Hierarchical models in the brain

The predictive mind and the experience of visual art work

Among the main challenges of the predictive brain/mind concept is how to link prediction at the neural level to prediction at the cognitive-psychological level and finding conceptually robust and empirically verifiable ways to harness this theoretical framework toward explaining higher-order mental and cognitive phenomena, including the subjective experience of aesthetic and symbolic forms. Building on the tentative prediction error account of visual art, this article extends the application of the predictive coding framework to the visual arts. It does so by linking this theoretical discussion to a subjective, phenomenological account of how a work of art is experienced. In order to engage more deeply with a work of art, viewers must be able to tune or adapt their prediction mechanism to recognize art as a specific class of objects whose ontological nature defies predictability, and they must be able to sustain a productive flow of predictions from low-level sensory, recognitional to abstract semantic, conceptual, and affective inferences. The affective component of the process of predictive error optimization that occurs when a viewer enters into dialog with a painting is constituted both by activating the affective affordances within the image and by the affective consequences of prediction error minimization itself. The predictive coding framework also has implications for the problem of the culturality of vision. A person's mindset, which determines what top-down expectations and predictions are generated, is co-constituted by culture-relative skills and knowledge, which form hyperpriors that operate in the perception of art.

Precision and neuronal dynamics in the human posterior parietal cortex during evidence accumulation

Primate studies show slow ramping activity in posterior parietal cortex (PPC) neurons during perceptual decision-making. These findings have inspired a rich theoretical literature to account for this activity. These accounts are largely unrelated to Bayesian theories of perception and predictive coding, a related formulation of perceptual inference in the cortical hierarchy. Here, we tested a key prediction of such hierarchical inference, namely that the estimated precision (reliability) of information ascending the cortical hierarchy plays a key role in determining both the speed of decision-making and the rate of increase of PPC activity. Using dynamic causal modelling of magnetoencephalographic (MEG) evoked responses, recorded during a simple perceptual decision-making task, we recover ramping-activity from an anatomically and functionally plausible network of regions, including early visual cortex, the middle temporal area (MT) and PPC. Precision, as reflected by the gain on pyramidal cell activity, was strongly correlated with both the speed of decision making and the slope of PPC ramping activity. Our findings indicate that the dynamics of neuronal activity in the human PPC during perceptual decision-making recapitulate those observed in the macaque, and in so doing we link observations from primate electrophysiology and human choice behaviour. Moreover, the synaptic gain control modulating these dynamics is consistent with predictive coding formulations of evidence accumulation.

Active inference, eye movements and oculomotor delays

This paper considers the problem of sensorimotor delays in the optimal control of (smooth) eye movements under uncertainty. Specifically, we consider delays in the visuo-oculomotor loop and their implications for active inference. Active inference uses a generalisation of Kalman filtering to provide Bayes optimal estimates of hidden states and action in generalised coordinates of motion. Representing hidden states in generalised coordinates provides a simple way of compensating for both sensory and oculomotor delays. The efficacy of this scheme is illustrated using neuronal simulations of pursuit initiation responses, with and without compensation. We then consider an extension of the generative model to simulate smooth pursuit eye movements-in which the visuo-oculomotor system believes both the target and its centre of gaze are attracted to a (hidden) point moving in the visual field. Finally, the generative model is equipped with a hierarchical structure, so that it can recognise and remember unseen (occluded) trajectories and emit anticipatory responses. These simulations speak to a straightforward and neurobiologically plausible solution to the generic problem of integrating information from different sources with different temporal delays and the particular difficulties encountered when a system-like the oculomotor system-tries to control its environment with delayed signals.

Construct validation of a DCM for resting state fMRI

Recently, there has been a lot of interest in characterising the connectivity of resting state brain networks. Most of the literature uses functional connectivity to examine these intrinsic brain networks. Functional connectivity has well documented limitations because of its inherent inability to identify causal interactions. Dynamic causal modelling (DCM) is a framework that allows for the identification of the causal (directed) connections among neuronal systems — known as effective connectivity. This technical note addresses the validity of a recently proposed DCM for resting state fMRI – as measured in terms of their complex cross spectral density – referred to as spectral DCM. Spectral DCM differs from (the alternative) stochastic DCM by parameterising neuronal fluctuations using scale free (i.e., power law) forms, rendering the stochastic model of neuronal activity deterministic. Spectral DCM not only furnishes an efficient estimation of model parameters but also enables the detection of group differences in effective connectivity, the form and amplitude of the neuronal fluctuations or both. We compare and contrast spectral and stochastic DCM models with endogenous fluctuations or state noise on hidden states. We used simulated data to first establish the face validity of both schemes and show that they can recover the model (and its parameters) that generated the data. We then used Monte Carlo simulations to assess the accuracy of both schemes in terms of their root mean square error. We also simulated group differences and compared the ability of spectral and stochastic DCMs to identify these differences. We show that spectral DCM was not only more accurate but also more sensitive to group differences. Finally, we performed a comparative evaluation using real resting state fMRI data (from an open access resource) to study the functional integration within default mode network using spectral and stochastic DCMs.

•

This paper provides construct validation of spectral DCM against stochastic DCM.

•

Spectral DCM is shown to be more accurate than stochastic DCM in terms of root mean square error.

•

Spectral DCM is shown to be more sensitive at identifying group differences.

