Neuronal dynamics of predictive coding

A sensory-motor theory of the neocortex

Recent neurophysiological and neuroanatomical studies suggest a close interaction between sensory and motor processes across the neocortex. Here, I propose that the neocortex implements active predictive coding (APC): each cortical area estimates both latent sensory states and actions (including potentially abstract actions internal to the cortex), and the cortex as a whole predicts the consequences of actions at multiple hierarchical levels. Feedback from higher areas modulates the dynamics of state and action networks in lower areas. I show how the same APC architecture can explain (1) how we recognize an object and its parts using eye movements, (2) why perception seems stable despite eye movements, (3) how we learn compositional representations, for example, part-whole hierarchies, (4) how complex actions can be planned using simpler actions, and (5) how we form episodic memories of sensory-motor experiences and learn abstract concepts such as a family tree. I postulate a mapping of the APC model to the laminar architecture of the cortex and suggest possible roles for cortico-cortical and cortico-subcortical pathways.

Goal-oriented representations in the human hippocampus during planning and navigation

Recent work in cognitive and systems neuroscience has suggested that the hippocampus might support planning, imagination, and navigation by forming cognitive maps that capture the abstract structure of physical spaces, tasks, and situations. Navigation involves disambiguating similar contexts, and the planning and execution of a sequence of decisions to reach a goal. Here, we examine hippocampal activity patterns in humans during a goal-directed navigation task to investigate how contextual and goal information are incorporated in the construction and execution of navigational plans. During planning, hippocampal pattern similarity is enhanced across routes that share a context and a goal. During navigation, we observe prospective activation in the hippocampus that reflects the retrieval of pattern information related to a key-decision point. These results suggest that, rather than simply representing overlapping associations or state transitions, hippocampal activity patterns are shaped by context and goals.

Rapid learning of predictive maps with STDP and theta phase precession

The predictive map hypothesis is a promising candidate principle for hippocampal function. A favoured formalisation of this hypothesis, called the successor representation, proposes that each place cell encodes the expected state occupancy of its target location in the near future. This predictive framework is supported by behavioural as well as electrophysiological evidence and has desirable consequences for both the generalisability and efficiency of reinforcement learning algorithms. However, it is unclear how the successor representation might be learnt in the brain. Error-driven temporal difference learning, commonly used to learn successor representations in artificial agents, is not known to be implemented in hippocampal networks. Instead, we demonstrate that spike-timing dependent plasticity (STDP), a form of Hebbian learning, acting on temporally compressed trajectories known as 'theta sweeps', is sufficient to rapidly learn a close approximation to the successor representation. The model is biologically plausible - it uses spiking neurons modulated by theta-band oscillations, diffuse and overlapping place cell-like state representations, and experimentally matched parameters. We show how this model maps onto known aspects of hippocampal circuitry and explains substantial variance in the temporal difference successor matrix, consequently giving rise to place cells that demonstrate experimentally observed successor representation-related phenomena including backwards expansion on a 1D track and elongation near walls in 2D. Finally, our model provides insight into the observed topographical ordering of place field sizes along the dorsal-ventral axis by showing this is necessary to prevent the detrimental mixing of larger place fields, which encode longer timescale successor representations, with more fine-grained predictions of spatial location.

Do after "not to do": Deinhibition in cognitive control

In daily life, we often need to inhibit a certain behavior or thought; however, sometimes we need to remove inhibition (deinhibition). Numerous studies have examined inhibition control, but it is unclear how deinhibition functions. In Experiment 1, we adopted a modified stop-signal task in which participants were instructed to immediately stop the prepared response to a stimulus appended by an accidental signal. The results showed that when the preceding trial was a stop-signal trial and participants successfully inhibited the action to the stimulus, the reaction time (RT) for the repeated stimuli in the current trial was significantly longer than that of the switched stimuli, reflecting the cost of deinhibition. Deinhibition ability is correlated with inhibitory control and cognitive flexibility. In Experiment 2, we manipulated stimulus onset asynchrony (SOA) between presentation of the stimuli and the stopping signals to exclude the interference of the signal preparation effect on the deinhibition cost. These findings suggest that an individual's deinhibition ability, as a previously ignored subcomponent of cognitive control, may play an important role in human adaptive behavior.

