Information

How is the biological error signal in predictive coding computed?

How is the biological error signal in predictive coding computed?


We are searching data for your request:

Forums and discussions:
Manuals and reference books:
Data from registers:
Wait the end of the search in all databases.
Upon completion, a link will appear to access the found materials.

I am quite inspired by the results obtained by PredNet, which implements a predictive coding model using artificial neural networks. They compute the prediction error as a simple subtraction, and then split the positive and negative into two different feature maps.

Within the predictive coding framework it is believed that biological brains also uses the prediction error to communicate between different areas. How is this error signal computed in biological brains?.

Update

I am especially interested in the mechanical aspects, such as how doe does the brain compare a prediction with the true signal?. Synthetic approaches generally just use a pixel-wise error, and I am wondering if the brain does something analogously.


I provided an answer to a similar question here that limitedly deals with the role of biological prediction errors.

Here's an excerpt of that answer:

… to answer this properly, we must first make it clear that there are potentially dozens, hundreds, or an arbitrarily high number of other "prediction error types" in use by the brain. Here are just a few major ways, hypothetically:

  • Lots of different neurotransmitters (e.g. dopamine)
  • The opening/closing of various ion channel species that regulate the membrane potential
  • Synaptic vescicles/receptors
  • Neuronal firing rates (as in bursting, a rapid succession of action potentials)
  • Temporal coding (relative firing times to the firing of other neurons)
  • And I can think of 10 other more-subtle and harder to explain possibilities, but that are just as important, off the top of my head

Keep in mind that each neuron also seems to have its own differentiated mechanisms for, both, interpreting and signaling prediction error. This complicates things further. For instance, one neurotransmitter may communicate prediction error to one particular neuron, but has no effect (or a different effect) on a different neuron. It may even be that neurotransmitter X must be present while temporal code Y happens for the event to be interpreted as a prediction error.

The study of biological mechanisms for prediction error is a very complicated thing that has no simple interpretation, as opposed to what you find in artificial approaches (presumably, like your PredNet example). While we have not yet uncovered how the brain computes or uses these mechanisms to encode and communicate prediction errors across neurons, what is obvious is that the brain has to be doing some kind of prediction error coding. However, if we try to oversimplify what the brain is doing, we are likely to not have a very intelligent model. The fragility of predictions, susceptibility to error from data complications, and limited nature of current artificial intelligence implementations lends to the idea that there is a lack of good ideas about how to implement and fully utilize prediction error.

The work I do is of a theoretical nature so your question is right up my alley. I have some unique ideas on how various biological prediction errors may work, but it requires a lot of background to understand. Unfortunately, I also have not published my ideas so they are definitely not peer reviewed. That makes me somewhat reluctant to mention my personal ideas as an answer.

It's possible that a kind of prediction error is used in all the listed mechanisms to fine tune the respective properties. Each mechanism likely has a unique role that is central to intelligence- from not only prediction errors encountered in the environment, but even to intrinsic behaviors and being able to predict the outcome of its own actions.


TL;DR: We don't know whether the brain really uses predictive coding or not. But neurally computing an error signal on a small scale is possible (see below).

Predictive coding is an hypothesis for a putative signal-processing mechanism used in vertebrate brains. As things stand presently (2017), mapping the hypothesis of predictive coding onto known neural structures and responses is a matter of active research (e.g. Choi et al. 2016, Zmarz & Keller 2016, Roth et al 2016).

On a small-scale neural level, an error signal could be partially computed through converging an excitatory "prediction" signal and an inhibitory "evidence" signal onto a single neuron. This is only a partial error signal, since it is rectified by the spiking threshold of the post-synaptic neuron.

The main issue is that the forward and backwards pathways need to follow very precise wiring patterns, and be quite tightly aligned, for the simple formulations of predictive coding to work. That doesn't mesh well with intuitions of what is biologically reasonable.

There are some approaches to map a predictive coding framework onto other mechanisms that may be more biologically plausible (e.g. Spratling 2008).


References

Huang, G. Is this a unified theory of the brain? New Scientist 2658, 30–33 (2008).

Friston K., Kilner, J. & Harrison, L. A free energy principle for the brain. J. Physiol. Paris 100, 70–87 (2006). An overview of the free-energy principle that describes its motivation and relationship to generative models and predictive coding. This paper focuses on perception and the neurobiological infrastructures involved.

Ashby, W. R. Principles of the self-organising dynamic system. J. Gen. Psychol. 37, 125–128 (1947).

Nicolis, G. & Prigogine, I. Self-Organisation in Non-Equilibrium Systems (Wiley, New York, 1977).

Haken, H. Synergistics: an Introduction. Non-Equilibrium Phase Transition and Self-Organisation in Physics, Chemistry and Biology 3rd edn (Springer, New York, 1983).

Kauffman, S. The Origins of Order: Self-Organization and Selection in Evolution (Oxford Univ. Press, Oxford, 1993).

Bernard, C. Lectures on the Phenomena Common to Animals and Plants (Thomas, Springfield, 1974).

Applebaum, D. Probability and Information: an Integrated Approach (Cambridge Univ. Press, Cambridge, UK, 2008).

Evans, D. J. A non-equilibrium free energy theorem for deterministic systems. Mol. Physics 101, 15551–11554 (2003).

Crauel, H. & Flandoli, F. Attractors for random dynamical systems. Probab. Theory Relat. Fields 100, 365–393 (1994).

Feynman, R. P. Statistical Mechanics: a Set of Lectures (Benjamin, Reading, Massachusetts, 1972).

Hinton, G. E. & von Cramp, D. Keeping neural networks simple by minimising the description length of weights. Proc. 6th Annu. ACM Conf. Computational Learning Theory 5–13 (1993).

MacKay. D. J. C. Free-energy minimisation algorithm for decoding and cryptoanalysis. Electron. Lett. 31, 445–447 (1995).

Neal, R. M. & Hinton, G. E. in Learning in Graphical Models (ed. Jordan, M. I.) 355–368 (Kluwer Academic, Dordrecht, 1998).

Itti, L. & Baldi, P. Bayesian surprise attracts human attention. Vision Res. 49, 1295–1306 (2009).

Friston, K., Daunizeau, J. & Kiebel, S. Active inference or reinforcement learning? PLoS ONE 4, e6421 (2009).

Knill, D. C. & Pouget, A. The Bayesian brain: the role of uncertainty in neural coding and computation. Trends Neurosci. 27, 712–719 (2004). A nice review of Bayesian theories of perception and sensorimotor control. Its focus is on Bayes optimality in the brain and the implicit nature of neuronal representations.

von Helmholtz, H. in Treatise on Physiological Optics Vol. III 3rd edn (Voss, Hamburg, 1909).