The medial temporal lobe-conduit of parallel connectivity: a model for attention, memory, and perception

Based on the notion that the brain is equipped with a hierarchical organization, which embodies environmental contingencies across many time scales, this paper suggests that the medial temporal lobe (MTL)—located deep in the hierarchy—serves as a bridge connecting supra- to infra—MTL levels. Bridging the upper and lower regions of the hierarchy provides a parallel architecture that optimizes information flow between upper and lower regions to aid attention, encoding, and processing of quick complex visual phenomenon. Bypassing intermediate hierarchy levels, information conveyed through the MTL “bridge” allows upper levels to make educated predictions about the prevailing context and accordingly select lower representations to increase the efficiency of predictive coding throughout the hierarchy. This selection or activation/deactivation is associated with endogenous attention. In the event that these “bridge” predictions are inaccurate, this architecture enables the rapid encoding of novel contingencies. A review of hierarchical models in relation to memory is provided along with a new theory, Medial-temporal-lobe Conduit for Parallel Connectivity (MCPC). In this scheme, consolidation is considered as a secondary process, occurring after a MTL-bridged connection, which eventually allows upper and lower levels to access each other directly. With repeated reactivations, as contingencies become consolidated, less MTL activity is predicted. Finally, MTL bridging may aid processing transient but structured perceptual events, by allowing communication between upper and lower levels without calling on intermediate levels of representation.

Autism, oxytocin and interoception

Autism is a pervasive developmental disorder characterized by profound social and verbal communication deficits, stereotypical motor behaviors, restricted interests, and cognitive abnormalities. Autism affects approximately 1% of children in developing countries. Given this prevalence, identifying risk factors and therapeutic interventions are pressing objectives—objectives that rest on neurobiologically grounded and psychologically informed theories about the underlying pathophysiology. In this article, we review the evidence that autism could result from a dysfunctional oxytocin system early in life. As a mediator of successful procreation, not only in the reproductive system, but also in the brain, oxytocin plays a crucial role in sculpting socio-sexual behavior. Formulated within a (Bayesian) predictive coding framework, we propose that oxytocin encodes the saliency or precision of interoceptive signals and enables the neuronal plasticity necessary for acquiring a generative model of the emotional and social 'self.' An aberrant oxytocin system in infancy could therefore help explain the marked deficits in language and social communication—as well as the sensory, autonomic, motor, behavioral, and cognitive abnormalities—seen in autism.

Temporal structure of consciousness and minimal self in schizophrenia

The concept of the minimal self refers to the consciousness of oneself as an immediate subject of experience. According to recent studies, disturbances of the minimal self may be a core feature of schizophrenia. They are emphasized in classical psychiatry literature and in phenomenological work. Impaired minimal self-experience may be defined as a distortion of one’s first-person experiential perspective as, for example, an “altered presence” during which the sense of the experienced self (“mineness”) is subtly affected, or “altered sense of demarcation,” i.e., a difficulty discriminating the self from the non-self. Little is known, however, about the cognitive basis of these disturbances. In fact, recent work indicates that disorders of the self are not correlated with cognitive impairments commonly found in schizophrenia such as working-memory and attention disorders. In addition, a major difficulty with exploring the minimal self experimentally lies in its definition as being non-self-reflexive, and distinct from the verbalized, explicit awareness of an “I.” In this paper, we shall discuss the possibility that disturbances of the minimal self observed in patients with schizophrenia are related to alterations in time processing. We shall review the literature on schizophrenia and time processing that lends support to this possibility. In particular we shall discuss the involvement of temporal integration windows on different time scales (implicit time processing) as well as duration perception disturbances (explicit time processing) in disorders of the minimal self. We argue that a better understanding of the relationship between time and the minimal self as well of issues of embodiment require research that looks more specifically at implicit time processing. Some methodological issues will be discussed.

Virtual reality and consciousness inference in dreaming

This article explores the notion that the brain is genetically endowed with an innate virtual reality generator that - through experience-dependent plasticity - becomes a generative or predictive model of the world. This model, which is most clearly revealed in rapid eye movement (REM) sleep dreaming, may provide the theater for conscious experience. Functional neuroimaging evidence for brain activations that are time-locked to rapid eye movements (REMs) endorses the view that waking consciousness emerges from REM sleep - and dreaming lays the foundations for waking perception. In this view, the brain is equipped with a virtual model of the world that generates predictions of its sensations. This model is continually updated and entrained by sensory prediction errors in wakefulness to ensure veridical perception, but not in dreaming. In contrast, dreaming plays an essential role in maintaining and enhancing the capacity to model the world by minimizing model complexity and thereby maximizing both statistical and thermodynamic efficiency. This perspective suggests that consciousness corresponds to the embodied process of inference, realized through the generation of virtual realities (in both sleep and wakefulness). In short, our premise or hypothesis is that the waking brain engages with the world to predict the causes of sensations, while in sleep the brain's generative model is actively refined so that it generates more efficient predictions during waking. We review the evidence in support of this hypothesis - evidence that grounds consciousness in biophysical computations whose neuronal and neurochemical infrastructure has been disclosed by sleep research.

Perspective: network-guided pattern formation of neural dynamics

The understanding of neural activity patterns is fundamentally linked to an understanding of how the brain's network architecture shapes dynamical processes. Established approaches rely mostly on deviations of a given network from certain classes of random graphs. Hypotheses about the supposed role of prominent topological features (for instance, the roles of modularity, network motifs or hierarchical network organization) are derived from these deviations. An alternative strategy could be to study deviations of network architectures from regular graphs (rings and lattices) and consider the implications of such deviations for self-organized dynamic patterns on the network. Following this strategy, we draw on the theory of spatio-temporal pattern formation and propose a novel perspective for analysing dynamics on networks, by evaluating how the self-organized dynamics are confined by network architecture to a small set of permissible collective states. In particular, we discuss the role of prominent topological features of brain connectivity, such as hubs, modules and hierarchy, in shaping activity patterns. We illustrate the notion of network-guided pattern formation with numerical simulations and outline how it can facilitate the understanding of neural dynamics.

