Desegregation of neuronal predictive processing.

Neural circuits construct internal 'world-models' to guide behavior. The predictive processing framework posits that neural activity signaling sensory predictions and concurrently computing prediction-errors is a signature of those internal models. Here, to understand how the brain generates predictions for complex sensorimotor signals, we investigate the emergence of high-dimensional, multi-modal predictive representations in recurrent networks. We find that robust predictive processing arises in a network with loose excitatory/inhibitory balance. Contrary to previous proposals of functionally specialized cell-types, the network exhibits desegregation of stimulus and prediction-error representations. We confirmed these model predictions by experimentally probing predictive-coding circuits using a rich stimulus-set to violate learned expectations. When constrained by data, our model further reveals and makes concrete testable experimental predictions for the distinct functional roles of excitatory and inhibitory neurons, and of neurons in different layers along a laminar hierarchy, in computing multi-modal predictions. These results together imply that in natural conditions, neural representations of internal models are highly distributed, yet structured to allow flexible readout of behaviorally-relevant information. The generality of our model advances the understanding of computation of internal models across species, by incorporating different types of predictive computations into a unified framework.

Sensation and expectation are embedded in mouse motor cortical activity.

During behavior, the motor cortex sends copies of motor-related signals to sensory cortices. Here, we combine closed-loop behavior with large-scale physiology, projection-pattern-specific recordings, and circuit perturbations to show that neurons in mouse secondary motor cortex (M2) encode sensation and are influenced by expectation. When a movement unexpectedly produces a sound, M2 becomes dominated by sound-evoked activity. Sound responses in M2 are inherited partially from the auditory cortex and are routed back to the auditory cortex, providing a path for the reciprocal exchange of sensory-motor information during behavior. When the acoustic consequences of a movement become predictable, M2 responses to self-generated sounds are selectively gated off. These changes in single-cell responses are reflected in population dynamics, which are influenced by both sensation and expectation. Together, these findings reveal the embedding of sensory and expectation signals in motor cortical activity.

Auditory neuroscience: Sounds make the face move.

Animals including humans often react to sounds by involuntarily moving their face and body. A new study shows that facial movements provide a simple and reliable readout of a mouse's hearing ability that is more sensitive than traditional measurements.

Learning within a sensory-motor circuit links action to expected outcome.

The cortex integrates sound- and movement-related signals to predict the acoustic consequences of behavior and detect violations from expectations. Although expectation- and prediction-related activity has been observed in the auditory cortex of humans, monkeys, and mice during vocal and non-vocal acoustic behaviors, the specific cortical circuitry required for forming memories, recalling expectations, and making predictions remains unknown. By combining closed-loop behavior, electrophysiological recordings, longitudinal pharmacology, and targeted optogenetic circuit activation, we identify a cortical locus for the emergence of expectation and error signals. Movement-related expectation signals and sound-related error signals emerge in parallel in the auditory cortex and are concentrated in largely distinct neurons, consistent with a compartmentalization of different prediction-related computations. On a trial-by-trial basis, expectation and error signals are correlated in auditory cortex, consistent with a local circuit implementation of an internal model. Silencing the auditory cortex during motor-sensory learning prevents the emergence of expectation signals and error signals, revealing the auditory cortex as a necessary node for learning to make predictions. Prediction-like signals can be experimentally induced in the auditory cortex, even in the absence of behavioral experience, by pairing optogenetic motor cortical activation with sound playback, indicating that cortical circuits are sufficient for movement-like predictive processing. Finally, motor-sensory experience realigns the manifold dimensions in which auditory cortical populations encode movement and sound, consistent with predictive processing. These findings show that prediction-related signals reshape auditory cortex dynamics during behavior and reveal a cortical locus for the emergence of expectation and error.

Movement-Related Modulation in Mouse Auditory Cortex Is Widespread Yet Locally Diverse.