MacKay, D. M. in Automata Studies (eds Shannon, C. E. & McCarthy, J.) 235–251 (Princeton Univ. Press, Princeton, 1956).

Neisser, U. Cognitive Psychology (Appleton-Century-Crofts, New York, 1967).

Gregory, R. L. Perceptual illusions and brain models. Proc. R. Soc. Lond. B Biol. Sci. 171, 179–196 (1968).

Gregory, R. L. Perceptions as hypotheses. Philos. Trans. R. Soc. Lond. B Biol. Sci. 290, 181–197 (1980).

Ballard, D. H., Hinton, G. E. & Sejnowski, T. J. Parallel visual computation. Nature 306, 21–26 (1983).

Kawato, M., Hayakawa, H. & Inui, T. A forward-inverse optics model of reciprocal connections between visual areas. Network: Computation in Neural Systems 4, 415–422 (1993).

Dayan, P., Hinton, G. E. & Neal, R. M. The Helmholtz machine. Neural Comput. 7, 889–904 (1995). This paper introduces the central role of generative models and variational approaches to hierarchical self-supervised learning and relates this to the function of bottom-up and top-down cortical processing pathways.

Lee, T. S. & Mumford, D. Hierarchical Bayesian inference in the visual cortex. J. Opt. Soc. Am. A Opt. Image Sci. Vis. 20, 1434–1448 (2003).

Kersten, D., Mamassian, P. & Yuille, A. Object perception as Bayesian inference. Annu. Rev. Psychol. 55, 271–304 (2004).

Friston, K. J. A theory of cortical responses. Philos. Trans. R. Soc. Lond. B Biol. Sci. 360, 815–836 (2005).

Beal, M. J. Variational Algorithms for Approximate Bayesian Inference. Thesis, University College London (2003).

Efron, B. & Morris, C. Stein's estimation rule and its competitors – an empirical Bayes approach. J. Am. Stats. Assoc. 68, 117–130 (1973).

Kass, R. E. & Steffey, D. Approximate Bayesian inference in conditionally independent hierarchical models (parametric empirical Bayes models). J. Am. Stat. Assoc. 407, 717–726 (1989).

Zeki, S. & Shipp, S. The functional logic of cortical connections. Nature 335, 311–317 (1988). Describes the functional architecture of cortical hierarchies with a focus on patterns of anatomical connections in the visual cortex. It emphasizes the role of functional segregation and integration (that is, message passing among cortical areas).

Felleman, D. J. & Van Essen, D. C. Distributed hierarchical processing in the primate cerebral cortex. Cereb. Cortex 1, 1–47 (1991).

Mesulam, M. M. From sensation to cognition. Brain 121, 1013–1052 (1998).

Sanger, T. Probability density estimation for the interpretation of neural population codes. J. Neurophysiol. 76, 2790–2793 (1996).

Zemel, R., Dayan, P. & Pouget, A. Probabilistic interpretation of population code. Neural Comput. 10, 403–430 (1998).

Paulin, M. G. Evolution of the cerebellum as a neuronal machine for Bayesian state estimation. J. Neural Eng. 2, S219–S234 (2005).

Ma, W. J., Beck, J. M., Latham, P. E. & Pouget, A. Bayesian inference with probabilistic population codes. Nature Neurosci. 9, 1432–1438 (2006).

Friston, K., Mattout, J., Trujillo-Barreto, N., Ashburner, J. & Penny, W. Variational free energy and the Laplace approximation. Neuroimage 34, 220–234 (2007).

Rao, R. P. & Ballard, D. H. Predictive coding in the visual cortex: a functional interpretation of some extra-classical receptive field effects. Nature Neurosci. 2, 79–87 (1998). Applies predictive coding to cortical processing to provide a compelling account of extra-classical receptive fields in the visual system. It emphasizes the importance of top-down projections in providing predictions, by modelling perceptual inference.

Mumford, D. On the computational architecture of the neocortex. II. The role of cortico-cortical loops. Biol. Cybern. 66, 241–251 (1992).

Friston, K. Hierarchical models in the brain. PLoS Comput. Biol. 4, e1000211 (2008).

Murray, S. O., Kersten, D., Olshausen, B. A., Schrater, P. & Woods, D. L. Shape perception reduces activity in human primary visual cortex. Proc. Natl Acad. Sci. USA 99, 15164–15169 (2002).

Garrido, M. I., Kilner, J. M., Kiebel, S. J. & Friston, K. J. Dynamic causal modeling of the response to frequency deviants. J. Neurophysiol. 101, 2620–2631 (2009).

Sherman, S. M. & Guillery, R. W. On the actions that one nerve cell can have on another: distinguishing “drivers” from “modulators”. Proc. Natl Acad. Sci. USA 95, 7121–7126 (1998).

Angelucci, A. & Bressloff, P. C. Contribution of feedforward, lateral and feedback connections to the classical receptive field center and extra-classical receptive field surround of primate V1 neurons. Prog. Brain Res. 154, 93–120 (2006).

Grossberg, S. Towards a unified theory of neocortex: laminar cortical circuits for vision and cognition. Prog. Brain Res. 165, 79–104 (2007).

Grossberg, S. & Versace, M. Spikes, synchrony, and attentive learning by laminar thalamocortical circuits. Brain Res. 1218, 278–312 (2008).

Barlow, H. in Sensory Communication (ed. Rosenblith, W.) 217–234 (MIT Press, Cambridge, Massachusetts, 1961).

Linsker, R. Perceptual neural organisation: some approaches based on network models and information theory. Annu. Rev. Neurosci. 13, 257–281 (1990).

Oja, E. Neural networks, principal components, and subspaces. Int. J. Neural Syst. 1, 61–68 (1989).

Bell, A. J. & Sejnowski, T. J. An information maximisation approach to blind separation and blind de-convolution. Neural Comput. 7, 1129–1159 (1995).

Atick, J. J. & Redlich, A. N. What does the retina know about natural scenes? Neural Comput. 4, 196–210 (1992).

Optican, L. & Richmond, B. J. Temporal encoding of two-dimensional patterns by single units in primate inferior cortex. III Information theoretic analysis. J. Neurophysiol. 57, 132–146 (1987).

Olshausen, B. A. & Field, D. J. Emergence of simple-cell receptive field properties by learning a sparse code for natural images. Nature 381, 607–609 (1996).