Rhythmic complexity and predictive coding: a novel approach to modeling rhythm and meter perception in music

Musical rhythm, consisting of apparently abstract intervals of accented temporal events, has a remarkable capacity to move our minds and bodies. How does the cognitive system enable our experiences of rhythmically complex music? In this paper, we describe some common forms of rhythmic complexity in music and propose the theory of predictive coding (PC) as a framework for understanding how rhythm and rhythmic complexity are processed in the brain. We also consider why we feel so compelled by rhythmic tension in music. First, we consider theories of rhythm and meter perception, which provide hierarchical and computational approaches to modeling. Second, we present the theory of PC, which posits a hierarchical organization of brain responses reflecting fundamental, survival-related mechanisms associated with predicting future events. According to this theory, perception and learning is manifested through the brain’s Bayesian minimization of the error between the input to the brain and the brain’s prior expectations. Third, we develop a PC model of musical rhythm, in which rhythm perception is conceptualized as an interaction between what is heard (“rhythm”) and the brain’s anticipatory structuring of music (“meter”). Finally, we review empirical studies of the neural and behavioral effects of syncopation, polyrhythm and groove, and propose how these studies can be seen as special cases of the PC theory. We argue that musical rhythm exploits the brain’s general principles of prediction and propose that pleasure and desire for sensorimotor synchronization from musical rhythm may be a result of such mechanisms.

Visual mismatch negativity: a predictive coding view

An increasing number of studies investigate the visual mismatch negativity (vMMN) or use the vMMN as a tool to probe various aspects of human cognition. This paper reviews the theoretical underpinnings of vMMN in the light of methodological considerations and provides recommendations for measuring and interpreting the vMMN. The following key issues are discussed from the experimentalist's point of view in a predictive coding framework: (1) experimental protocols and procedures to control "refractoriness" effects; (2) methods to control attention; (3) vMMN and veridical perception.

Computational model of visual hallucination in dementia with Lewy bodies

Patients with dementia with Lewy bodies (DLB) frequently experience visual hallucination (VH), which has been aptly described as people seeing things that are not there. The distinctive character of VH in DLB necessitates a new theory of visual cognition. We have conducted a series of studies with the aim to understand the mechanism of this dysfunction of the cognitive system. We have proposed that if we view the disease from the internal mechanism of neurocognitive processes, and if also take into consideration recent experimental data on conduction abnormality, at least some of the symptoms can be understood within the framework of network (or disconnection) syndromes. This paper describes the problem from a computational aspect and tries to determine whether conduction disturbances in a computational model can in fact produce a "computational" hallucination under appropriate assumptions.

The functional anatomy of schizophrenia: A dynamic causal modeling study of predictive coding

This paper tests the hypothesis that patients with schizophrenia have a deficit in selectively attending to predictable events. We used dynamic causal modeling (DCM) of electrophysiological responses - to predictable and unpredictable visual targets - to quantify the effective connectivity within and between cortical sources in the visual hierarchy in 25 schizophrenia patients and 25 age-matched controls. We found evidence for marked differences between normal subjects and schizophrenia patients in the strength of extrinsic backward connections from higher hierarchical levels to lower levels within the visual system. In addition, we show that not only do schizophrenia subjects have abnormal connectivity but also that they fail to adjust or optimize this connectivity when events can be predicted. Thus, the differential intrinsic recurrent connectivity observed during processing of predictable versus unpredictable targets was markedly attenuated in schizophrenia patients compared with controls, suggesting a failure to modulate the sensitivity of neurons responsible for passing sensory information of prediction errors up the visual cortical hierarchy. The findings support the proposed role of abnormal connectivity in the neuropathology and pathophysiology of schizophrenia.

Sensory processing and the rubber hand illusion--an evoked potentials study

The rubber hand illusion (RHI) paradigm--in which illusory bodily ownership is induced by synchronous tactile stimulation of a participant's (hidden) hand and a (visible) surrogate--allows one to investigate how the brain resolves conflicting multisensory evidence during perceptual inference. To identify the functional anatomy of the RHI, we used multichannel EEG, acquired under three conditions of tactile stimulation. Evoked potentials were averaged from EEG signals registered to the timing of brushstrokes to the participant's hand. The participant's hand was stroked either in the absence of an artificial hand (REAL) or synchronously with an artificial hand, which either lay in an anatomically plausible (CONGRUENT) or impossible (INCONGRUENT) position. The illusion was reliably elicited in the CONGRUENT condition. For right-hand stimulation, significant differences between conditions emerged at the sensor level around 55 msec after the brushstroke at left frontal and right parietal electrodes. Response amplitudes were smaller for illusory (CONGRUENT) compared with nonillusory (INCONGRUENT and REAL) conditions in the contralateral perirolandic region (pre- and postcentral gyri), superior and inferior parietal lobule, whereas veridical perception of the artificial hand (INCONGRUENT) amplified responses at a scalp region overlying the contralateral postcentral gyrus and inferior parietal lobule compared with the remaining two conditions. Left-hand stimulation produced similar contralateral patterns. These results are consistent with predictive coding models of multisensory integration and may reflect the attenuation of somatosensory precision that is required to resolve perceptual hypotheses about conflicting multisensory input.