Neurons in the mouse auditory cortex are strongly influenced by behavior, including both suppression and enhancement of sound-evoked responses during movement. The mouse auditory cortex comprises multiple fields with different roles in sound processing and distinct connectivity to movement-related centers of the brain. Here, we asked whether movement-related modulation in male mice might differ across auditory cortical fields, thereby contributing to the heterogeneity of movement-related modulation at the single-cell level. We used wide-field calcium imaging to identify distinct cortical fields and cellular-resolution two-photon calcium imaging to visualize the activity of layer 2/3 excitatory neurons within each field. We measured each neuron's responses to three sound categories (pure tones, chirps, and amplitude-modulated white noise) as mice rested and ran on a non-motorized treadmill. We found that individual neurons in each cortical field typically respond to just one sound category. Some neurons are only active during rest and others during locomotion, and those that are responsive across conditions retain their sound-category tuning. The effects of locomotion on sound-evoked responses vary at the single-cell level, with both suppression and enhancement of neural responses, and the net modulatory effect of locomotion is largely conserved across cortical fields. Movement-related modulation in auditory cortex also reflects more complex behavioral patterns, including instantaneous running speed and nonlocomotor movements such as grooming and postural adjustments, with similar patterns seen across all auditory cortical fields. Our findings underscore the complexity of movement-related modulation throughout the mouse auditory cortex and indicate that movement-related modulation is a widespread phenomenon.

Sensation and expectation are embedded in mouse motor cortical activity.

During behavior, the motor cortex sends copies of motor-related signals to sensory cortices. It remains unclear whether these corollary discharge signals strictly encode movement or whether they also encode sensory experience and expectation. Here, we combine closed-loop behavior with large-scale physiology, projection-pattern specific recordings, and circuit perturbations to show that neurons in mouse secondary motor cortex (M2) encode sensation and are influenced by expectation. When a movement unexpectedly produces a sound, M2 becomes dominated by sound-evoked activity. Sound responses in M2 are inherited partially from the auditory cortex and are routed back to the auditory cortex, providing a path for the dynamic exchange of sensory-motor information during behavior. When the acoustic consequences of a movement become predictable, M2 responses to self-generated sounds are selectively gated off. These changes in single-cell responses are reflected in population dynamics, which are influenced by both sensation and expectation. Together, these findings reveal the rich embedding of sensory and expectation signals in motor cortical activity.

Stimulus-Specific Prediction Error Neurons in Mouse Auditory Cortex.

Comparing expectation with experience is an important neural computation performed throughout the brain and is a hallmark of predictive processing. Experiments that alter the sensory outcome of an animal's behavior reveal enhanced neural responses to unexpected self-generated stimuli, indicating that populations of neurons in sensory cortex may reflect prediction errors (PEs), mismatches between expectation and experience. However, enhanced neural responses to self-generated stimuli could also arise through nonpredictive mechanisms, such as the movement-based facilitation of a neuron's inherent sound responses. If sensory prediction error neurons exist in sensory cortex, it is unknown whether they manifest as general error responses, or respond with specificity to errors in distinct stimulus dimensions. To answer these questions, we trained mice of either sex to expect the outcome of a simple sound-generating behavior and recorded auditory cortex activity as mice heard either the expected sound or sounds that deviated from expectation in one of multiple distinct dimensions. Our data reveal that the auditory cortex learns to suppress responses to self-generated sounds along multiple acoustic dimensions simultaneously. We identify a distinct population of auditory cortex neurons that are not responsive to passive sounds or to the expected sound but that encode prediction errors. These prediction error neurons are abundant only in animals with a learned motor-sensory expectation, and encode one or two specific violations rather than a generic error signal. Together, these findings reveal that cortical predictions about self-generated sounds have specificity in multiple simultaneous dimensions and that cortical prediction error neurons encode specific violations from expectation.SIGNIFICANCE STATEMENT Audette et. al record neural activity in the auditory cortex while mice perform a sound-generating forelimb movement and measure neural responses to sounds that violate an animal's expectation in different ways. They find that predictions about self-generated sounds are highly specific across multiple stimulus dimensions and that a population of typically nonsound-responsive neurons respond to sounds that violate an animal's expectation in a specific way. These results identify specific prediction error (PE) signals in the mouse auditory cortex and suggest that errors may be calculated early in sensory processing.

Fish who act on impulse must learn the consequences.

Animals learn internal models that link specific behaviors to their anticipated sensory outcomes. In this issue of Neuron, Wallach and Sawtell1 discover that freely moving fish learn how the sensory outcome of a single behavior changes with local context.