Cognitive Neural Mechanism of Backward Inhibition and Deinhibition: A Review

Task switching is one of the typical paradigms to study cognitive control. When switching back to a recently inhibited task (e.g., “A” in an ABA sequence), the performance is often worse compared to a task without N-2 task repetitions (e.g., CBA). This difference is called the backward inhibitory effect (BI effect), which reflects the process of overcoming residual inhibition from a recently performed task (i.e., deinhibition). The neural mechanism of backward inhibition and deinhibition has received a lot of attention in the past decade. Multiple brain regions, including the frontal lobe, parietal, basal ganglia, and cerebellum, are activated during deinhibition. The event-related potentials (ERP) studies have shown that deinhibition process is reflected in the P1/N1 and P3 components, which might be related to early attention control, context updating, and response selection, respectively. Future research can use a variety of new paradigms to separate the neural mechanisms of BI and deinhibition.

Implicit learning of temporal behavior in complex dynamic environments

Humans can automatically detect and learn to exploit repeated aspects (regularities) of the environment. Timing research suggests that such learning is not only used to anticipate what will happen, but also when it will happen. However, in timing experiments, the intervals to be timed are presented in isolation from other stimuli and explicitly cued, contrasting with naturalistic environments in which intervals are embedded in a constant stream of events and individuals are hardly aware of them. It is unclear whether laboratory findings from timing research translate to a more ecologically valid, implicit environment. Here we show in a game-like experiment, specifically designed to measure naturalistic behavior, that participants implicitly use regular intervals to anticipate future events, even when these intervals are constantly interrupted by irregular yet behaviorally relevant events. This finding extends previous research by showing that individuals not only detect such regularities but can also use this knowledge to decide when to act in a complex environment. Furthermore, this finding demonstrates that this type of learning can occur independently from the ordinal sequence of motor actions, which contrasts this work with earlier motor learning studies. Taken together, our results demonstrate that regularities in the time between events are implicitly monitored and used to predict and act on what happens when, thereby showing that laboratory findings from timing research can generalize to naturalistic environments. Additionally, with the development of our game-like experiment, we demonstrate an approach to test cognitive theories in less controlled, ecologically more valid environments.

Remembering rewarding futures: A simulation-selection model of the hippocampus

Despite tremendous progress, the neural circuit dynamics underlying hippocampal mnemonic processing remain poorly understood. We propose a new model for hippocampal function-the simulation-selection model-based on recent experimental findings and neuroecological considerations. Under this model, the mammalian hippocampus evolved to simulate and evaluate arbitrary navigation sequences. Specifically, we suggest that CA3 simulates unexperienced navigation sequences in addition to remembering experienced ones, and CA1 selects from among these CA3-generated sequences, reinforcing those that are likely to maximize reward during offline idling states. High-value sequences reinforced in CA1 may allow flexible navigation toward a potential rewarding location during subsequent navigation. We argue that the simulation-selection functions of the hippocampus have evolved in mammals mostly because of the unique navigational needs of land mammals. Our model may account for why the mammalian hippocampus has evolved not only to remember, but also to imagine episodes, and how this might be implemented in its neural circuits.

Hippocampal Neural Circuits Respond to Optogenetic Pacing of Theta Frequencies by Generating Accelerated Oscillation Frequencies

Biological oscillations can be controlled by a small population of rhythmic pacemaker cells, or in the brain, they also can emerge from complex cellular and circuit-level interactions. Whether and how these mechanisms are combined to give rise to oscillatory patterns that govern cognitive function are not well understood. For example, the activity of hippocampal networks is temporally coordinated by a 7- to 9-Hz local field potential (LFP) theta rhythm, yet many individual cells decouple from the LFP frequency to oscillate at frequencies ∼1 Hz higher. To better understand the network interactions that produce these complex oscillatory patterns, we asked whether the relative frequency difference between LFP and individual cells is retained when the LFP frequency is perturbed experimentally. We found that rhythmic optogenetic stimulation of medial septal GABAergic neurons controlled the hippocampal LFP frequency outside of the endogenous theta range, even during behavioral states when endogenous mechanisms would otherwise have generated 7- to 9-Hz theta oscillations. While the LFP frequency matched the optogenetically induced stimulation frequency, the oscillation frequency of individual hippocampal cells remained broadly distributed, and in a subset of cells including interneurons, it was accelerated beyond the new base LFP frequency. The inputs from septal GABAergic neurons to the hippocampus, therefore, do not appear to directly control the cellular oscillation frequency but rather engage cellular and circuit mechanisms that accelerate the rhythmicity of individual cells. Thus, theta oscillations are an example of cortical oscillations that combine inputs from a subcortical pacemaker with local computations to generate complex oscillatory patterns that support cognitive functions.