Simoncelli, E. P. & Olshausen, B. A. Natural image statistics and neural representation. Annu. Rev. Neurosci. 24, 1193–1216 (2001). A nice review of information theory in visual processing. It covers natural scene statistics and empirical tests of the efficient coding hypothesis in individual neurons and populations of neurons.

Friston, K. J. The labile brain. III. Transients and spatio-temporal receptive fields. Philos. Trans. R. Soc. Lond. B Biol. Sci. 355, 253–265 (2000).

Bialek, W., Nemenman, I. & Tishby, N. Predictability, complexity, and learning. Neural Comput. 13, 2409–2463 (2001).

Lewen, G. D., Bialek, W. & de Ruyter van Steveninck, R. R. Neural coding of naturalistic motion stimuli. Network 12, 317–329 (2001).

Laughlin, S. B. Efficiency and complexity in neural coding. Novartis Found. Symp. 239, 177–187 (2001).

Tipping, M. E. Sparse Bayesian learning and the Relevance Vector Machine. J. Machine Learn. Res. 1, 211–244 (2001).

Paus, T., Keshavan, M. & Giedd, J. N. Why do many psychiatric disorders emerge during adolescence? Nature Rev. Neurosci. 9, 947–957 (2008).

Gilestro, G. F., Tononi, G. & Cirelli, C. Widespread changes in synaptic markers as a function of sleep and wakefulness in Drosophila. Science 324, 109–112 (2009).

Roweis, S. & Ghahramani, Z. A unifying review of linear Gaussian models. Neural Comput. 11, 305–345 (1999).

Hebb, D. O. The Organization of Behaviour (Wiley, New York, 1949).

Paulsen, O. & Sejnowski, T. J. Natural patterns of activity and long-term synaptic plasticity. Curr. Opin. Neurobiol. 10, 172–179 (2000).

von der Malsburg, C. The Correlation Theory of Brain Function. Internal Report 81–82, Dept. Neurobiology, Max-Planck-Institute for Biophysical Chemistry (1981).

Singer, W. & Gray, C. M. Visual feature integration and the temporal correlation hypothesis. Annu. Rev. Neurosci. 18, 555–586 (1995).

Bienenstock, E. L., Cooper, L. N. & Munro, P. W. Theory for the development of neuron selectivity: orientation specificity and binocular interaction in visual cortex. J. Neurosci. 2, 32–48 (1982).

Abraham, W. C. & Bear, M. F. Metaplasticity: the plasticity of synaptic plasticity. Trends Neurosci. 19, 126–130 (1996).

Pareti, G. & De Palma, A. Does the brain oscillate? The dispute on neuronal synchronization. Neurol. Sci. 25, 41–47 (2004).

Leutgeb, S., Leutgeb, J. K., Moser, M. B. & Moser, E. I. Place cells, spatial maps and the population code for memory. Curr. Opin. Neurobiol. 15, 738–746 (2005).

Durstewitz, D. & Seamans, J. K. Beyond bistability: biophysics and temporal dynamics of working memory. Neuroscience 139, 119–133 (2006).

Anishchenko, A. & Treves, A. Autoassociative memory retrieval and spontaneous activity bumps in small-world networks of integrate-and-fire neurons. J. Physiol. Paris 100, 225–236 (2006).

Abbott, L. F., Varela, J. A., Sen, K. & Nelson, S. B. Synaptic depression and cortical gain control. Science 275, 220–224 (1997).

Yu, A. J. & Dayan, P. Uncertainty, neuromodulation and attention. Neuron 46, 681–692 (2005).

Doya, K. Metalearning and neuromodulation. Neural Netw. 15, 495–506 (2002).

Chawla, D., Lumer, E. D. & Friston, K. J. The relationship between synchronization among neuronal populations and their mean activity levels. Neural Comput. 11, 1389–1411 (1999).

Fries, P., Womelsdorf, T., Oostenveld, R. & Desimone, R. The effects of visual stimulation and selective visual attention on rhythmic neuronal synchronization in macaque area V4. J. Neurosci. 28, 4823–4835 (2008).

Womelsdorf, T. & Fries, P. Neuronal coherence during selective attentional processing and sensory-motor integration. J. Physiol. Paris 100, 182–193 (2006).

Desimone, R. Neural mechanisms for visual memory and their role in attention. Proc. Natl Acad. Sci. USA 93, 13494–13499 (1996). A nice review of mnemonic effects (such as repetition suppression) on neuronal responses and how they bias the competitive interactions between stimulus representations in the cortex. It provides a good perspective on attentional mechanisms in the visual system that is empirically grounded.

Treisman, A. Feature binding, attention and object perception. Philos. Trans. R. Soc. Lond. B Biol. Sci. 353, 1295–1306 (1998).

Maunsell, J. H. & Treue, S. Feature-based attention in visual cortex. Trends Neurosci. 29, 317–322 (2006).

Spratling, M. W. Predictive-coding as a model of biased competition in visual attention. Vision Res. 48, 1391–1408 (2008).

Reynolds, J. H. & Heeger, D. J. The normalization model of attention. Neuron 61, 168–185 (2009).

Schroeder, C. E., Mehta, A. D. & Foxe, J. J. Determinants and mechanisms of attentional modulation of neural processing. Front. Biosci. 6, D672–D684 (2001).

Hirayama, J., Yoshimoto, J. & Ishii, S. Bayesian representation learning in the cortex regulated by acetylcholine. Neural Netw. 17, 1391–1400 (2004).

Edelman, G. M. Neural Darwinism: selection and reentrant signaling in higher brain function. Neuron 10, 115–125 (1993).

Knobloch, F. Altruism and the hypothesis of meta-selection in human evolution. J. Am. Acad. Psychoanal. 29, 339–354 (2001).

Friston, K. J., Tononi, G., Reeke, G. N. Jr, Sporns, O. & Edelman, G. M. Value-dependent selection in the brain: simulation in a synthetic neural model. Neuroscience 59, 229–243 (1994).

Sutton, R. S. & Barto, A. G. Toward a modern theory of adaptive networks: expectation and prediction. Psychol. Rev. 88, 135–170 (1981).

Montague, P. R., Dayan, P., Person, C. & Sejnowski, T. J. Bee foraging in uncertain environments using predictive Hebbian learning. Nature 377, 725–728 (1995). A computational treatment of behaviour that combines ideas from optimal control theory and dynamic programming with the neurobiology of reward. This provided an early example of value learning in the brain.

