Neural processing as causal inference

Perception is about making sense, that is, understanding what events in the outside world caused the sensory observations. Consistent with this intuition, many aspects of human behavior confronting noise and ambiguity are well explained by principles of causal inference. Extending these insights, recent studies have applied the same powerful set of tools to perceptual processing at the neural level. According to these approaches, microscopic neural structures solve elementary probabilistic tasks and can be combined to construct hierarchical predictive models of the sensory input. This framework suggests that variability in neural responses reflects the inherent uncertainty associated with sensory interpretations and that sensory neurons are active predictors rather than passive filters of their inputs. Causal inference can account parsimoniously and quantitatively for non-linear dynamical properties in single synapses, single neurons and sensory receptive fields.     

Functional brain networks underlying perceptual switching: auditory streaming and verbal transformations

Recent studies have shown that auditory scene analysis involves distributed neural sites below, in, and beyond the auditory cortex (AC). However, it remains unclear what role each site plays and how they interact in the formation and selection of auditory percepts. We addressed this issue through perceptual multistability phenomena, namely, spontaneous perceptual switching in auditory streaming (AS) for a sequence of repeated triplet tones, and perceptual changes for a repeated word, known as verbal transformations (VTs). An event-related fMRI analysis revealed brain activity timelocked to perceptual switching in the cerebellum for AS, in frontal areas for VT, and the AC and thalamus for both. The results suggest that motor-based prediction, produced by neural networks outside the auditory system, plays essential roles in the segmentation of acoustic sequences both in AS and VT. The frequency of perceptual switching was determined by a balance between the activation of two sites, which are proposed to be involved in exploring novel perceptual organization and stabilizing current perceptual organization. The effect of the gene polymorphism of catechol-O-methyltransferase (COMT) on individual variations in switching frequency suggests that the balance of exploration and stabilization is modulated by catecholamines such as dopamine and noradrenalin. These mechanisms would support the noteworthy flexibility of auditory scene analysis.     

Multisensory Causal Inference in the Brain

At any given moment, our brain processes multiple inputs from its different sensory modalities (vision, hearing, touch, etc.). In deciphering this array of sensory information, the brain has to solve two problems: (1) which of the inputs originate from the same object and should be integrated and (2) for the sensations originating from the same object, how best to integrate them. Recent behavioural studies suggest that the human brain solves these problems using optimal probabilistic inference, known as Bayesian causal inference. However, how and where the underlying computations are carried out in the brain have remained unknown. By combining neuroimaging-based decoding techniques and computational modelling of behavioural data, a new study now sheds light on how multisensory causal inference maps onto specific brain areas. The results suggest that the complexity of neural computations increases along the visual hierarchy and link specific components of the causal inference process with specific visual and parietal regions.     

Where’s the Noise? Key Features of Spontaneous Activity and Neural Variability Arise through Learning in a Deterministic Network

Even in the absence of sensory stimulation the brain is spontaneously active. This background “noise” seems to be the dominant cause of the notoriously high trial-to-trial variability of neural recordings. Recent experimental observations have extended our knowledge of trial-to-trial variability and spontaneous activity in several directions: 1. Trial-to-trial variability systematically decreases following the onset of a sensory stimulus or the start of a motor act. 2. Spontaneous activity states in sensory cortex outline the region of evoked sensory responses. 3. Across development, spontaneous activity aligns itself with typical evoked activity patterns. 4. The spontaneous brain activity prior to the presentation of an ambiguous stimulus predicts how the stimulus will be interpreted. At present it is unclear how these observations relate to each other and how they arise in cortical circuits. Here we demonstrate that all of these phenomena can be accounted for by a deterministic self-organizing recurrent neural network model (SORN), which learns a predictive model of its sensory environment. The SORN comprises recurrently coupled populations of excitatory and inhibitory threshold units and learns via a combination of spike-timing dependent plasticity (STDP) and homeostatic plasticity mechanisms. Similar to balanced network architectures, units in the network show irregular activity and variable responses to inputs. Additionally, however, the SORN exhibits sequence learning abilities matching recent findings from visual cortex and the network’s spontaneous activity reproduces the experimental findings mentioned above. Intriguingly, the network’s behaviour is reminiscent of sampling-based probabilistic inference, suggesting that correlates of sampling-based inference can develop from the interaction of STDP and homeostasis in deterministic networks. We conclude that key observations on spontaneous brain activity and the variability of neural responses can be accounted for by a simple deterministic recurrent neural network which learns a predictive model of its sensory environment via a combination of generic neural plasticity mechanisms.     