Predictions, perception, and a sense of self

In recent years there has been a paradigm shift in theoretical neuroscience in which the brain-as a passive processor of sensory information-is now considered an active organ of inference, generating predictions and hypotheses about the causes of its sensations. In this commentary, we try to convey the basic ideas behind this perspective, describe their neurophysiologic underpinnings, and highlight the potential importance of this formulation for clinical neuroscience. The formalism it provides-and the implementation of active inference in the brain-may have the potential to reveal aspects of functional neuroanatomy that are compromised in conditions ranging from Parkinson disease to schizophrenia. In particular, many neurologic and neuropsychiatric conditions may be understandable in terms of a failure to modulate the postsynaptic gain of neuronal populations reporting prediction errors during action and perception. From the perspective of the predictive brain, this represents a failure to encode the precision of-or confidence in-sensory information. We propose that the predictive or inferential perspective on brain function offers novel insights into brain diseases.

Simultaneous recordings from the primary visual cortex and lateral geniculate nucleus reveal rhythmic interactions and a cortical source for γ-band oscillations

Oscillatory synchronization of neuronal activity has been proposed as a mechanism to modulate effective connectivity between interacting neuronal populations. In the visual system, oscillations in the gamma-frequency range (30-100 Hz) are thought to subserve corticocortical communication. To test whether a similar mechanism might influence subcortical-cortical communication, we recorded local field potential activity from retinotopically aligned regions in the lateral geniculate nucleus (LGN) and primary visual cortex (V1) of alert macaque monkeys viewing stimuli known to produce strong cortical gamma-band oscillations. As predicted, we found robust gamma-band power in V1. In contrast, visual stimulation did not evoke gamma-band activity in the LGN. Interestingly, an analysis of oscillatory phase synchronization of LGN and V1 activity identified synchronization in the alpha (8-14 Hz) and beta (15-30 Hz) frequency bands. Further analysis of directed connectivity revealed that alpha-band interactions mediated corticogeniculate feedback processing, whereas beta-band interactions mediated geniculocortical feedforward processing. These results demonstrate that although the LGN and V1 display functional interactions in the lower frequency bands, gamma-band activity in the alert monkey is largely an emergent property of cortex.

Computational psychiatry: the brain as a phantastic organ

In this Review, we discuss advances in computational neuroscience that relate to psychiatry. We review computational psychiatry in terms of the ambitions of investigators, emerging domains of application, and future work. Our focus is on theoretical formulations of brain function that put subjective beliefs and behaviour within formal (computational) frameworks-frameworks that can be grounded in neurophysiology down to the level of synaptic mechanisms. Understanding the principles that underlie the brain's functional architecture might be essential for an informed phenotyping of psychopathology in terms of its pathophysiological underpinnings. We focus on active (Bayesian) inference and predictive coding. Specifically, we show how basic principles of neuronal computation can be used to explain psychopathology, ranging from impoverished theory of mind in autism to abnormalities of smooth pursuit eye movements in schizophrenia.

Model averaging, optimal inference, and habit formation

Postulating that the brain performs approximate Bayesian inference generates principled and empirically testable models of neuronal function-the subject of much current interest in neuroscience and related disciplines. Current formulations address inference and learning under some assumed and particular model. In reality, organisms are often faced with an additional challenge-that of determining which model or models of their environment are the best for guiding behavior. Bayesian model averaging-which says that an agent should weight the predictions of different models according to their evidence-provides a principled way to solve this problem. Importantly, because model evidence is determined by both the accuracy and complexity of the model, optimal inference requires that these be traded off against one another. This means an agent's behavior should show an equivalent balance. We hypothesize that Bayesian model averaging plays an important role in cognition, given that it is both optimal and realizable within a plausible neuronal architecture. We outline model averaging and how it might be implemented, and then explore a number of implications for brain and behavior. In particular, we propose that model averaging can explain a number of apparently suboptimal phenomena within the framework of approximate (bounded) Bayesian inference, focusing particularly upon the relationship between goal-directed and habitual behavior.

An aberrant precision account of autism

Autism is a neurodevelopmental disorder characterized by problems with social-communication, restricted interests and repetitive behavior. A recent and thought-provoking article presented a normative explanation for the perceptual symptoms of autism in terms of a failure of Bayesian inference (Pellicano and Burr, 2012). In response, we suggested that when Bayesian inference is grounded in its neural instantiation—namely, predictive coding—many features of autistic perception can be attributed to aberrant precision (or beliefs about precision) within the context of hierarchical message passing in the brain (Friston et al., 2013). Here, we unpack the aberrant precision account of autism. Specifically, we consider how empirical findings—that speak directly or indirectly to neurobiological mechanisms—are consistent with the aberrant encoding of precision in autism; in particular, an imbalance of the precision ascribed to sensory evidence relative to prior beliefs.

Contrast gain control and horizontal interactions in V1: a DCM study

Using high-density electrocorticographic recordings – from awake-behaving monkeys – and dynamic causal modelling, we characterised contrast dependent gain control in visual cortex, in terms of synaptic rate constants and intrinsic connectivity. Specifically, we used neural field models to quantify the balance of excitatory and inhibitory influences; both in terms of the strength and spatial dispersion of horizontal intrinsic connections. Our results allow us to infer that increasing contrast increases the sensitivity or gain of superficial pyramidal cells to inputs from spiny stellate populations. Furthermore, changes in the effective spatial extent of horizontal coupling nuance the spatiotemporal filtering properties of cortical laminae in V1 — effectively preserving higher spatial frequencies. These results are consistent with recent non-invasive human studies of contrast dependent changes in the gain of pyramidal cells elaborating forward connections — studies designed to test specific hypotheses about precision and gain control based on predictive coding. Furthermore, they are consistent with established results showing that the receptive fields of V1 units shrink with increasing visual contrast.

•

A new observation model suitable for ECoG recordings.

•

An canonical microcircuit field model.

•

A DCM treatment of multiple experimental conditions and trial-specific effects.

•

Excitation-inhibition balance in terms of strength and dispersion.