Movement-related modulation in mouse auditory cortex is widespread yet locally diverse.

Neurons in the mouse auditory cortex are strongly influenced by behavior, including both suppression and enhancement of sound-evoked responses during movement. The mouse auditory cortex comprises multiple fields with different roles in sound processing and distinct connectivity to movement-related centers of the brain. Here, we asked whether movement-related modulation might differ across auditory cortical fields, thereby contributing to the heterogeneity of movement-related modulation at the single-cell level. We used wide-field calcium imaging to identify distinct cortical fields followed by cellular-resolution two-photon calcium imaging to visualize the activity of layer 2/3 excitatory neurons within each field. We measured each neuron's responses to three sound categories (pure tones, chirps, and amplitude modulated white noise) as mice rested and ran on a non-motorized treadmill. We found that individual neurons in each cortical field typically respond to just one sound category. Some neurons are only active during rest and others during locomotion, and those that are responsive across conditions retain their sound-category tuning. The effects of locomotion on sound-evoked responses vary at the single-cell level, with both suppression and enhancement of neural responses, and the net modulatory effect of locomotion is largely conserved across cortical fields. Movement-related modulation in auditory cortex also reflects more complex behavioral patterns, including instantaneous running speed and non-locomotor movements such as grooming and postural adjustments, with similar patterns seen across all auditory cortical fields. Our findings underscore the complexity of movement-related modulation throughout the mouse auditory cortex and indicate that movement-related modulation is a widespread phenomenon.

Stimulus-specific prediction error neurons in mouse auditory cortex.

Comparing expectation with experience is an important neural computation performed throughout the brain and is a hallmark of predictive processing. Experiments that alter the sensory outcome of an animal's behavior reveal enhanced neural responses to unexpected self-generated stimuli, indicating that populations of neurons in sensory cortex may reflect prediction errors - mismatches between expectation and experience. However, enhanced neural responses to self-generated stimuli could also arise through non-predictive mechanisms, such as the movement-based facilitation of a neuron's inherent sound responses. If sensory prediction error neurons exist in sensory cortex, it is unknown whether they manifest as general error responses, or respond with specificity to errors in distinct stimulus dimensions. To answer these questions, we trained mice to expect the outcome of a simple sound-generating behavior and recorded auditory cortex activity as mice heard either the expected sound or sounds that deviated from expectation in one of multiple distinct dimensions. Our data reveal that the auditory cortex learns to suppress responses to self-generated sounds along multiple acoustic dimensions simultaneously. We identify a distinct population of auditory cortex neurons that are not responsive to passive sounds or to the expected sound but that explicitly encode prediction errors. These prediction error neurons are abundant only in animals with a learned motor-sensory expectation, and encode one or two specific violations rather than a generic error signal.

Precise movement-based predictions in the mouse auditory cortex.

Many of the sensations experienced by an organism are caused by their own actions, and accurately anticipating both the sensory features and timing of self-generated stimuli is crucial to a variety of behaviors. In the auditory cortex, neural responses to self-generated sounds exhibit frequency-specific suppression, suggesting that movement-based predictions may be implemented early in sensory processing. However, it remains unknown whether this modulation results from a behaviorally specific and temporally precise prediction, nor is it known whether corresponding expectation signals are present locally in the auditory cortex. To address these questions, we trained mice to expect the precise acoustic outcome of a forelimb movement using a closed-loop sound-generating lever. Dense neuronal recordings in the auditory cortex revealed suppression of responses to self-generated sounds that was specific to the expected acoustic features, to a precise position within the movement, and to the movement that was coupled to sound during training. Prediction-based suppression was concentrated in L2/3 and L5, where deviations from expectation also recruited a population of prediction-error neurons that was otherwise unresponsive. Recording in the absence of sound revealed abundant movement signals in deep layers that were biased toward neurons tuned to the expected sound, as well as expectation signals that were present throughout the cortex and peaked at the time of expected auditory feedback. Together, these findings identify distinct populations of auditory cortical neurons with movement, expectation, and error signals consistent with a learned internal model linking an action to its specific acoustic outcome.

Reflections of action in sensory cortex.