Schultz, W. Predictive reward signal of dopamine neurons. J. Neurophysiol. 80, 1–27 (1998).

Daw, N. D. & Doya, K. The computational neurobiology of learning and reward. Curr. Opin. Neurobiol. 16, 199–204 (2006).

Redgrave, P. & Gurney, K. The short-latency dopamine signal: a role in discovering novel actions? Nature Rev. Neurosci. 7, 967–975 (2006).

Berridge, K. C. The debate over dopamine's role in reward: the case for incentive salience. Psychopharmacology (Berl.) 191, 391–431 (2007).

Sella, G. & Hirsh, A. E. The application of statistical physics to evolutionary biology. Proc. Natl Acad. Sci. USA 102, 9541–9546 (2005).

Rescorla, R. A. & Wagner, A. R. in Classical Conditioning II: Current Research and Theory (eds Black, A. H. & Prokasy, W. F.) 64–99 (Appleton Century Crofts, New York, 1972).

Bellman, R. On the Theory of Dynamic Programming. Proc. Natl Acad. Sci. USA 38, 716–719 (1952).

Watkins, C. J. C. H. & Dayan, P. Q-learning. Mach. Learn. 8, 279–292 (1992).

Todorov, E. in Advances in Neural Information Processing Systems (eds Scholkopf, B., Platt, J. & Hofmann T.) 19, 1369–1376 (MIT Press, 2006).

Camerer, C. F. Behavioural studies of strategic thinking in games. Trends Cogn. Sci. 7, 225–231 (2003).

Smith, J. M. & Price, G. R. The logic of animal conflict. Nature 246, 15–18 (1973).

Nash, J. Equilibrium points in n-person games. Proc. Natl Acad. Sci. USA 36, 48–49 (1950).

Wolpert, D. M. & Miall, R. C. Forward models for physiological motor control. Neural Netw. 9, 1265–1279 (1996).

Todorov, E. & Jordan, M. I. Smoothness maximization along a predefined path accurately predicts the speed profiles of complex arm movements. J. Neurophysiol. 80, 696–714 (1998).

Tseng, Y. W., Diedrichsen, J., Krakauer, J. W., Shadmehr, R. & Bastian, A. J. Sensory prediction-errors drive cerebellum-dependent adaptation of reaching. J. Neurophysiol. 98, 54–62 (2007).

Bays, P. M. & Wolpert, D. M. Computational principles of sensorimotor control that minimize uncertainty and variability. J. Physiol. 578, 387–396 (2007). A nice overview of computational principles in motor control. Its focus is on representing uncertainty and optimal estimation when extracting the sensory information required for motor planning.

Shadmehr, R. & Krakauer, J. W. A computational neuroanatomy for motor control. Exp. Brain Res. 185, 359–381 (2008).

Verschure, P. F., Voegtlin, T. & Douglas, R. J. Environmentally mediated synergy between perception and behaviour in mobile robots. Nature 425, 620–624 (2003).

Cohen, J. D., McClure, S. M. & Yu, A. J. Should I stay or should I go? How the human brain manages the trade-off between exploitation and exploration. Philos. Trans. R. Soc. Lond. B Biol. Sci. 362, 933–942 (2007).

Ishii, S., Yoshida, W. & Yoshimoto, J. Control of exploitation-exploration meta-parameter in reinforcement learning. Neural Netw. 15, 665–687 (2002).

Usher, M., Cohen, J. D., Servan-Schreiber, D., Rajkowski, J. & Aston-Jones, G. The role of locus coeruleus in the regulation of cognitive performance. Science 283, 549–554 (1999).

Voigt, C. A., Kauffman, S. & Wang, Z. G. Rational evolutionary design: the theory of in vitro protein evolution. Adv. Protein Chem. 55, 79–160 (2000).

Freeman, W. J. Characterization of state transitions in spatially distributed, chaotic, nonlinear, dynamical systems in cerebral cortex. Integr. Physiol. Behav. Sci. 29, 294–306 (1994).

Tsuda, I. Toward an interpretation of dynamic neural activity in terms of chaotic dynamical systems. Behav. Brain Sci. 24, 793–810 (2001).

Jirsa, V. K., Friedrich, R., Haken, H. & Kelso, J. A. A theoretical model of phase transitions in the human brain. Biol. Cybern. 71, 27–35 (1994). This paper develops a theoretical model (based on synergetics and nonlinear oscillator theory) that reproduces observed dynamics and suggests a formulation of biophysical coupling among brain systems.

Breakspear, M. & Stam, C. J. Dynamics of a neural system with a multiscale architecture. Philos. Trans. R. Soc. Lond. B Biol. Sci. 360, 1051–1074 (2005).

Bressler, S. L. & Tognoli, E. Operational principles of neurocognitive networks. Int. J. Psychophysiol. 60, 139–148 (2006).

Werner, G. Brain dynamics across levels of organization. J. Physiol. Paris 101, 273–279 (2007).

Pasquale, V., Massobrio, P., Bologna, L. L., Chiappalone, M. & Martinoia, S. Self-organization and neuronal avalanches in networks of dissociated cortical neurons. Neuroscience 153, 1354–1369 (2008).

Kitzbichler, M. G., Smith, M. L., Christensen, S. R. & Bullmore, E. Broadband criticality of human brain network synchronization. PLoS Comput. Biol. 5, e1000314 (2009).

Rabinovich, M., Huerta, R. & Laurent, G. Transient dynamics for neural processing. Science 321 48–50 (2008).

Tschacher, W. & Hake, H. Intentionality in non-equilibrium systems? The functional aspects of self-organised pattern formation. New Ideas Psychol. 25, 1–15 (2007).

Maturana, H. R. & Varela, F. De máquinas y seres vivos (Editorial Universitaria, Santiago, 1972). English translation available in Maturana, H. R. & Varela, F. in Autopoiesis and Cognition (Reidel, Dordrecht, 1980).

Fiorillo, C. D., Tobler, P. N. & Schultz, W. Discrete coding of reward probability and uncertainty by dopamine neurons. Science 299, 1898–1902 (2003).

Niv, Y., Duff, M. O. & Dayan, P. Dopamine, uncertainty and TD learning. Behav. Brain Funct. 1, 6 (2005).

Fletcher, P. C. & Frith, C. D. Perceiving is believing: a Bayesian approach to explaining the positive symptoms of schizophrenia. Nature Rev. Neurosci. 10, 48–58 (2009).

Phillips, W. A. & Silverstein, S. M. Convergence of biological and psychological perspectives on cognitive coordination in schizophrenia. Behav. Brain Sci. 26, 65–82 (2003).