Social and Contextual Constraints on Embodied Perception

A number of papers have challenged research on physiological and psychological influences on perception by claiming to show that such findings can be explained by nonperceptual factors such as demand characteristics. Relatedly, calls for separating perception from judgment have been issued. However, such efforts fail to consider key processes known to shape judgment processes: people's inability to report accurately on their judgments, conversational dynamics of experimental research contexts, and misattribution and discounting processes. Indeed, the fact that initially observed effects of embodied influences disappear is predicted by an extensive amount of literature on judgments studied within social psychology. Thus, findings from such studies suggest that the initially presumed underlying processes are at work-namely, functional considerations that are informative in the context of preparing the body for action. In this article, I provide suggestions on how to conduct research on perception within the social constraints of experimental contexts.

Learning of anticipatory responses in single neurons of the human medial temporal lobe

Neuronal processes underlying the formation of new associations in the human brain are not yet well understood. Here human participants, implanted with depth electrodes in the brain, learned arbitrary associations between images presented in an ordered, predictable sequence. During learning we recorded from medial temporal lobe (MTL) neurons that responded to at least one of the pictures in the sequence (the preferred stimulus). We report that as a result of learning, single MTL neurons show asymmetric shifts in activity and start firing earlier in the sequence in anticipation of their preferred stimulus. These effects appear relatively early in learning, after only 11 exposures to the stimulus sequence. The anticipatory neuronal responses emerge while the subjects became faster in reporting the next item in the sequence. These results demonstrate flexible representations that could support learning of new associations between stimuli in a sequence, in single neurons in the human MTL.

The medial entorhinal cortex is necessary for temporal organization of hippocampal neuronal activity

The superficial layers of the medial entorhinal cortex (MEC) are the major input to the hippocampus. The high proportion of spatially modulated cells, including grid cells and border cells, in these layers suggests that the MEC inputs to the hippocampus are critical for the representation of space in the hippocampus. However, selective manipulations of the MEC do not completely abolish hippocampal spatial firing. To therefore determine whether other hippocampal firing characteristics depend more critically on MEC inputs, we recorded from hippocampal CA1 cells in rats with MEC lesions. Strikingly, theta phase precession was substantially disrupted, even during periods of stable spatial firing. Our findings indicate that MEC inputs to the hippocampus are required for the temporal organization of hippocampal firing patterns and suggest that cognitive functions that depend on precise neuronal sequences within the hippocampal theta cycle are particularly dependent on the MEC.

Speed controls the amplitude and timing of the hippocampal gamma rhythm

Cortical and hippocampal gamma oscillations have been implicated in many behavioral tasks. The hippocampus is required for spatial navigation where animals run at varying speeds. Hence we tested the hypothesis that the gamma rhythm could encode the running speed of mice. We found that the amplitude of slow (20–45 Hz) and fast (45–120 Hz) gamma rhythms in the hippocampal local field potential (LFP) increased with running speed. The speed-dependence of gamma amplitude was restricted to a narrow range of theta phases where gamma amplitude was maximal, called the preferred theta phase of gamma. The preferred phase of slow gamma precessed to lower values with increasing running speed. While maximal fast and slow gamma occurred at coincident phases of theta at low speeds, they became progressively more theta-phase separated with increasing speed. These results demonstrate a novel influence of speed on the amplitude and timing of the hippocampal gamma rhythm which could contribute to learning of temporal sequences and navigation.

Routing the flow of sensory signals using plastic responses to bursts and isolated spikes: experiment and theory