Nearly every movement that one makes produces a corresponding set of sensations. The simple fact that much of our sensory world is driven by our own actions underscores one of the major computations that our brains execute every day: to interpret the sensory world even as we interact with and change it. It should not be surprising therefore that activity in sensory cortex reflects not only incoming sensory inputs but also ongoing movement and behavioral state. With a focus on the mouse as a model organism, this review highlights recent findings revealing the widespread modulation of sensory cortex across diverse movements, the circuitry through which movement-related inputs are integrated with sensory signals, and the computational and perceptual roles that motor-sensory integration may serve within the brain.

Signatures of Microevolutionary Processes in Phylogenetic Patterns.

Phylogenetic trees are representations of evolutionary relationships among species and contain signatures of the processes responsible for the speciation events they display. Inferring processes from tree properties, however, is challenging. To address this problem, we analyzed a spatially-explicit model of speciation where genome size and mating range can be controlled. We simulated parapatric and sympatric (narrow and wide mating range, respectively) radiations and constructed their phylogenetic trees, computing structural properties such as tree balance and speed of diversification. We showed that parapatric and sympatric speciation are well separated by these structural tree properties. Balanced trees with constant rates of diversification only originate in sympatry and genome size affected both the balance and the speed of diversification of the simulated trees. Comparison with empirical data showed that most of the evolutionary radiations considered to have developed in parapatry or sympatry are in good agreement with model predictions. Even though additional forces other than spatial restriction of gene flow, genome size, and genetic incompatibilities, do play a role in the evolution of species formation, the microevolutionary processes modeled here capture signatures of the diversification pattern of evolutionary radiations, regarding the symmetry and speed of diversification of lineages.

A cortical filter that learns to suppress the acoustic consequences of movement.

Sounds can arise from the environment and also predictably from many of our own movements, such as vocalizing, walking, or playing music. The capacity to anticipate these movement-related (reafferent) sounds and distinguish them from environmental sounds is essential for normal hearing1,2, but the neural circuits that learn to anticipate the often arbitrary and changeable sounds that result from our movements remain largely unknown. Here we developed an acoustic virtual reality (aVR) system in which a mouse learned to associate a novel sound with its locomotor movements, allowing us to identify the neural circuit mechanisms that learn to suppress reafferent sounds and to probe the behavioural consequences of this predictable sensorimotor experience. We found that aVR experience gradually and selectively suppressed auditory cortical responses to the reafferent frequency, in part by strengthening motor cortical activation of auditory cortical inhibitory neurons that respond to the reafferent tone. This plasticity is behaviourally adaptive, as aVR-experienced mice showed an enhanced ability to detect non-reafferent tones during movement. Together, these findings describe a dynamic sensory filter that involves motor cortical inputs to the auditory cortex that can be shaped by experience to selectively suppress the predictable acoustic consequences of movement.

How Movement Modulates Hearing.

Hearing is often viewed as a passive process: Sound enters the ear, triggers a cascade of activity through the auditory system, and culminates in an auditory percept. In contrast to a passive process, motor-related signals strongly modulate the auditory system from the eardrum to the cortex. The motor modulation of auditory activity is most well documented during speech and other vocalizations but also can be detected during a wide variety of other sound-generating behaviors. An influential idea is that these motor-related signals suppress neural responses to predictable movement-generated sounds, thereby enhancing sensitivity to environmental sounds during movement while helping to detect errors in learned acoustic behaviors, including speech and musicianship. Findings in humans, monkeys, songbirds, and mice provide new insights into the circuits that convey motor-related signals to the auditory system, while lending support to the idea that these signals function predictively to facilitate hearing and vocal learning.

The mutation-drift balance in spatially structured populations.

In finite populations the action of neutral mutations is balanced by genetic drift, leading to a stationary distribution of alleles that displays a transition between two different behaviors. For small mutation rates most individuals will carry the same allele at equilibrium, whereas for high mutation rates of the alleles will be randomly distributed with frequencies close to one half for a biallelic gene. For well-mixed haploid populations the mutation threshold is µc=1/2N, where N is the population size. In this paper we study how spatial structure affects this mutation threshold. Specifically, we study the stationary allele distribution for populations placed on regular networks where connected nodes represent potential mating partners. We show that the mutation threshold is sensitive to spatial structure only if the number of potential mates is very small. In this limit, the mutation threshold decreases substantially, increasing the diversity of the population at considerably low mutation rates. Defining kc as the degree of the network for which the mutation threshold drops to half of its value in well-mixed populations we show that kc grows slowly as a function of the population size, following a power law. Our calculations and simulations are based on the Moran model and on a mapping between the Moran model with mutations and the voter model with opinion makers.