Predictive coding, precision and synchrony

Abstract Gotts, Chow and Martin provide a very nice review of repetition priming and suppression that reaches a compelling conclusion-we need to look more closely at synchronization in learning and priming. Indeed, current modeling work focuses on this issue-namely, the dynamic causal modeling of electrophysiological responses to address the role of synchrony in Bayesian explaining away. This commentary revisits the nature and relationships among the four theories in Gotts et al. and nuances some of their empirical predictions. In particular, I emphasize precision or uncertainty in predictive coding as a unifying consideration.

A neurocomputational model of the mismatch negativity

The mismatch negativity (MMN) is an event related potential evoked by violations of regularity. Here, we present a model of the underlying neuronal dynamics based upon the idea that auditory cortex continuously updates a generative model to predict its sensory inputs. The MMN is then modelled as the superposition of the electric fields evoked by neuronal activity reporting prediction errors. The process by which auditory cortex generates predictions and resolves prediction errors was simulated using generalised (Bayesian) filtering--a biologically plausible scheme for probabilistic inference on the hidden states of hierarchical dynamical models. The resulting scheme generates realistic MMN waveforms, explains the qualitative effects of deviant probability and magnitude on the MMN - in terms of latency and amplitude--and makes quantitative predictions about the interactions between deviant probability and magnitude. This work advances a formal understanding of the MMN and--more generally--illustrates the potential for developing computationally informed dynamic causal models of empirical electromagnetic responses.

Active inference, sensory attenuation and illusions

Active inference provides a simple and neurobiologically plausible account of how action and perception are coupled in producing (Bayes) optimal behaviour. This can be seen most easily as minimising prediction error: we can either change our predictions to explain sensory input through perception. Alternatively, we can actively change sensory input to fulfil our predictions. In active inference, this action is mediated by classical reflex arcs that minimise proprioceptive prediction error created by descending proprioceptive predictions. However, this creates a conflict between action and perception; in that, self-generated movements require predictions to override the sensory evidence that one is not actually moving. However, ignoring sensory evidence means that externally generated sensations will not be perceived. Conversely, attending to (proprioceptive and somatosensory) sensations enables the detection of externally generated events but precludes generation of actions. This conflict can be resolved by attenuating the precision of sensory evidence during movement or, equivalently, attending away from the consequences of self-made acts. We propose that this Bayes optimal withdrawal of precise sensory evidence during movement is the cause of psychophysical sensory attenuation. Furthermore, it explains the force-matching illusion and reproduces empirical results almost exactly. Finally, if attenuation is removed, the force-matching illusion disappears and false (delusional) inferences about agency emerge. This is important, given the negative correlation between sensory attenuation and delusional beliefs in normal subjects--and the reduction in the magnitude of the illusion in schizophrenia. Active inference therefore links the neuromodulatory optimisation of precision to sensory attenuation and illusory phenomena during the attribution of agency in normal subjects. It also provides a functional account of deficits in syndromes characterised by false inference and impaired movement--like schizophrenia and Parkinsonism--syndromes that implicate abnormal modulatory neurotransmission.

Expectation and attention in hierarchical auditory prediction

Hierarchical predictive coding suggests that attention in humans emerges from increased precision in probabilistic inference, whereas expectation biases attention in favor of contextually anticipated stimuli. We test these notions within auditory perception by independently manipulating top-down expectation and attentional precision alongside bottom-up stimulus predictability. Our findings support an integrative interpretation of commonly observed electrophysiological signatures of neurodynamics, namely mismatch negativity (MMN), P300, and contingent negative variation (CNV), as manifestations along successive levels of predictive complexity. Early first-level processing indexed by the MMN was sensitive to stimulus predictability: here, attentional precision enhanced early responses, but explicit top-down expectation diminished it. This pattern was in contrast to later, second-level processing indexed by the P300: although sensitive to the degree of predictability, responses at this level were contingent on attentional engagement and in fact sharpened by top-down expectation. At the highest level, the drift of the CNV was a fine-grained marker of top-down expectation itself. Source reconstruction of high-density EEG, supported by intracranial recordings, implicated temporal and frontal regions differentially active at early and late levels. The cortical generators of the CNV suggested that it might be involved in facilitating the consolidation of context-salient stimuli into conscious perception. These results provide convergent empirical support to promising recent accounts of attention and expectation in predictive coding.

A theoretical and experimental study of neuromorphic atomic switch networks for reservoir computing

Atomic switch networks (ASNs) have been shown to generate network level dynamics that resemble those observed in biological neural networks. To facilitate understanding and control of these behaviors, we developed a numerical model based on the synapse-like properties of individual atomic switches and the random nature of the network wiring. We validated the model against various experimental results highlighting the possibility to functionalize the network plasticity and the differences between an atomic switch in isolation and its behaviors in a network. The effects of changing connectivity density on the nonlinear dynamics were examined as characterized by higher harmonic generation in response to AC inputs. To demonstrate their utility for computation, we subjected the simulated network to training within the framework of reservoir computing and showed initial evidence of the ASN acting as a reservoir which may be optimized for specific tasks by adjusting the input gain. The work presented represents steps in a unified approach to experimentation and theory of complex systems to make ASNs a uniquely scalable platform for neuromorphic computing.