Processing complex sensory environments efficiently requires a diverse array of neural coding strategies. Neural codes relying on specific temporal patterning of action potentials may offer advantages over using solely spike rate codes. In particular, stimulus-dependent burst firing may carry additional information that isolated spikes do not. We use the well characterized electrosensory system of weakly electric fish to address how stimulus-dependent burst firing can determine the flow of information in feedforward neural circuits with different forms of short-term synaptic plasticity. Pyramidal cells in the electrosensory lateral line lobe burst in response to low-frequency, local (prey) signals. We show that the ability of pyramidal cells to code for local signals in the presence of additional high-frequency, global (communication) stimuli is uncompromised, while burst firing is reduced. We developed a bursting neuron model to understand how these effects, in particular noise-induced burst suppression, arise from interplay between incoming sensory signals and intrinsic neuronal dynamics. Finally, we examined how postsynaptic target populations preferentially respond to one of the two sensory mixtures (local vs local plus global) depending on whether the populations are in receipt of facilitating or depressing synapses. This form of feedforward neural architecture may allow for efficient information flow in the same neural pathway via either isolated or burst spikes, where the mechanisms by which stimuli are encoded are adaptable and sensitive to a diverse array of stimulus and contextual mixtures.

Predictive coding

Predictive coding is a unifying framework for understanding redundancy reduction and efficient coding in the nervous system. By transmitting only the unpredicted portions of an incoming sensory signal, predictive coding allows the nervous system to reduce redundancy and make full use of the limited dynamic range of neurons. Starting with the hypothesis of efficient coding as a design principle in the sensory system, predictive coding provides a functional explanation for a range of neural responses and many aspects of brain organization. The lateral and temporal antagonism in receptive fields in the retina and lateral geniculate nucleus occur naturally as a consequence of predictive coding of natural images. In the higher visual system, predictive coding provides an explanation for oriented receptive fields and contextual effects as well as the hierarchical reciprocally connected organization of the cortex. Predictive coding has also been found to be consistent with a variety of neurophysiological and psychophysical data obtained from different areas of the brain. WIREs Cogni Sci 2011 2 580-593 DOI: 10.1002/wcs.142 For further resources related to this article, please visit the WIREs website.

Effects of timing and movement uncertainty implicate the temporo-parietal junction in the prediction of forthcoming motor actions

The concept of predictive coding supposes the brain to build predictions of forthcoming events in order to decrease the computational load, thereby facilitating efficient reactions. In contrast, increasing uncertainty, i.e., lower predictability, should increase reaction time and neural activity due to reactive processing and believe updating. We used functional magnetic resonance imaging (fMRI) to scan subjects reacting to briefly presented arrows pointing to either side by pressing a button with the corresponding index finger. Predictability of these stimuli was manipulated along the independently varied factors "response type" (known hand or random, i.e., unknown order) and "timing" (fixed or variable intervals between stimuli). Behavioural data showed a significant reaction-time advantage when either factor was predictable, confirming the hypothesised reduction in computational load. On the neural level, only the right temporo-parietal junction showed enhanced activation upon both increased task and timing uncertainty. Moreover, activity in this region also positively correlated with reaction time. There was, however, a dissociation between both factors in the frontal lobe, as increased timing uncertainty recruited right BA 44, whereas increased response uncertainty activated the right ventral premotor cortex, the pre-SMA and the DLPFC. In line with the theoretical framework of predictive coding as a load-saving mechanism no brain region showed significantly increased activity in the lower uncertainty conditions or correlated negatively with reaction times. This study hence provided behavioural and neuroimaging evidence for predictive motor coding and points to a key role of the right temporo-parietal junction in its implementation.

Hippocampal Erk mechanisms linking prediction error to fear extinction: roles of shock expectancy and contextual aversive valence

Extinction of fear requires learning that anticipated aversive events no longer occur. Animal models reveal that sustained phosphorylation of the extracellular signal-regulated kinase (Erk) in hippocampal CA1 neurons plays an important role in this process. However, the key signals triggering and regulating the activity of Erk are not known. By varying the degree of expected and delivered aversive reinforcement, we demonstrate that Erk specifically responds to prediction errors of contextual aversive events. An increase of somatonuclear phospho-Erk (pErk) within principal CA1 neurons was observed only when the expectation of contextual foot shock was violated, but not when the context was consistently nonreinforced or reinforced by foot shock. The rate of error detection, Erk signaling, and fear extinction markedly depended on shock expectancy and the aversive valence of the context, as revealed by comparison of groups trained with single, continuous, or partial reinforcement. On the basis of these findings, the hippocampal Erk response to prediction errors of aversive outcome is proposed as a unique mechanism of fear extinction. Improving the detection and processing of these errors has the potential to attenuate fear responses in patients with anxiety disorders.

The emergence of grid cells: Intelligent design or just adaptation?