Task-induced neural covariability as a signature of approximate Bayesian learning and inference

Perception is often characterized computationally as an inference process in which uncertain or ambiguous sensory inputs are combined with prior expectations. Although behavioral studies have shown that observers can change their prior expectations in the context of a task, robust neural signatures of task-specific priors have been elusive. Here, we analytically derive such signatures under the general assumption that the responses of sensory neurons encode posterior beliefs that combine sensory inputs with task-specific expectations. Specifically, we derive predictions for the task-dependence of correlated neural variability and decision-related signals in sensory neurons. The qualitative aspects of our results are parameter-free and specific to the statistics of each task. The predictions for correlated variability also differ from predictions of classic feedforward models of sensory processing and are therefore a strong test of theories of hierarchical Bayesian inference in the brain. Importantly, we find that Bayesian learning predicts an increase in so-called “differential correlations” as the observer’s internal model learns the stimulus distribution, and the observer’s behavioral performance improves. This stands in contrast to classic feedforward encoding/decoding models of sensory processing, since such correlations are fundamentally information-limiting. We find support for our predictions in data from existing neurophysiological studies across a variety of tasks and brain areas. Finally, we show in simulation how measurements of sensory neural responses can reveal information about a subject’s internal beliefs about the task. Taken together, our results reinterpret task-dependent sources of neural covariability as signatures of Bayesian inference and provide new insights into their cause and their function.     

Abstract coding of audiovisual speech: beyond sensory representation

Is there a neural representation of speech that transcends its sensory properties? Using fMRI, we investigated whether there are brain areas where neural activity during observation of sublexical audiovisual input corresponds to a listener's speech percept (what is "heard") independent of the sensory properties of the input. A target audiovisual stimulus was preceded by stimuli that (1) shared the target's auditory features (auditory overlap), (2) shared the target's visual features (visual overlap), or (3) shared neither the target's auditory or visual features but were perceived as the target (perceptual overlap). In two left-hemisphere regions (pars opercularis, planum polare), the target invoked less activity when it was preceded by the perceptually overlapping stimulus than when preceded by stimuli that shared one of its sensory components. This pattern of neural facilitation indicates that these regions code sublexical speech at an abstract level corresponding to that of the speech percept.     

Predictive Coding or Evidence Accumulation? False Inference and Neuronal Fluctuations

Perceptual decisions can be made when sensory input affords an inference about what generated that input. Here, we report findings from two independent perceptual experiments conducted during functional magnetic resonance imaging (fMRI) with a sparse event-related design. The first experiment, in the visual modality, involved forced-choice discrimination of coherence in random dot kinematograms that contained either subliminal or periliminal motion coherence. The second experiment, in the auditory domain, involved free response detection of (non-semantic) near-threshold acoustic stimuli. We analysed fluctuations in ongoing neural activity, as indexed by fMRI, and found that neuronal activity in sensory areas (extrastriate visual and early auditory cortex) biases perceptual decisions towards correct inference and not towards a specific percept. Hits (detection of near-threshold stimuli) were preceded by significantly higher activity than both misses of identical stimuli or false alarms, in which percepts arise in the absence of appropriate sensory input. In accord with predictive coding models and the free-energy principle, this observation suggests that cortical activity in sensory brain areas reflects the precision of prediction errors and not just the sensory evidence or prediction errors per se.     

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.     