Diploid versus haploid models of neutral speciation.

Neutral models of speciation based on isolation by distance and assortative mating, termed topopatric, have shown to be successful in describing abundance distributions and species-area relationships. Previous works have considered this type of process in the context of haploid genomes. Here we discuss the implementation of two schemes of dominance to analyze the effects of diploidy: a complete dominance model in which one allele dominates over the other and a perfect codominant model in which heterozygous genotypes give rise to a third phenotype. In the case of complete dominance, we observe that speciation requires stronger spatial inbreeding in comparison to the haploid model. For perfect codominance, instead, speciation demands stronger genetic assortativeness. Nevertheless, once speciation is established, the three models predict the same abundance distributions even at the quantitative level, revealing the robustness of the original mechanism to describe biodiversity features.

Binary dynamics on star networks under external perturbations.

We study a binary dynamical process that is a representation of the voter model with two candidates and opinion makers. The voters are represented by nodes of a network of social contacts with internal states labeled 0 or 1 and nodes that are connected can influence each other. The network is also perturbed by opinion makers, a set of external nodes whose states are frozen in 0 or 1 and that can influence all nodes of the network. The quantity of interest is the probability of finding m nodes in state 1 at time t. Here we study this process on star networks, which are simple representations of hubs found in complex systems, and compare the results with those obtained for networks that are fully connected. In both cases a transition from disordered to ordered equilibrium states is observed as the number of external nodes becomes small. For fully connected networks the probability distribution becomes uniform at the critical point. For star networks, on the other hand, we show that the equilibrium distribution splits in two peaks, reflecting the two possible states of the central node. We obtain approximate analytical solutions for the equilibrium distribution that clarify the role of the central node in the process. We show that the network topology also affects the time scale of oscillations in single realizations of the dynamics, which are much faster for the star network. Finally, extending the analysis to two stars we compare our results with simulations in simple scale-free networks.

Error catastrophe in populations under similarity-essential recombination.

Organisms are often more likely to exchange genetic information with others that are similar to themselves. One of the most widely accepted mechanisms of RNA virus recombination requires substantial sequence similarity between the parental RNAs and is termed similarity-essential recombination. This mechanism may be considered analogous to assortative mating, an important form of non-random mating that can be found in animals and plants. Here we study the dynamics of haplotype frequencies in populations evolving under similarity-essential recombination. Haplotypes are represented by a genome of B biallelic loci and the Hamming distance between individuals is used as a criterion for recombination. We derive the evolution equations for the haplotype frequencies assuming that recombination does not occur if the genetic distance is larger than a critical value G and that mutation occurs at a rate µ per locus. Additionally, uniform crossover is considered. Although no fitness is directly associated to the haplotypes, we show that frequency-dependent selection emerges dynamically and governs the haplotype distribution. A critical mutation rate µc can be identified as the error threshold transition, beyond which this selective information cannot be stored. For µ<µc the distribution consists of a dominant sequence surrounded by a cloud of closely related sequences, characterizing a quasispecies. For µ>µc the distribution becomes uniform, with all haplotypes having the same frequency. In the case of extreme assortativeness, where individuals only recombine with others identical to themselves (G=0), the error threshold results µc=1/4, independently of the genome size. For weak assortativity (G=B-1)µc=2(-(B+1)) and for the case of no assortativity (G=B) µc=0. We compute the mutation threshold for 0

Motor-related signals in the auditory system for listening and learning.

In the auditory system, corollary discharge signals are theorized to facilitate normal hearing and the learning of acoustic behaviors, including speech and music. Despite clear evidence of corollary discharge signals in the auditory cortex and their presumed importance for hearing and auditory-guided motor learning, the circuitry and function of corollary discharge signals in the auditory cortex are not well described. In this review, we focus on recent developments in the mouse and songbird that provide insights into the circuitry that transmits corollary discharge signals to the auditory system and the function of these signals in the context of hearing and vocal learning.