Friston, K. & Kiebel, S. Cortical circuits for perceptual inference. Neural Netw. 22, 1093–1104 (2009).


We thank Scott Brincat for assistance with surgeries and data preprocessing and Morteza Moazami and Jefferson Roy for assistance with surgeries and animal training. We also thank the MIT veterinary staff and animal caretakers for their excellent support. We also thank Jaan Aru and Bruno Gomes for comments on the manuscript and Miles Whittington for many useful conversations. This work was supported by National Institutes of Mental Health Grant R37MH087027 and 5K99MH116100-02, Office of Naval Research Multidisciplinary University Research Initiatives Grant N00014-16-1-2832, and the MIT Picower Institute Faculty Innovation Fund.

↵ 2 Present address: Department of Psychology and Vanderbilt Brain Institute, Vanderbilt University, Nashville, TN 37240.


10 Jul 2019: Smout CA, Tang MF, Garrido MI, Mattingley JB (2019) Correction: Attention promotes the neural encoding of prediction errors. PLOS Biology 17(7): e3000368. https://doi.org/10.1371/journal.pbio.3000368 View correction

The encoding of sensory information in the human brain is thought to be optimised by two principal processes: ‘prediction’ uses stored information to guide the interpretation of forthcoming sensory events, and ‘attention’ prioritizes these events according to their behavioural relevance. Despite the ubiquitous contributions of attention and prediction to various aspects of perception and cognition, it remains unknown how they interact to modulate information processing in the brain. A recent extension of predictive coding theory suggests that attention optimises the expected precision of predictions by modulating the synaptic gain of prediction error units. Because prediction errors code for the difference between predictions and sensory signals, this model would suggest that attention increases the selectivity for mismatch information in the neural response to a surprising stimulus. Alternative predictive coding models propose that attention increases the activity of prediction (or ‘representation’) neurons and would therefore suggest that attention and prediction synergistically modulate selectivity for ‘feature information’ in the brain. Here, we applied forward encoding models to neural activity recorded via electroencephalography (EEG) as human observers performed a simple visual task to test for the effect of attention on both mismatch and feature information in the neural response to surprising stimuli. Participants attended or ignored a periodic stream of gratings, the orientations of which could be either predictable, surprising, or unpredictable. We found that surprising stimuli evoked neural responses that were encoded according to the difference between predicted and observed stimulus features, and that attention facilitated the encoding of this type of information in the brain. These findings advance our understanding of how attention and prediction modulate information processing in the brain, as well as support the theory that attention optimises precision expectations during hierarchical inference by increasing the gain of prediction errors.


Why visual perception is a decision process

Summary: Prediction errors play a role in the context of dynamic perceptual events that take place within fractions of a second. Findings support the hypothesis that visual perception occurs as a result of a decision process.

Source: RUB

Neuroscientists at the Ruhr-Universität Bochum (RUB) together with colleagues at the Freiburg University show that this is not strictly the case. Instead, they show that prediction errors can occasionally appear as visual illusion when viewing rapid image sequences. Thus, rather than being explained away prediction errors remain in fact accessible at final processing stages forming perception. Previous theories of predictive coding need therefore to be revised. The study is reported in Plos One on 4. May 2020.

Our visual system starts making predictions within a few milliseconds

To fixate objects in the outside world, our eyes perform far more than one hundred thousand of rapid movements per day called saccades. However, as soon as our eyes rest about 100 milliseconds, the brain starts making predictions. Differences between previous and current image contents are then forwarded to subsequent processing stages as prediction errors. The advantage to deal with differences instead of complete image information is obvious: similar to video compression techniques the data volume is drastically reduced. Another advantage turns up literally only at second sight: statistically, there is a high probability that the next saccade lands on locations where differences to previous image contents are largest. Thus, calculating potential changes of image content as the differences to previous contents prepares the visual system early on for new input.

To test whether the brain uses indeed such a strategy, the authors presented rapid sequences of two images to human volunteers. In the first image two gratings were superimposed, in the second image only one of the gratings was present. The task was to report the orientation of the last seen single grating. In most cases, the participants correctly reported the orientation of the present orientation, as expected. However, surprisingly, in some cases an orientation was perceived that was exactly orthogonal to the present orientation. That is, participants saw sometimes the difference between the previous superimposed gratings and the present single grating. „Seeing the difference instead of the real current input is here a visual illusion that can be interpreted as directly seeing the prediction error,” says Robert Staadt from the Institute of Neural Computation of the RUB, first author of the study.

Avoiding the pigeonhole benefits flexibility

“Within the framework of the predictive coding theory, prediction errors are mostly conceived in the context of higher cognitive functions that are coupled to conscious expectations. However, we demonstrate that prediction errors also play a role in the context of highly dynamic perceptual events that take place within fractions of a second,” explains Dr. Dirk Jancke, head of the Optical Imaging Group at the Institute of Neural Computation. The present study reveals that the visual system simultaneously keeps up information about past, current, and possible future image contents. Such a strategy allows both stability and flexibility when viewing rapid image sequences. “Altogether, our results support hypotheses that consider perception as a result of a decision process,” says Jancke. Hence, prediction errors should not be sorted out too early, as they might become relevant for following events.

Besides straightforward physical parameters like stimulus duration, brightness, and contrast, other, more elusive factors that characterize psychological features might be involved. Image is in the public domain.

Visual perception underlies decision making

In next studies the scientists will scrutinize the sets of parameters that drive the perceptual illusion most effectively. Besides straightforward physical parameters like stimulus duration, brightness, and contrast, other, more elusive factors that characterize psychological features might be involved. The authors’ long-term perspective is the development of practical visual tests that can be used for an early diagnosis of cognitive disorders connected to rapid perceptual decision processes.

Funding: The study was partly financed through grants of the Collaborative Research Centre (CRC) 874 at RUB, which is supported by the German Research Foundation since 2010. The CRC “Integration and representation of sensory processes” investigates how sensory signals generate neuronal maps, and result in complex behavior and memory formation.


Reinforcement learning framework

Reinforcement learning (RL) is perhaps the most influential framework developed to describe how an agent learns by interacting with its environment. RL is derived from the behaviorist view of animal behavior, in which an organism’s knowledge of the world is exclusively modeled based on its behavior. Crucially, RL theories focus on mechanistic accounts for behaviors based on several learning-related parameters established from empirical sources.