Circular inferences in schizophrenia

A considerable number of recent experimental and computational studies suggest that subtle impairments of excitatory to inhibitory balance or regulation are involved in many neurological and psychiatric conditions. The current paper aims to relate, specifically and quantitatively, excitatory to inhibitory imbalance with psychotic symptoms in schizophrenia. Considering that the brain constructs hierarchical causal models of the external world, we show that the failure to maintain the excitatory to inhibitory balance results in hallucinations as well as in the formation and subsequent consolidation of delusional beliefs. Indeed, the consequence of excitatory to inhibitory imbalance in a hierarchical neural network is equated to a pathological form of causal inference called 'circular belief propagation'. In circular belief propagation, bottom-up sensory information and top-down predictions are reverberated, i.e. prior beliefs are misinterpreted as sensory observations and vice versa. As a result, these predictions are counted multiple times. Circular inference explains the emergence of erroneous percepts, the patient's overconfidence when facing probabilistic choices, the learning of 'unshakable' causal relationships between unrelated events and a paradoxical immunity to perceptual illusions, which are all known to be associated with schizophrenia.

From birdsong to human speech recognition: bayesian inference on a hierarchy of nonlinear dynamical systems

Our knowledge about the computational mechanisms underlying human learning and recognition of sound sequences, especially speech, is still very limited. One difficulty in deciphering the exact means by which humans recognize speech is that there are scarce experimental findings at a neuronal, microscopic level. Here, we show that our neuronal-computational understanding of speech learning and recognition may be vastly improved by looking at an animal model, i.e., the songbird, which faces the same challenge as humans: to learn and decode complex auditory input, in an online fashion. Motivated by striking similarities between the human and songbird neural recognition systems at the macroscopic level, we assumed that the human brain uses the same computational principles at a microscopic level and translated a birdsong model into a novel human sound learning and recognition model with an emphasis on speech. We show that the resulting Bayesian model with a hierarchy of nonlinear dynamical systems can learn speech samples such as words rapidly and recognize them robustly, even in adverse conditions. In addition, we show that recognition can be performed even when words are spoken by different speakers and with different accents—an everyday situation in which current state-of-the-art speech recognition models often fail. The model can also be used to qualitatively explain behavioral data on human speech learning and derive predictions for future experiments.

Neuroscience still lacks a concrete explanation of how humans recognize speech. Even though neuroimaging techniques are helpful in determining the brain areas involved in speech recognition, there are rarely mechanistic explanations at a neuronal level. Here, we assume that songbirds and humans solve a very similar task: extracting information from sound wave modulations produced by a singing bird or a speaking human. Given strong evidence that both humans and songbirds, although genetically very distant, converged to a similar solution, we combined the vast amount of neurobiological findings for songbirds with nonlinear dynamical systems theory to develop a hierarchical, Bayesian model which explains fundamental functions in recognition of sound sequences. We found that the resulting model is good at learning and recognizing human speech. We suggest that this translated model can be used to qualitatively explain or predict experimental data, and the underlying mechanism can be used to construct improved automatic speech recognition algorithms.

Minimal self-models and the free energy principle

The term “minimal phenomenal selfhood” (MPS) describes the basic, pre-reflective experience of being a self (Blanke and Metzinger, 2009). Theoretical accounts of the minimal self have long recognized the importance and the ambivalence of the body as both part of the physical world, and the enabling condition for being in this world (Gallagher, 2005a; Grafton, 2009). A recent account of MPS (Metzinger, 2004a) centers on the consideration that minimal selfhood emerges as the result of basic self-modeling mechanisms, thereby being founded on pre-reflective bodily processes. The free energy principle (FEP; Friston, 2010) is a novel unified theory of cortical function built upon the imperative that self-organizing systems entail hierarchical generative models of the causes of their sensory input, which are optimized by minimizing free energy as an approximation of the log-likelihood of the model. The implementation of the FEP via predictive coding mechanisms and in particular the active inference principle emphasizes the role of embodiment for predictive self-modeling, which has been appreciated in recent publications. In this review, we provide an overview of these conceptions and illustrate thereby the potential power of the FEP in explaining the mechanisms underlying minimal selfhood and its key constituents, multisensory integration, interoception, agency, perspective, and the experience of mineness. We conclude that the conceptualization of MPS can be well mapped onto a hierarchical generative model furnished by the FEP and may constitute the basis for higher-level, cognitive forms of self-referral, as well as the understanding of other minds.

Free energy, precision and learning: the role of cholinergic neuromodulation

Acetylcholine (ACh) is a neuromodulatory transmitter implicated in perception and learning under uncertainty. This study combined computational simulations and pharmaco-electroencephalography in humans, to test a formulation of perceptual inference based upon the free energy principle. This formulation suggests that ACh enhances the precision of bottom-up synaptic transmission in cortical hierarchies by optimizing the gain of supragranular pyramidal cells. Simulations of a mismatch negativity paradigm predicted a rapid trial-by-trial suppression of evoked sensory prediction error (PE) responses that is attenuated by cholinergic neuromodulation. We confirmed this prediction empirically with a placebo-controlled study of cholinesterase inhibition. Furthermore, using dynamic causal modeling, we found that drug-induced differences in PE responses could be explained by gain modulation in supragranular pyramidal cells in primary sensory cortex. This suggests that ACh adaptively enhances sensory precision by boosting bottom-up signaling when stimuli are predictable, enabling the brain to respond optimally under different levels of environmental uncertainty.

Information and efficiency in the nervous system--a synthesis

In systems biology, questions concerning the molecular and cellular makeup of an organism are of utmost importance, especially when trying to understand how unreliable components—like genetic circuits, biochemical cascades, and ion channels, among others—enable reliable and adaptive behaviour. The repertoire and speed of biological computations are limited by thermodynamic or metabolic constraints: an example can be found in neurons, where fluctuations in biophysical states limit the information they can encode—with almost 20–60% of the total energy allocated for the brain used for signalling purposes, either via action potentials or by synaptic transmission. Here, we consider the imperatives for neurons to optimise computational and metabolic efficiency, wherein benefits and costs trade-off against each other in the context of self-organised and adaptive behaviour. In particular, we try to link information theoretic (variational) and thermodynamic (Helmholtz) free-energy formulations of neuronal processing and show how they are related in a fundamental way through a complexity minimisation lemma.