Decoding Accuracy in Supplementary Motor Cortex Correlates with Perceptual Sensitivity to Tactile Roughness

Perceptual sensitivity to tactile roughness varies across individuals for the same degree of roughness. A number of neurophysiological studies have investigated the neural substrates of tactile roughness perception, but the neural processing underlying the strong individual differences in perceptual roughness sensitivity remains unknown. In this study, we explored the human brain activation patterns associated with the behavioral discriminability of surface texture roughness using functional magnetic resonance imaging (fMRI). First, a whole-brain searchlight multi-voxel pattern analysis (MVPA) was used to find brain regions from which we could decode roughness information. The searchlight MVPA revealed four brain regions showing significant decoding results: the supplementary motor area (SMA), contralateral postcentral gyrus (S1), and superior portion of the bilateral temporal pole (STP). Next, we evaluated the behavioral roughness discrimination sensitivity of each individual using the just-noticeable difference (JND) and correlated this with the decoding accuracy in each of the four regions. We found that only the SMA showed a significant correlation between neuronal decoding accuracy and JND across individuals; Participants with a smaller JND (i.e., better discrimination ability) exhibited higher decoding accuracy from their voxel response patterns in the SMA. Our findings suggest that multivariate voxel response patterns presented in the SMA represent individual perceptual sensitivity to tactile roughness and people with greater perceptual sensitivity to tactile roughness are likely to have more distinct neural representations of different roughness levels in their SMA.     

Subjective visual perception: from local processing to emergent phenomena of brain activity

The combination of electrophysiological recordings with ambiguous visual stimulation made possible the detection of neurons that represent the content of subjective visual perception and perceptual suppression in multiple cortical and subcortical brain regions. These neuronal populations, commonly referred to as the neural correlates of consciousness, are more likely to be found in the temporal and prefrontal cortices as well as the pulvinar, indicating that the content of perceptual awareness is represented with higher fidelity in higher-order association areas of the cortical and thalamic hierarchy, reflecting the outcome of competitive interactions between conflicting sensory information resolved in earlier stages. However, despite the significant insights into conscious perception gained through monitoring the activities of single neurons and small, local populations, the immense functional complexity of the brain arising from correlations in the activity of its constituent parts suggests that local, microscopic activity could only partially reveal the mechanisms involved in perceptual awareness. Rather, the dynamics of functional connectivity patterns on a mesoscopic and macroscopic level could be critical for conscious perception. Understanding these emergent spatio-temporal patterns could be informative not only for the stability of subjective perception but also for spontaneous perceptual transitions suggested to depend either on the dynamics of antagonistic ensembles or on global intrinsic activity fluctuations that may act upon explicit neural representations of sensory stimuli and induce perceptual reorganization. Here, we review the most recent results from local activity recordings and discuss the potential role of effective, correlated interactions during perceptual awareness.     

Tactile rivalry demonstrated with an ambiguous apparent-motion quartet

When observers view ambiguous visual stimuli, their perception will often alternate between the possible interpretations, a phenomenon termed perceptual rivalry [1]. To induce perceptual rivalry in the tactile domain, we developed a new tactile illusion, based on the visual apparent-motion quartet [2]. Pairs of 200 ms vibrotactile stimuli were applied to the finger pad at intervals separated by 300 ms. The location of each successive stimulus pair alternated between the opposing diagonal corners of the approximately 1 cm(2) stimulation array. This stimulation sequence led all participants to report switches between the perception of motion traveling either up and down or left and right across their fingertip. Adaptation to tactile stimulation biased toward one direction caused subsequent ambiguous stimulation to be experienced in the opposing direction. In contrast, when consecutive trials of ambiguous stimulation were presented, motion was generally perceived in the direction consistent with the motion reported in the previous trial. Voluntary eye movements induced shifts in the tactile perception toward a motion axis aligned along a world-centered coordinate frame. Because the tactile quartet results in switching perceptual states despite unvaried sensory input, it is ideally suited to future studies of the neural processes associated with conscious tactile perception.     

Bistable Perception Modeled as Competing Stochastic Integrations at Two Levels

We propose a novel explanation for bistable perception, namely, the collective dynamics of multiple neural populations that are individually meta-stable. Distributed representations of sensory input and of perceptual state build gradually through noise-driven transitions in these populations, until the competition between alternative representations is resolved by a threshold mechanism. The perpetual repetition of this collective race to threshold renders perception bistable. This collective dynamics – which is largely uncoupled from the time-scales that govern individual populations or neurons – explains many hitherto puzzling observations about bistable perception: the wide range of mean alternation rates exhibited by bistable phenomena, the consistent variability of successive dominance periods, and the stabilizing effect of past perceptual states. It also predicts a number of previously unsuspected relationships between observable quantities characterizing bistable perception. We conclude that bistable perception reflects the collective nature of neural decision making rather than properties of individual populations or neurons.     