Both humans and nonhuman animals are excellent models for a variety of learning and decision-making tasks that are grounded on RL theories. Describing learning and learned outcomes through mathematical models is a powerful way to make explicit and testable predictions about how an organism will behave in a particular context and how they will make decisions that take into account internal states, such as motivation and subjective value. 27 The RL framework can capture seemingly complex behaviors with relatively simple yet elegant rules, as in the famous Rescorla–Wagner model. 28 Although various RL models differ in how they describe different cognitive phenomena, they share several core elements, such as the rate of learning or the salience of stimuli, to fit the specifics of learning and decision-making processes.

RL has its roots and applications in both engineering and psychology. RL has its core foundations in the work of Richard Bellman, most famous for developing the Bellman optimality equation and dynamic programming. The more widely appreciated root of RL is conceptualizing how organisms gather information from their environment to learn and make decisions. RL requires an agent that moves through different states, or contexts, in a given environment. Other necessary components include a reward signal, a value function, and a policy. Reward outcome is central to all forms of RL and consists of a quantity the agent gets as a result of its actions within the environment. The agent then computes a value function using that reward outcome that calculates the expected value of certain states/contexts as well as the conjunction of specific states and actions. The agent uses these value functions to develop a set of preferred actions, known as a policy. A model of the environment is an optional component of RL that can provide the organism with guidance on how to move from state to state.

In dynamic programming, developed by Bellman for engineering applications, a complete model of the environment is required. This idea requires the action of an agent to be guided by the expected payoff of the action in addition to the total expected payoff of potential actions in hypothetical future states. 29 The same principle applies to temporal discounting (TD) models, the predominant form of RL model applied in psychological studies of humans and other animals. 30 TD learning notably differs from dynamic programming, as it does not require any model of the environment. Instead, learning is accomplished by comparing expected reward to actual reward after a certain transition in time. This difference is the reward prediction error. This prediction error is used to update the value function and, ultimately, the policy of an agent interacting with its environment. Prediction error signaling is indeed the fundamental attribute of the original models of learning. 28 In simple terms, a prediction error calculates the difference between what the animal expects to have happen and what actually happens to the animal on a given event or trial. 31 This can also be described as an error signal. 32


An alternative to the two-factor model of delusions

Fletcher & Frith ( Reference Fletcher and Frith 2009) have argued that a two-stage theory of psychosis in schizophrenia is unnecessary, as both perception and belief rely on predictions and updating those predictions in the light of evidence (so-called prediction error). Frith has shown the importance in health of being able to monitor one's own thoughts and actions subconsciously, and of being able to predict the consequences of those actions using the so-called efference-copy or corollary discharge mechanism. He and others have shown that patients with schizophrenia do indeed have problems in predictive coding and monitoring of their actions (Feinberg, Reference Feinberg 1978 McGuire et al. Reference McGuire, Silbersweig, Wright, Murray, David, Frackowiak and Frith 1995 Shergill et al. Reference Shergill, Samson, Bays, Frith and Wolpert 2005 Mathalon & Ford, Reference Mathalon and Ford 2008). By linking this theory of abnormal predictive coding in psychosis to evidence that dopamine-driven prediction error signalling is also abnormal in psychosis, Fletcher and Frith argue that abnormal predictive coding at multiple levels of brain complexity could lead to both false perceptions and false beliefs in schizophrenia. This provides a unitary psychological process that could underlie all positive psychotic symptoms. This simplicity and parsimony is a very attractive aspect of the theory a disadvantage is that this parsimony may not be reflected at a biological level. For example, it is unlikely that a single neurotransmitter system could be responsible for predictive coding throughout the brain. Whilst predictive coding for rewards may be intricately linked to dopamine, it is unlikely that predictive coding for visual perception will have the same biological basis. Empirical work combining pharmacological, imaging and psychological studies is required to test this theory.


Author Summary

Perception inevitably depends on combining sensory input with prior expectations. This is particularly critical for identifying degraded input. However, the underlying neural mechanism by which expectations influence sensory processing is unclear. Predictive Coding theories suggest that the brain passes forward the unexpected part of the sensory input while expected properties are suppressed (i.e., Prediction Error). However, evidence to rule out the opposite mechanism in which the expected part of the sensory input is enhanced or sharpened (i.e., Sharpening) has been lacking. In this study, we investigate the neural mechanisms by which sensory clarity and prior knowledge influence the perception of degraded speech. A univariate measure of brain activity obtained from functional magnetic resonance imaging (fMRI) is in line with both neural mechanisms (Prediction Error and Sharpening). However, combining multivariate fMRI measures with computational simulations allows us to determine the underlying mechanism. Our key finding was an interaction between sensory input and prior expectations: for unexpected speech, increasing speech clarity increases the amount of information represented in sensory brain areas. In contrast, for speech that matches prior expectations, increasing speech clarity reduces the amount of this information. Our observations are uniquely simulated by a model of speech perception that includes Prediction Errors.

Citation: Blank H, Davis MH (2016) Prediction Errors but Not Sharpened Signals Simulate Multivoxel fMRI Patterns during Speech Perception. PLoS Biol 14(11): e1002577. https://doi.org/10.1371/journal.pbio.1002577

Academic Editor: Robert Zatorre, McGill University, CANADA

Received: May 10, 2016 Accepted: October 19, 2016 Published: November 15, 2016

Copyright: © 2016 Blank, Davis. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Data Availability: The data is available at https://osf.io/2ze9n/ doi: 10.17605/OSF.IO/2ZE9N

Funding: This research was supported by UK Medical Research Council (MRC) funding of the Cognition and Brain Sciences Unit MC-A060-5PQ80 (to MHD). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Competing interests: The authors have declared that no competing interests exist.

Abbreviations: fMRI, functional magnetic resonance imaging IFG, inferior frontal gyrus RDM, representational dissimilarity matrix ROI, region of interest RSA, representational similarity analysis STG, superior temporal gyrus STS, superior temporal sulcus


Does signal reduction imply predictive coding in models of spoken word recognition?

Pervasive behavioral and neural evidence for predictive processing has led to claims that language processing depends upon predictive coding. Formally, predictive coding is a computational mechanism where only deviations from top-down expectations are passed between levels of representation. In many cognitive neuroscience studies, a reduction of signal for expected inputs is taken as being diagnostic of predictive coding. In the present work, we show that despite not explicitly implementing prediction, the TRACE model of speech perception exhibits this putative hallmark of predictive coding, with reductions in total lexical activation, total lexical feedback, and total phoneme activation when the input conforms to expectations. These findings may indicate that interactive activation is functionally equivalent or approximant to predictive coding or that caution is warranted in interpreting neural signal reduction as diagnostic of predictive coding.