Individual medial entorhinal cortex (mEC) 'grid' cells provide a representation of space that appears to be essentially invariant across environments, modulo simple transformations, in contrast to multiple, rapidly acquired hippocampal maps; it may therefore be established gradually during rodent development. We explore with a simplified mathematical model the possibility that the self-organization of multiple grid fields into a triangular grid pattern may be a single-cell process, driven by firing rate adaptation and slowly varying spatial inputs. A simple analytical derivation indicates that triangular grids are favored asymptotic states of the self-organizing system, and computer simulations confirm that such states are indeed reached during a model learning process, provided it is sufficiently slow to effectively average out fluctuations. The interactions among local ensembles of grid units serve solely to stabilize a common grid orientation. Spatial information, in the real mEC network, may be provided by any combination of feedforward cortical afferents and feedback hippocampal projections from place cells, since either input alone is likely sufficient to yield grid fields.

Hippocampus-independent phase precession in entorhinal grid cells

Theta-phase precession in hippocampal place cells is one of the best-studied experimental models of temporal coding in the brain. Theta-phase precession is a change in spike timing in which the place cell fires at progressively earlier phases of the extracellular theta rhythm as the animal crosses the spatially restricted firing field of the neuron. Within individual theta cycles, this phase advance results in a compressed replication of the firing sequence of consecutively activated place cells along the animal's trajectory, at a timescale short enough to enable spike-time-dependent plasticity between neurons in different parts of the sequence. The neuronal circuitry required for phase precession has not yet been established. The fact that phase precession can be seen in hippocampal output stuctures such as the prefrontal cortex suggests either that efferent structures inherit the precession from the hippocampus or that it is generated locally in those structures. Here we show that phase precession is expressed independently of the hippocampus in spatially modulated grid cells in layer II of medial entorhinal cortex, one synapse upstream of the hippocampus. Phase precession is apparent in nearly all principal cells in layer II but only sparsely in layer III. The precession in layer II is not blocked by inactivation of the hippocampus, suggesting that the phase advance is generated in the grid cell network. The results point to possible mechanisms for grid formation and raise the possibility that hippocampal phase precession is inherited from entorhinal cortex.

A public confession: the retina trumpets its failed predictions

To compensate for delays of phototransduction, the retina anticipates the future by extrapolating the position of a moving object. But what if the object's motion changes, and the extrapolation is wrong? In this issue of Neuron, Schwartz and colleagues show that these prediction failures trigger a large burst of firing that helps to rapidly correct the neural representation of the object's new position.

Integrated mechanisms of anticipation and rate-of-change computations in cortical circuits

Local neocortical circuits are characterized by stereotypical physiological and structural features that subserve generic computational operations. These basic computations of the cortical microcircuit emerge through the interplay of neuronal connectivity, cellular intrinsic properties, and synaptic plasticity dynamics. How these interacting mechanisms generate specific computational operations in the cortical circuit remains largely unknown. Here, we identify the neurophysiological basis of both the rate of change and anticipation computations on synaptic inputs in a cortical circuit. Through biophysically realistic computer simulations and neuronal recordings, we show that the rate-of-change computation is operated robustly in cortical networks through the combination of two ubiquitous brain mechanisms: short-term synaptic depression and spike-frequency adaptation. We then show how this rate-of-change circuit can be embedded in a convergently connected network to anticipate temporally incoming synaptic inputs, in quantitative agreement with experimental findings on anticipatory responses to moving stimuli in the primary visual cortex. Given the robustness of the mechanism and the widespread nature of the physiological machinery involved, we suggest that rate-of-change computation and temporal anticipation are principal, hard-wired functions of neural information processing in the cortical microcircuit.

The cerebral cortex is the region of the brain whose intricate connectivity and physiology is thought to subserve most computations required for effective action in mammals. Through biophysically realistic computer simulation and experimental recordings in brain tissue, the authors show how a specific combination of physiological mechanisms often found in neurons of the cortex transforms an input signal into another signal that represents the rate of change of the slower components of the input. This is the first report of a neurobiological implementation of an approximate mathematical derivative in the cortex. Further, such a signal integrates naturally into a neurobiologically simple network that is able to generate a linear prediction of its inputs. Anticipation of information is a primary concern of spatially extended organisms which are subject to neural delays, and it has been demonstrated at various different levels: from the retina to sensori-motor integration. We present here a simple and general mechanism for anticipation that can operate incrementally within local circuits of the cortex, to compensate for time-consuming computations and conduction delays and thus contribute to effective real-time action.