Bifurcation study of a neural field competition model with an application to perceptual switching in motion integration

Perceptual multistability is a phenomenon in which alternate interpretations of a fixed stimulus are perceived intermittently. Although correlates between activity in specific cortical areas and perception have been found, the complex patterns of activity and the underlying mechanisms that gate multistable perception are little understood. Here, we present a neural field competition model in which competing states are represented in a continuous feature space. Bifurcation analysis is used to describe the different types of complex spatio-temporal dynamics produced by the model in terms of several parameters and for different inputs. The dynamics of the model was then compared to human perception investigated psychophysically during long presentations of an ambiguous, multistable motion pattern known as the barberpole illusion. In order to do this, the model is operated in a parameter range where known physiological response properties are reproduced whilst also working close to bifurcation. The model accounts for characteristic behaviour from the psychophysical experiments in terms of the type of switching observed and changes in the rate of switching with respect to contrast. In this way, the modelling study sheds light on the underlying mechanisms that drive perceptual switching in different contrast regimes. The general approach presented is applicable to a broad range of perceptual competition problems in which spatial interactions play a role.     

Evolution of heterogeneous perceptual limits and indifference in competitive foraging

The collective behaviour of animal and human groups emerges from the individual decisions and actions of their constituent members. Recent research has revealed many ways in which the behaviour of groups can be influenced by differences amongst their constituent individuals. The existence of individual differences that have implications for collective behaviour raises important questions. How are these differences generated and maintained? Are individual differences driven by exogenous factors, or are they a response to the social dilemmas these groups face? Here I consider the classic case of patch selection by foraging agents under conditions of social competition. I introduce a multilevel model wherein the perceptual sensitivities of agents evolve in response to their foraging success or failure over repeated patch selections. This model reveals a bifurcation in the population, creating a class of agents with no perceptual sensitivity. These agents exploit the social environment to avoid the costs of accurate perception, relying on other agents to make fitness rewards insensitive to the choice of foraging patch. This provides a individual-based evolutionary basis for models incorporating perceptual limits that have been proposed to explain observed deviations from the Ideal Free Distribution (IFD) in empirical studies, while showing that the common assumption in such models that agents share identical sensory limits is likely false. Further analysis of the model shows how agents develop perceptual strategic niches in response to environmental variability. The emergence of agents insensitive to reward differences also has implications for societal resource allocation problems, including the use of financial and prediction markets as mechanisms for aggregating collective wisdom.     

Neural control of three-dimensional eye and head movements

Although the eyes and head can potentially rotate about any three-dimensional axis during orienting gaze shifts, behavioral recordings have shown that certain lawful strategies--such as Listing's law and Donders' law--determine which axis is used for a particular sensory input. Here, we review recent advances in understanding the neuromuscular mechanisms for these laws, the neural mechanisms that control three-dimensional head posture, and the neural mechanisms that coordinate three-dimensional eye orientation with head motion. Finally, we consider how the brain copes with the perceptual consequences of these motor acts.     

Rhythms of Consciousness: Binocular Rivalry Reveals Large-Scale Oscillatory Network Dynamics Mediating Visual Perception

Consciousness has been proposed to emerge from functionally integrated large-scale ensembles of gamma-synchronous neural populations that form and dissolve at a frequency in the theta band. We propose that discrete moments of perceptual experience are implemented by transient gamma-band synchronization of relevant cortical regions, and that disintegration and reintegration of these assemblies is time-locked to ongoing theta oscillations. In support of this hypothesis we provide evidence that (1) perceptual switching during binocular rivalry is time-locked to gamma-band synchronizations which recur at a theta rate, indicating that the onset of new conscious percepts coincides with the emergence of a new gamma-synchronous assembly that is locked to an ongoing theta rhythm; (2) localization of the generators of these gamma rhythms reveals recurrent prefrontal and parietal sources; (3) theta modulation of gamma-band synchronization is observed between and within the activated brain regions. These results suggest that ongoing theta-modulated-gamma mechanisms periodically reintegrate a large-scale prefrontal-parietal network critical for perceptual experience. Moreover, activation and network inclusion of inferior temporal cortex and motor cortex uniquely occurs on the cycle immediately preceding responses signaling perceptual switching. This suggests that the essential prefrontal-parietal oscillatory network is expanded to include additional cortical regions relevant to tasks and perceptions furnishing consciousness at that moment, in this case image processing and response initiation, and that these activations occur within a time frame consistent with the notion that conscious processes directly affect behaviour.     