Toward a second-person neuroscience

In spite of the remarkable progress made in the burgeoning field of social neuroscience, the neural mechanisms that underlie social encounters are only beginning to be studied and could – paradoxically – be seen as representing the “dark matter” of social neuroscience. Recent conceptual and empirical developments consistently indicate the need for investigations that allow the study of real-time social encounters in a truly interactive manner. This suggestion is based on the premise that social cognition is fundamentally different when we are in interaction with others rather than merely observing them. In this article, we outline the theoretical conception of a second-person approach to other minds and review evidence from neuroimaging, psychophysiological studies, and related fields to argue for the development of a second-person neuroscience, which will help neuroscience to really “go social”; this may also be relevant for our understanding of psychiatric disorders construed as disorders of social cognition.

Emotional valence and the free-energy principle

The free-energy principle has recently been proposed as a unified Bayesian account of perception, learning and action. Despite the inextricable link between emotion and cognition, emotion has not yet been formulated under this framework. A core concept that permeates many perspectives on emotion is valence, which broadly refers to the positive and negative character of emotion or some of its aspects. In the present paper, we propose a definition of emotional valence in terms of the negative rate of change of free-energy over time. If the second time-derivative of free-energy is taken into account, the dynamics of basic forms of emotion such as happiness, unhappiness, hope, fear, disappointment and relief can be explained. In this formulation, an important function of emotional valence turns out to regulate the learning rate of the causes of sensory inputs. When sensations increasingly violate the agent's expectations, valence is negative and increases the learning rate. Conversely, when sensations increasingly fulfil the agent's expectations, valence is positive and decreases the learning rate. This dynamic interaction between emotional valence and learning rate highlights the crucial role played by emotions in biological agents' adaptation to unexpected changes in their world.

On disturbed time continuity in schizophrenia: an elementary impairment in visual perception?

Schizophrenia is associated with a series of visual perception impairments, which might impact on the patients' every day life and be related to clinical symptoms. However, the heterogeneity of the visual disorders make it a challenge to understand both the mechanisms and the consequences of these impairments, i.e., the way patients experience the outer world. Based on earlier psychiatry literature, we argue that issues regarding time might shed a new light on the disorders observed in patients with schizophrenia. We will briefly review the mechanisms involved in the sense of time continuity and clinical evidence that they are impaired in patients with schizophrenia. We will then summarize a recent experimental approach regarding the coding of time-event structure in time, namely the ability to discriminate between simultaneous and asynchronous events. The use of an original method of analysis allowed us to distinguish between explicit and implicit judgments of synchrony. We showed that for SOAs below 20 ms neither patients nor controls fuse events in time. On the contrary subjects distinguish events at an implicit level even when judging them as synchronous. In addition, the implicit responses of patients and controls differ qualitatively. It is as if controls always put more weight on the last occurred event, whereas patients have a difficulty to follow events in time at an implicit level. In patients, there is a clear dissociation between results at short and large asynchronies, that suggest selective mechanisms for the implicit coding of time-event structure. These results might explain the disruption of the sense of time continuity in patients. We argue that this line of research might also help us to better understand the mechanisms of the visual impairments in patients and how they see their environment.

Grounding predictive coding models in empirical neuroscience research

Clark makes a convincing case for the merits of conceptualizing brains as hierarchical prediction machines. This perspective has the potential to provide an elegant and powerful general theory of brain function, but it will ultimately stand or fall with evidence from basic neuroscience research. Here, we characterize the status quo of that evidence and highlight important avenues for future investigations.

Defining statistical perceptions with an empirical Bayesian approach

Extracting statistical structures (including textures or contrasts) from a natural stimulus is a central challenge in both biological and engineering contexts. This study interprets the process of statistical recognition in terms of hyperparameter estimations and free-energy minimization procedures with an empirical Bayesian approach. This mathematical interpretation resulted in a framework for relating physiological insights in animal sensory systems to the functional properties of recognizing stimulus statistics. We applied the present theoretical framework to two typical models of natural images that are encoded by a population of simulated retinal neurons, and demonstrated that the resulting cognitive performances could be quantified with the Fisher information measure. The current enterprise yielded predictions about the properties of human texture perception, suggesting that the perceptual resolution of image statistics depends on visual field angles, internal noise, and neuronal information processing pathways, such as the magnocellular, parvocellular, and koniocellular systems. Furthermore, the two conceptually similar natural-image models were found to yield qualitatively different predictions, striking a note of warning against confusing the two models when describing a natural image.

The free-energy self: a predictive coding account of self-recognition

Recognising and representing one's self as distinct from others is a fundamental component of self-awareness. However, current theories of self-recognition are not embedded within global theories of cortical function and therefore fail to provide a compelling explanation of how the self is processed. We present a theoretical account of the neural and computational basis of self-recognition that is embedded within the free-energy account of cortical function. In this account one's body is processed in a Bayesian manner as the most likely to be "me". Such probabilistic representation arises through the integration of information from hierarchically organised unimodal systems in higher-level multimodal areas. This information takes the form of bottom-up "surprise" signals from unimodal sensory systems that are explained away by top-down processes that minimise the level of surprise across the brain. We present evidence that this theoretical perspective may account for the findings of psychological and neuroimaging investigations into self-recognition and particularly evidence that representations of the self are malleable, rather than fixed as previous accounts of self-recognition might suggest.