Backward shift of head direction tuning curves of the anterior thalamus: comparison with CA1 place fields

The head direction cell system is composed of multiple regions associated with the hippocampal formation. The dynamics of head direction tuning curves (HDTCs) were compared with those of hippocampal place fields. In both familiar and cue-altered environments, as a rat ran an increasing number of laps on a track, the center of mass (COM) of the HDTC tended to shift backward, similar to shifting observed in place cells. However, important differences existed between these cells in terms of the shift patterns relative to the cue-altered conditions, the proportion of backward versus forward shifts, and the time course of shift resetting. The demonstration of backward COM shifts in head direction cells and place cells suggests that similar plasticity mechanisms (such as temporally asymmetric LTP induction or spike timing-dependent plasticity) may be at work in both brain systems, and these processes may reflect a general mechanism for storing learned sequences of neural activity patterns.

Attending to remember and remembering to attend

Attention and memory are intimately linked. Two functional imaging studies in this issue of Neuron provide novel evidence for this powerful, reciprocal relationship. Turk-Browne and colleagues report that attention simultaneously facilitates the formation of both implicit and explicit memories, while Summerfield and colleagues demonstrate that memory for the past can guide the allocation of attention in the present. Together, these elegant studies reveal bidirectional interactions between attention and memory.

What can a neuron learn with spike-timing-dependent plasticity?

Spiking neurons are very flexible computational modules, which can implement with different values of their adjustable synaptic parameters an enormous variety of different transformations F from input spike trains to output spike trains. We examine in this letter the question to what extent a spiking neuron with biologically realistic models for dynamic synapses can be taught via spike-timing-dependent plasticity (STDP) to implement a given transformation F. We consider a supervised learning paradigm where during training, the output of the neuron is clamped to the target signal (teacher forcing). The well-known perceptron convergence theorem asserts the convergence of a simple supervised learning algorithm for drastically simplified neuron models (McCulloch-Pitts neurons). We show that in contrast to the perceptron convergence theorem, no theoretical guarantee can be given for the convergence of STDP with teacher forcing that holds for arbitrary input spike patterns. On the other hand, we prove that average case versions of the perceptron convergence theorem hold for STDP in the case of uncorrelated and correlated Poisson input spike trains and simple models for spiking neurons. For a wide class of cross-correlation functions of the input spike trains, the resulting necessary and sufficient condition can be formulated in terms of linear separability, analogously as the well-known condition of learnability by perceptrons. However, the linear separability criterion has to be applied here to the columns of the correlation matrix of the Poisson input. We demonstrate through extensive computer simulations that the theoretically predicted convergence of STDP with teacher forcing also holds for more realistic models for neurons, dynamic synapses, and more general input distributions. In addition, we show through computer simulations that these positive learning results hold not only for the common interpretation of STDP, where STDP changes the weights of synapses, but also for a more realistic interpretation suggested by experimental data where STDP modulates the initial release probability of dynamic synapses.

A theory of cortical responses

This article concerns the nature of evoked brain responses and the principles underlying their generation. We start with the premise that the sensory brain has evolved to represent or infer the causes of changes in its sensory inputs. The problem of inference is well formulated in statistical terms. The statistical fundaments of inference may therefore afford important constraints on neuronal implementation. By formulating the original ideas of Helmholtz on perception, in terms of modern-day statistical theories, one arrives at a model of perceptual inference and learning that can explain a remarkable range of neurobiological facts.It turns out that the problems of inferring the causes of sensory input (perceptual inference) and learning the relationship between input and cause (perceptual learning) can be resolved using exactly the same principle. Specifically, both inference and learning rest on minimizing the brain's free energy, as defined in statistical physics. Furthermore, inference and learning can proceed in a biologically plausible fashion. Cortical responses can be seen as the brain's attempt to minimize the free energy induced by a stimulus and thereby encode the most likely cause of that stimulus. Similarly, learning emerges from changes in synaptic efficacy that minimize the free energy, averaged over all stimuli encountered. The underlying scheme rests on empirical Bayes and hierarchical models of how sensory input is caused. The use of hierarchical models enables the brain to construct prior expectations in a dynamic and context-sensitive fashion. This scheme provides a principled way to understand many aspects of cortical organization and responses. The aim of this article is to encompass many apparently unrelated anatomical, physiological and psychophysical attributes of the brain within a single theoretical perspective. In terms of cortical architectures, the theoretical treatment predicts that sensory cortex should be arranged hierarchically, that connections should be reciprocal and that forward and backward connections should show a functional asymmetry (forward connections are driving, whereas backward connections are both driving and modulatory). In terms of synaptic physiology, it predicts associative plasticity and, for dynamic models, spike-timing-dependent plasticity. In terms of electrophysiology, it accounts for classical and extra classical receptive field effects and long-latency or endogenous components of evoked cortical responses. It predicts the attenuation of responses encoding prediction error with perceptual learning and explains many phenomena such as repetition suppression, mismatch negativity (MMN) and the P300 in electroencephalography. In psychophysical terms, it accounts for the behavioural correlates of these physiological phenomena, for example, priming and global precedence. The final focus of this article is on perceptual learning as measured with the MMN and the implications for empirical studies of coupling among cortical areas using evoked sensory responses.