End-to-end neural system identification with neural information flow

Neural information flow (NIF) provides a novel approach for system identification in neuroscience. It models the neural computations in multiple brain regions and can be trained end-to-end via stochastic gradient descent from noninvasive data. NIF models represent neural information processing via a network of coupled tensors, each encoding the representation of the sensory input contained in a brain region. The elements of these tensors can be interpreted as cortical columns whose activity encodes the presence of a specific feature in a spatiotemporal location. Each tensor is coupled to the measured data specific to a brain region via low-rank observation models that can be decomposed into the spatial, temporal and feature receptive fields of a localized neuronal population. Both these observation models and the convolutional weights defining the information processing within regions are learned end-to-end by predicting the neural signal during sensory stimulation. We trained a NIF model on the activity of early visual areas using a large-scale fMRI dataset recorded in a single participant. We show that we can recover plausible visual representations and population receptive fields that are consistent with empirical findings.     

United we sense, divided we fail: context-driven perception of ambiguous visual stimuli

Ambiguous visual stimuli provide the brain with sensory information that contains conflicting evidence for multiple mutually exclusive interpretations. Two distinct aspects of the phenomenological experience associated with viewing ambiguous visual stimuli are the apparent stability of perception whenever one perceptual interpretation is dominant, and the instability of perception that causes perceptual dominance to alternate between perceptual interpretations upon extended viewing. This review summarizes several ways in which contextual information can help the brain resolve visual ambiguities and construct temporarily stable perceptual experiences. Temporal context through prior stimulation or internal brain states brought about by feedback from higher cortical processing levels may alter the response characteristics of specific neurons involved in rivalry resolution. Furthermore, spatial or crossmodal context may strengthen the neuronal representation of one of the possible perceptual interpretations and consequently bias the rivalry process towards it. We suggest that contextual influences on perceptual choices with ambiguous visual stimuli can be highly informative about the neuronal mechanisms of context-driven inference in the general processes of perceptual decision-making.     

Attention controls multisensory perception via two distinct mechanisms at different levels of the cortical hierarchy

To form a percept of the multisensory world, the brain needs to integrate signals from common sources weighted by their reliabilities and segregate those from independent sources. Previously, we have shown that anterior parietal cortices combine sensory signals into representations that take into account the signals’ causal structure (i.e., common versus independent sources) and their sensory reliabilities as predicted by Bayesian causal inference. The current study asks to what extent and how attentional mechanisms can actively control how sensory signals are combined for perceptual inference. In a pre- and postcueing paradigm, we presented observers with audiovisual signals at variable spatial disparities. Observers were precued to attend to auditory or visual modalities prior to stimulus presentation and postcued to report their perceived auditory or visual location. Combining psychophysics, functional magnetic resonance imaging (fMRI), and Bayesian modelling, we demonstrate that the brain moulds multisensory inference via two distinct mechanisms. Prestimulus attention to vision enhances the reliability and influence of visual inputs on spatial representations in visual and posterior parietal cortices. Poststimulus report determines how parietal cortices flexibly combine sensory estimates into spatial representations consistent with Bayesian causal inference. Our results show that distinct neural mechanisms control how signals are combined for perceptual inference at different levels of the cortical hierarchy.

A combination of psychophysics, computational modelling and fMRI reveals novel insights into how the brain controls the binding of information across the senses, such as the voice and lip movements of a speaker.     

The neural systems that mediate human perceptual decision making

Perceptual decision making is the act of choosing one option or course of action from a set of alternatives on the basis of available sensory evidence. Thus, when we make such decisions, sensory information must be interpreted and translated into behaviour. Neurophysiological work in monkeys performing sensory discriminations, combined with computational modelling, has paved the way for neuroimaging studies that are aimed at understanding decision-related processes in the human brain. Here we review findings from human neuroimaging studies in conjunction with data analysis methods that can directly link decisions and signals in the human brain on a trial-by-trial basis. This leads to a new view about the neural basis of human perceptual decision-making processes.