to the right, the fly uses its wings to turn its body to the right, presumably to try to maintain what it perceives as its current heading direction (Reichardt and Poggio 1976). However, flies do not exhibit this behavior following lesions to subsets of the 50 or so identified neurons in the lobular plate that respond to wide-field visual motion (Hausen and Wehrhahn 1983). The fact that these neurons are involved in stabilizing heading direction, which is critical for chasing potential mating partners and avoiding obstacles, suggests that there has probably been strong evolutionary pressure on their performance. One of these neurons, H1, is particularly accessible experimentally, allowing stable extracellular recordings for hours and even days.

Neurons in the auditory cortex can lock with millisecond precision to the fine timing of acoustic stimuli, but it is not known whether this precise spike timing can be used to guide decisions. We used chronically implanted microelectrode pairs to stimulate neurons in the rat auditory cortex directly. Here we demonstrate that rats can exploit differences in the timing of cortical activity as short as three milliseconds to guide decisions.

The sparse coding hypothesis has enjoyed much success in predicting response properties of simple cells in primary visual cortex (V1) based solely on the statistics of natural scenes. In typical sparse coding models, model neuron activities and receptive fields are optimized to accurately represent input stimuli using the least amount of neural activity. As these networks develop to represent a given class of stimulus, the receptive fields are refined so that they capture the most important stimulus features. Intuitively, this is expected to result in sparser network activity over time. Recent experiments, however, show that stimulus-evoked activity in ferret V1 becomes less sparse during development, presenting an apparent challenge to the sparse coding hypothesis. Here we demonstrate that some sparse coding models, such as those employing homeostatic mechanisms on neural firing rates, can exhibit decreasing sparseness during learning, while still achieving good agreement with mature V1 receptive field shapes and a reasonably sparse mature network state. We conclude that observed developmental trends do not rule out sparseness as a principle of neural coding per se: a mature network can perform sparse coding even if sparseness decreases somewhat during development. To make comparisons between model and physiological receptive fields, we introduce a new nonparametric method for comparing receptive field shapes using image registration techniques., Author Summary The popular sparse coding theory posits that the receptive fields of visual cortical neurons maximize the efficiency of the neural representation of natural images. Models implementing this idea typically minimize a combination of the error in reconstructing natural images from neural activities, and the average level of activity in the model neurons. In simulations, these models are presented with natural images and the RFs then develop so as to increase representation efficiency. After a long developmental period, the model RFs typically agree well with those observed experimentally in visual cortex. Since the models seek to minimize (for a given level of reconstruction error) the neural activity levels, the average levels of neural activity might be expected to decrease as the models develop. In the developing mammalian cortex, visual RFs are also modified during development, so the sparse coding hypothesis might appear to suggest that activity levels should decrease during development. Recent experiments with young ferrets show the opposite trend: mature animals tend to have more active visual cortices. Herein, we demonstrate that, depending on the models' initial conditions, some sparse coding models can exhibit increasing activity levels while learning the same types of RFs that are observed in visual cortex: the developmental data do not preclude sparse coding.

How do neuronal populations in the auditory cortex represent acoustic stimuli? Although sound-evoked neural responses in the anesthetized auditory cortex are mainly transient, recent experiments in the unanesthetized preparation have emphasized subpopulations with other response properties. To quantify the relative contributions of these different subpopulations in the awake preparation, we have estimated the representation of sounds across the neuronal population using a representative ensemble of stimuli. We used cell-attached recording with a glass electrode, a method for which single-unit isolation does not depend on neuronal activity, to quantify the fraction of neurons engaged by acoustic stimuli (tones, frequency modulated sweeps, white-noise bursts, and natural stimuli) in the primary auditory cortex of awake head-fixed rats. We find that the population response is sparse, with stimuli typically eliciting high firing rates (>20 spikes/second) in less than 5% of neurons at any instant. Some neurons had very low spontaneous firing rates (, Author Summary How do neuronal populations in the auditory cortex represent sounds? Although sound-evoked neural responses in the anesthetized auditory cortex are mainly transient, recent experiments in the unanesthetized preparation have emphasized subpopulations with other response properties. We quantified the relative contributions of these different subpopulations in the auditory cortex of awake head-fixed rats. We recorded neuronal activity using cell-attached recordings with a glass electrode—a method for which isolation of individual neurons does not depend on neuronal activity—while probing neurons with a representative ensemble of sounds. Our data therefore address the question: What is the typical response to a particular stimulus? We find that the population response is sparse, with sounds typically eliciting high activity in less than 5% of neurons at any instant. The overall population response was well described by a lognormal distribution, rather than the exponential distribution that is often reported. Our results represent, to our knowledge, the first quantitative evidence for sparse representations of sounds in the unanesthetized auditory cortex. These results are compatible with a model in which most neurons are silent much of the time, and in which representations are composed of small dynamic subsets of highly active neurons., Patch clamp recordings in the auditory cortex of unanesthetized rats reveal that the population response to sounds is sparse and that most neurons are silent most of the time.

SummaryIt is unclear why there are so many more neurons in sensory cortex than in the sensory periphery. One possibility is that these “extra” neurons are used to overcome cortical noise and faithfully represent the acoustic stimulus. Another possibility is that even after overcoming cortical noise, there is “excess representational bandwidth” available and that this bandwidth is used to represent conjunctions of auditory and nonauditory information for computation. Here, we discuss recent data about neuronal reliability in auditory cortex showing that cortical noise may not be as high as was previously believed. Although at present, the data suggest that auditory cortex neurons can be more reliable than those in the visual cortex, we speculate that the principles governing cortical computation are universal and that visual and other cortical areas can also exploit strategies based on similarly high-fidelity activity.