A Theory of Cortical Responses

The Wellcome Department of Imaging Neuroscience, Institute of Neurology, University College London, 12 Queen Square, London WC1N 3BG, UK.
Philosophical Transactions of The Royal Society B Biological Sciences (Impact Factor: 7.06). 05/2005; 360(1456):815-36. DOI: 10.1098/rstb.2005.1622
Source: PubMed


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.

Download full-text


Available from: Karl J Friston
  • Source
    • "The mechanics of somatosensory perception has begun to be investigated in terms of its dependency on the executive functions of frontoparietal networks as well as the " salience network " including anterior insula and midcingulate cortex. Much of this investigation has been based on " Hierarchical Predictive Coding (HPC) " accounts of perception [for example (Rao and Ballard, 1999;Friston, 2005Friston, , 2008] that may shed light on body misperceptions and neuroplasticity in CRPS. Here, we outline this theoretic approach, and in subsequent sections, we review supporting empirical data from neuroimaging studies of somatosensory perception and finally explore how this perspective could form the basis for identifying neurocognitive phenotypes of patients with CRPS. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Complex regional pain syndrome (CRPS) is a chronic, debilitating pain condition that usually arises after trauma to a limb, but its precise etiology remains elusive. Novel clinical signs based on body perceptual disturbances have been reported, but their pathophysiological mechanisms remain poorly understood. Investigators have used functional neuroimaging techniques (including MEG, EEG, fMRI, and PET) to study changes mainly within the somatosensory and motor cortices. Here, we provide a focused review of the neuroimaging research findings that have generated insights into the potential neurocognitive and neuroplastic mechanisms underlying perceptual disturbances in CRPS. Neuroimaging findings, particularly with regard to somatosensory processing, have been promising but limited by a number of technique-specific factors (such as the complexity of neuroimaging investigations, poor spatial resolution of EEG/MEG, and use of modeling procedures that do not draw causal inferences) and more general factors including small samples sizes and poorly characterized patients. These factors have led to an underappreciation of the potential heterogeneity of pathophysiology that may underlie variable clinical presentation in CRPS. Also, until now, neurological deficits have been predominantly investigated separately from perceptual and cognitive disturbances. Here, we highlight the need to identify neurocognitive phenotypes of patients with CRPS that are underpinned by causal explanations for perceptual disturbances. We suggest that a combination of larger cohorts, patient phenotyping, the use of both high temporal, and spatial resolution neuroimaging methods, and the identification of simplified biomarkers is likely to be the most fruitful approach to identifying neurocognitive phenotypes in CRPS. Based on our review, we explain how such phenotypes could be characterized in terms of hierarchical models of perception and corresponding disturbances in recurrent processing involving the somatosensory, salience and executive brain networks. We also draw attention to complementary neurological factors that may explain some CRPS symptoms, including the possibility of central neuroinflammation and neuronal atrophy, and how these phenomena may overlap but be partially separable from neurocognitive deficits.
    Full-text · Article · Feb 2016 · Frontiers in Human Neuroscience
    • "As such they do not depend on higher-level inputs and on the conscious interpretations of the environment. Mismatch Negativity (MMN) is a perfect candidate for testing such predictions in oddball paradigms (Baldeweg, 2006;Friston, 2005;Stefanics & Czigler, 2012;Wacongne, Changeux, & Dehaene, 2012). However, as we have seen above, the difference between rare and frequent stimuli can merely be explained by bottom-up RS effects without the necessity of predictive coding mechanisms (Kaliukhovich & Vogels, 2014). "
    [Show abstract] [Hide abstract]
    ABSTRACT: While in earlier work various local or bottom-up neural mechanisms were proposed to give rise to repetition suppression (RS), current theories suggest that top-down processes play a role in determining the repetition related reduction of the neural responses. In the current review we summarise those results, which support the role of these top-down processes, concentrating on the Bayesian models of predictive coding (PC). Such models assume that RS is related to the statistical probabilities of prior stimulus occurrences and to the future predictability of these stimuli. Here we review the current results that support or argue against this explanation. We point out that the heterogeneity of experimental manipulations that are thought to reflect predictive processes are likely to measure different processing steps, making their direct comparison difficult. In addition we emphasize the importance of identifying these sub-processes and clarifying their role in explaining RS. Finally, we propose a two-stage model for explaining the relationships of repetition and expectation phenomena in the human cortex.
    No preview · Article · Jan 2016 · Cortex
  • Source
    • "Prediction errors, on the other hand, are conveyed by ascending (driving) connections that use fast AMPA-type glutamate receptors. This asymmetry is not only supported by empirical differences between ascending and descending projections but can be deduced from the form of Equations 4 and 6: note that expectations are driven by a linear mixture of ascending prediction errors, while prediction errors depend upon non-linear functions of descending expectations (Friston, 2005). Thus, pharmacological manipulations of NMDA-and ACh-dependent processing should primarily influence learning and attention, leaving inference per se relatively intact. "
    [Show abstract] [Hide abstract]
    ABSTRACT: This paper presents a review of theoretical and empirical work on repetition suppression in the context of predictive coding. Predictive coding is a neurobiologically plausible scheme explaining how biological systems might perform perceptual inference and learning. From this perspective, repetition suppression is a manifestation of minimising prediction error through adaptive changes in predictions about the content and precision of sensory inputs. Simulations of artificial neural hierarchies provide a principled way of understanding how repetition suppression – at different time scales – can be explained in terms of inference and learning implemented under predictive coding. This formulation of repetition suppression is supported by results of numerous empirical studies of repetition suppression and its contextual determinants.
    Full-text · Article · Jan 2016 · Cortex
Show more

We use cookies to give you the best possible experience on ResearchGate. Read our cookies policy to learn more.