The Prediction Error Account of Art – Current Formulation and its Contradiction

The basic tenet of Van de Cruys and Wagemans (2011) model is that a temporary state of unpredictability (or prediction error) is important for the emergence of perceptual pleasure vis-a-vis a work of art. Understanding perception in terms of predictions means that it is possible for perceptual configuration to induce different sequences of affect and to do so partly independently of the particular content of perception (TPEA, 1040). Accordingly, artists are supposed to intentionally create incongruities (perceived as prediction errors) that may not be possible in a natural visual environment, and viewers are able to tolerate and even enjoy the unpredictability because they expect to be surprised in their encounters with art (TPEA, 1041). By delaying prediction confirmation, artists create a positive affect: the viewer quickly runs into incongruities, which presumably generate an arousal aimed at reducing prediction errors. It is this incompatibility (or prediction error) that is the source of some of the emotionality of a work of art. In other words, artists intuitively attempt to strike the optimal balance between predictability and surprise. The mental effort required of a person in order to cope with the prediction error is a condition sine qua non for registering the perceptual pleasure of a Gestalt formation (prediction error reduction). According to Van de Cruys and Wagemans (2011): “Only by using minimal prediction errors painters can ensure that viewers will obtain their reward and not give up prematurely. Final gratification postponed as long as the artist has hidden in the painting enough micro reward the viewer can discover. ” (TPEA, 1050).

This concept is appealing for a number of reasons. First, it offers a much stronger explanatory framework than the neuroscientific and psychological models of art experience, which focus exclusively on a bottom–up account of visual processing (Zeki, 1999 Shimamura, 2013) and neglect or downplay the role of the top𠄽own, inferential activity of brain/mind. Second, although not specifically stated by the authors, the model responds to the so-called �rk room problem,” the apparent paradox that, in order to minimize surprise, agents should avoid sensory stimulation altogether and should proceed directly to the least stimulating environment and stay there they should take up a position in the nearest �rk room” and never move again. Neatly summarized: avoid surprises and you will last longer. Predictive coding theorists offer a simple solution to the dark room scenario: prior beliefs render dark rooms surprising. That is, agents that predict rich stimulating environments will find the �rk room” surprising and will leave at the earliest opportunity. The postulate of surprise minimization therefore by no means inhibits subjects from active, exploratory behavior and novelty-seeking, including presumably an aesthetic experience (Friston et al., 2012b Schwartenbeck et al., 2013).

Third, the model is apparently compatible with some well-established and respected art-historical theories, notably Gombrich’s (1960) theories of the prognostic character of the perception of pictures, the role of the beholder’s share and the viewer’s inferences in perception, and his notion of the artist working through a cycle of scheme and correction. Indeed, Gombrich’s (1960) famous maxim that “[t]o read the artist’s picture is to mobilize our memories and experience of the visible world and to test his image through tentative projections” leaves the door open to the Bayesian brain perspective an interesting challenge would be to recast Gombrich’s account in the explanatory terms of predictive coding, but this will not be pursued here. Furthermore, it provides scientific footing to some philosophical interpretations of aesthetic experience, most notably Gadamer’s (1975/2004) hermeneutical scenario, which highlights the activity of the perceiving subject vis-à-vis the aesthetic object. Gadamer describes the nature of this exchange as ongoing and dynamic, suggesting that understanding is an open-ended or at least an extended process that does not end the moment the representational content is identified or the information embedded in the work of art obtained, but also includes a more complex response and understanding: 𠇊ll encounter with the language of art is an encounter with an unfinished event and is itself part of this event. There is no absolute progress and no final exhaustion of what lies in work of art” (Gadamer, 1975/2004, p. 85).


References

Huang, G. Is this a unified theory of the brain? New Scientist 2658, 30–33 (2008).

Friston K., Kilner, J. & Harrison, L. A free energy principle for the brain. J. Physiol. Paris 100, 70–87 (2006). An overview of the free-energy principle that describes its motivation and relationship to generative models and predictive coding. This paper focuses on perception and the neurobiological infrastructures involved.

Ashby, W. R. Principles of the self-organising dynamic system. J. Gen. Psychol. 37, 125–128 (1947).

Nicolis, G. & Prigogine, I. Self-Organisation in Non-Equilibrium Systems (Wiley, New York, 1977).

Haken, H. Synergistics: an Introduction. Non-Equilibrium Phase Transition and Self-Organisation in Physics, Chemistry and Biology 3rd edn (Springer, New York, 1983).

Kauffman, S. The Origins of Order: Self-Organization and Selection in Evolution (Oxford Univ. Press, Oxford, 1993).

Bernard, C. Lectures on the Phenomena Common to Animals and Plants (Thomas, Springfield, 1974).

Applebaum, D. Probability and Information: an Integrated Approach (Cambridge Univ. Press, Cambridge, UK, 2008).

Evans, D. J. A non-equilibrium free energy theorem for deterministic systems. Mol. Physics 101, 15551–11554 (2003).

Crauel, H. & Flandoli, F. Attractors for random dynamical systems. Probab. Theory Relat. Fields 100, 365–393 (1994).

Feynman, R. P. Statistical Mechanics: a Set of Lectures (Benjamin, Reading, Massachusetts, 1972).

Hinton, G. E. & von Cramp, D. Keeping neural networks simple by minimising the description length of weights. Proc. 6th Annu. ACM Conf. Computational Learning Theory 5–13 (1993).

MacKay. D. J. C. Free-energy minimisation algorithm for decoding and cryptoanalysis. Electron. Lett. 31, 445–447 (1995).

Neal, R. M. & Hinton, G. E. in Learning in Graphical Models (ed. Jordan, M. I.) 355–368 (Kluwer Academic, Dordrecht, 1998).

Itti, L. & Baldi, P. Bayesian surprise attracts human attention. Vision Res. 49, 1295–1306 (2009).

Friston, K., Daunizeau, J. & Kiebel, S. Active inference or reinforcement learning? PLoS ONE 4, e6421 (2009).

Knill, D. C. & Pouget, A. The Bayesian brain: the role of uncertainty in neural coding and computation. Trends Neurosci. 27, 712–719 (2004). A nice review of Bayesian theories of perception and sensorimotor control. Its focus is on Bayes optimality in the brain and the implicit nature of neuronal representations.

von Helmholtz, H. in Treatise on Physiological Optics Vol. III 3rd edn (Voss, Hamburg, 1909).