The computational anatomy of psychosis

This paper considers psychotic symptoms in terms of false inferences or beliefs. It is based on the notion that the brain is an inference machine that actively constructs hypotheses to explain or predict its sensations. This perspective provides a normative (Bayes-optimal) account of action and perception that emphasizes probabilistic representations; in particular, the confidence or precision of beliefs about the world. We will consider hallucinosis, abnormal eye movements, sensory attenuation deficits, catatonia, and delusions as various expressions of the same core pathology: namely, an aberrant encoding of precision. From a cognitive perspective, this represents a pernicious failure of metacognition (beliefs about beliefs) that can confound perceptual inference. In the embodied setting of active (Bayesian) inference, it can lead to behaviors that are paradoxically more accurate than Bayes-optimal behavior. Crucially, this normative account is accompanied by a neuronally plausible process theory based upon hierarchical predictive coding. In predictive coding, precision is thought to be encoded by the post-synaptic gain of neurons reporting prediction error. This suggests that both pervasive trait abnormalities and florid failures of inference in the psychotic state can be linked to factors controlling post-synaptic gain - such as NMDA receptor function and (dopaminergic) neuromodulation. We illustrate these points using biologically plausible simulations of perceptual synthesis, smooth pursuit eye movements and attribution of agency - that all use the same predictive coding scheme and pathology: namely, a reduction in the precision of prior beliefs, relative to sensory evidence.

Not so primitive: context-sensitive meta-learning about unattended sound sequences

Mismatch negativity (MMN), an evoked response potential elicited when a “deviant” sound violates a regularity in the auditory environment, is integral to auditory scene processing and has been used to demonstrate “primitive intelligence” in auditory short-term memory. Using a new multiple-context and -timescale protocol we show that MMN magnitude displays a context-sensitive modulation depending on changes in the probability of a deviant at multiple temporal scales. We demonstrate a primacy bias causing asymmetric evidence-based modulation of predictions about the environment, and we demonstrate that learning how to learn about deviant probability (meta-learning) induces context-sensitive variation in the accessibility of predictive long-term memory representations that underpin the MMN. The existence of the bias and meta-learning are consistent with automatic attributions of behavioral salience governing relevance-filtering processes operating outside of awareness.

Task-dependent changes in cross-level coupling between single neurons and oscillatory activity in multiscale networks

Understanding the principles governing the dynamic coordination of functional brain networks remains an important unmet goal within neuroscience. How do distributed ensembles of neurons transiently coordinate their activity across a variety of spatial and temporal scales? While a complete mechanistic account of this process remains elusive, evidence suggests that neuronal oscillations may play a key role in this process, with different rhythms influencing both local computation and long-range communication. To investigate this question, we recorded multiple single unit and local field potential (LFP) activity from microelectrode arrays implanted bilaterally in macaque motor areas. Monkeys performed a delayed center-out reach task either manually using their natural arm (Manual Control, MC) or under direct neural control through a brain-machine interface (Brain Control, BC). In accord with prior work, we found that the spiking activity of individual neurons is coupled to multiple aspects of the ongoing motor beta rhythm (10–45 Hz) during both MC and BC, with neurons exhibiting a diversity of coupling preferences. However, here we show that for identified single neurons, this beta-to-rate mapping can change in a reversible and task-dependent way. For example, as beta power increases, a given neuron may increase spiking during MC but decrease spiking during BC, or exhibit a reversible shift in the preferred phase of firing. The within-task stability of coupling, combined with the reversible cross-task changes in coupling, suggest that task-dependent changes in the beta-to-rate mapping play a role in the transient functional reorganization of neural ensembles. We characterize the range of task-dependent changes in the mapping from beta amplitude, phase, and inter-hemispheric phase differences to the spike rates of an ensemble of simultaneously-recorded neurons, and discuss the potential implications that dynamic remapping from oscillatory activity to spike rate and timing may hold for models of computation and communication in distributed functional brain networks.

How is the functional role of a particular neuron established within an ensemble? The concept of a neural tuning curve – the mapping from input variables such as movement direction to output firing rate – has proven useful in investigating neural function. However, prior work shows that tuning curves are not fixed but may be remapped as a function of task demands – presumably via high-level mechanisms of cognitive control. How is this accomplished? Brain rhythms may play a causal role in this process, but the coupling of single cells to network activity remains poorly understood. We investigated the coupling between rhythmic beta activity and spiking as macaques performed two different tasks. This coupling can be described in terms of a function that maps oscillatory amplitude and phase to instantaneous spike rate. Similarly to direction tuning, this “internal” tuning curve also exhibits task-dependent changes. We characterize these changes across a large ensemble of simultaneously-recorded cells, and consider some of the neuro-computational implications presented by cross-level coupling between single cells and large-scale networks. In particular, relative to the slow time-scale of behavior, the observed beta-to-rate mappings may prove useful for modulating winner-take-all dynamics on intermediate time-scales and relative spike timing on fast time-scales.

Canonical microcircuits for predictive coding

This Perspective considers the influential notion of a canonical (cortical) microcircuit in light of recent theories about neuronal processing. Specifically, we conciliate quantitative studies of microcircuitry and the functional logic of neuronal computations. We revisit the established idea that message passing among hierarchical cortical areas implements a form of Bayesian inference-paying careful attention to the implications for intrinsic connections among neuronal populations. By deriving canonical forms for these computations, one can associate specific neuronal populations with specific computational roles. This analysis discloses a remarkable correspondence between the microcircuitry of the cortical column and the connectivity implied by predictive coding. Furthermore, it provides some intuitive insights into the functional asymmetries between feedforward and feedback connections and the characteristic frequencies over which they operate.