The effect of learning on the function of monkey extrastriate visual cortex

One of the most remarkable capabilities of the adult brain is its ability to learn and continuously adapt to an ever-changing environment. While many studies have documented how learning improves the perception and identification of visual stimuli, relatively little is known about how it modifies the underlying neural mechanisms. We trained monkeys to identify natural images that were degraded by interpolation with visual noise. We found that learning led to an improvement in monkeys' ability to identify these indeterminate visual stimuli. We link this behavioral improvement to a learning-dependent increase in the amount of information communicated by V4 neurons. This increase was mediated by a specific enhancement in neural activity. Our results reveal a mechanism by which learning increases the amount of information that V4 neurons are able to extract from the visual environment. This suggests that V4 plays a key role in resolving indeterminate visual inputs by coordinated interaction between bottom-up and top-down processing streams.

In monkeys trained to identify natural images embedded in noise, changes are seen in the information about learned stimuli signaled by V4 neurons

Computational constraints between retrieving the past and predicting the future, and the CA3-CA1 differentiation

The differentiation between the CA3 and CA1 fields of the mammalian hippocampus is one of the salient traits that set it apart from the organization of the homologue medial wall in reptiles and birds. CA3 is widely thought to function as an autoassociator, but what do we need CA1 for? Based on evidence for a specific role of CA1 in temporal processing, I have explored the hypothesis that the differentiation between CA3 and CA1 may help solve a computational conflict. The conflict is between pattern completion, or integrating current sensory information on the basis of memory, and prediction, or moving from one pattern to the next in a stored sequence. CA3 would take care of the former, while CA1 would concentrate on the latter. I have found the hypothesis to be only weakly supported by neural network simulations. The conflict indeed exists, but two mechanisms that would relate more directly to a functional CA3-CA1 differentiation were found unable to produce genuine prediction. Instead, a simple mechanism based on firing frequency adaptation in pyramidal cells was found to be sufficient for prediction, with the degree of adaptation as the crucial parameter balancing retrieval with prediction. The differentiation between the architectures of CA3 and CA1 has a minor but significant, and positive, effect on this balance. In particular, for a fixed anticipatory interval in the model, it increases significantly the information content of hippocampal outputs. There may therefore be just a simple quantitative advantage in differentiating the connectivity of the two fields. Moreover, different degrees of adaptation in CA3 and CA1 cells were not found to lead to better performance, further undermining the notion of a functional dissociation.

Role of experience and oscillations in transforming a rate code into a temporal code

In the vast majority of brain areas, the firing rates of neurons, averaged over several hundred milliseconds to several seconds, can be strongly modulated by, and provide accurate information about, properties of their inputs. This is referred to as the rate code. However, the biophysical laws of synaptic plasticity require precise timing of spikes over short timescales (<10 ms). Hence it is critical to understand the physiological mechanisms that can generate precise spike timing in vivo, and the relationship between such a temporal code and a rate code. Here we propose a mechanism by which a temporal code can be generated through an interaction between an asymmetric rate code and oscillatory inhibition. Consistent with the predictions of our model, the rate and temporal codes of hippocampal pyramidal neurons are highly correlated. Furthermore, the temporal code becomes more robust with experience. The resulting spike timing satisfies the temporal order constraints of hebbian learning. Thus, oscillations and receptive field asymmetry may have a critical role in temporal sequence learning.