MacKay, D. M. in Automata Studies (eds Shannon, C. E. & McCarthy, J.) 235–251 (Princeton Univ. Press, Princeton, 1956).

Neisser, U. Cognitive Psychology (Appleton-Century-Crofts, New York, 1967).

Gregory, R. L. Perceptual illusions and brain models. Proc. R. Soc. Lond. B Biol. Sci. 171, 179–196 (1968).

Gregory, R. L. Perceptions as hypotheses. Philos. Trans. R. Soc. Lond. B Biol. Sci. 290, 181–197 (1980).

Ballard, D. H., Hinton, G. E. & Sejnowski, T. J. Parallel visual computation. Nature 306, 21–26 (1983).

Kawato, M., Hayakawa, H. & Inui, T. A forward-inverse optics model of reciprocal connections between visual areas. Network: Computation in Neural Systems 4, 415–422 (1993).

Dayan, P., Hinton, G. E. & Neal, R. M. The Helmholtz machine. Neural Comput. 7, 889–904 (1995). This paper introduces the central role of generative models and variational approaches to hierarchical self-supervised learning and relates this to the function of bottom-up and top-down cortical processing pathways.

Lee, T. S. & Mumford, D. Hierarchical Bayesian inference in the visual cortex. J. Opt. Soc. Am. A Opt. Image Sci. Vis. 20, 1434–1448 (2003).

Kersten, D., Mamassian, P. & Yuille, A. Object perception as Bayesian inference. Annu. Rev. Psychol. 55, 271–304 (2004).

Friston, K. J. A theory of cortical responses. Philos. Trans. R. Soc. Lond. B Biol. Sci. 360, 815–836 (2005).

Beal, M. J. Variational Algorithms for Approximate Bayesian Inference. Thesis, University College London (2003).

Efron, B. & Morris, C. Stein's estimation rule and its competitors – an empirical Bayes approach. J. Am. Stats. Assoc. 68, 117–130 (1973).

Kass, R. E. & Steffey, D. Approximate Bayesian inference in conditionally independent hierarchical models (parametric empirical Bayes models). J. Am. Stat. Assoc. 407, 717–726 (1989).

Zeki, S. & Shipp, S. The functional logic of cortical connections. Nature 335, 311–317 (1988). Describes the functional architecture of cortical hierarchies with a focus on patterns of anatomical connections in the visual cortex. It emphasizes the role of functional segregation and integration (that is, message passing among cortical areas).

Felleman, D. J. & Van Essen, D. C. Distributed hierarchical processing in the primate cerebral cortex. Cereb. Cortex 1, 1–47 (1991).

Mesulam, M. M. From sensation to cognition. Brain 121, 1013–1052 (1998).

Sanger, T. Probability density estimation for the interpretation of neural population codes. J. Neurophysiol. 76, 2790–2793 (1996).

Zemel, R., Dayan, P. & Pouget, A. Probabilistic interpretation of population code. Neural Comput. 10, 403–430 (1998).

Paulin, M. G. Evolution of the cerebellum as a neuronal machine for Bayesian state estimation. J. Neural Eng. 2, S219–S234 (2005).

Ma, W. J., Beck, J. M., Latham, P. E. & Pouget, A. Bayesian inference with probabilistic population codes. Nature Neurosci. 9, 1432–1438 (2006).

Friston, K., Mattout, J., Trujillo-Barreto, N., Ashburner, J. & Penny, W. Variational free energy and the Laplace approximation. Neuroimage 34, 220–234 (2007).

Rao, R. P. & Ballard, D. H. Predictive coding in the visual cortex: a functional interpretation of some extra-classical receptive field effects. Nature Neurosci. 2, 79–87 (1998). Applies predictive coding to cortical processing to provide a compelling account of extra-classical receptive fields in the visual system. It emphasizes the importance of top-down projections in providing predictions, by modelling perceptual inference.

Mumford, D. On the computational architecture of the neocortex. II. The role of cortico-cortical loops. Biol. Cybern. 66, 241–251 (1992).

Friston, K. Hierarchical models in the brain. PLoS Comput. Biol. 4, e1000211 (2008).

Murray, S. O., Kersten, D., Olshausen, B. A., Schrater, P. & Woods, D. L. Shape perception reduces activity in human primary visual cortex. Proc. Natl Acad. Sci. USA 99, 15164–15169 (2002).

Garrido, M. I., Kilner, J. M., Kiebel, S. J. & Friston, K. J. Dynamic causal modeling of the response to frequency deviants. J. Neurophysiol. 101, 2620–2631 (2009).

Sherman, S. M. & Guillery, R. W. On the actions that one nerve cell can have on another: distinguishing “drivers” from “modulators”. Proc. Natl Acad. Sci. USA 95, 7121–7126 (1998).

Angelucci, A. & Bressloff, P. C. Contribution of feedforward, lateral and feedback connections to the classical receptive field center and extra-classical receptive field surround of primate V1 neurons. Prog. Brain Res. 154, 93–120 (2006).

Grossberg, S. Towards a unified theory of neocortex: laminar cortical circuits for vision and cognition. Prog. Brain Res. 165, 79–104 (2007).

Grossberg, S. & Versace, M. Spikes, synchrony, and attentive learning by laminar thalamocortical circuits. Brain Res. 1218, 278–312 (2008).

Barlow, H. in Sensory Communication (ed. Rosenblith, W.) 217–234 (MIT Press, Cambridge, Massachusetts, 1961).

Linsker, R. Perceptual neural organisation: some approaches based on network models and information theory. Annu. Rev. Neurosci. 13, 257–281 (1990).

Oja, E. Neural networks, principal components, and subspaces. Int. J. Neural Syst. 1, 61–68 (1989).

Bell, A. J. & Sejnowski, T. J. An information maximisation approach to blind separation and blind de-convolution. Neural Comput. 7, 1129–1159 (1995).

Atick, J. J. & Redlich, A. N. What does the retina know about natural scenes? Neural Comput. 4, 196–210 (1992).

Optican, L. & Richmond, B. J. Temporal encoding of two-dimensional patterns by single units in primate inferior cortex. III Information theoretic analysis. J. Neurophysiol. 57, 132–146 (1987).

Olshausen, B. A. & Field, D. J. Emergence of simple-cell receptive field properties by learning a sparse code for natural images. Nature 381, 607–609 (1996).

Simoncelli, E. P. & Olshausen, B. A. Natural image statistics and neural representation. Annu. Rev. Neurosci. 24, 1193–1216 (2001). A nice review of information theory in visual processing. It covers natural scene statistics and empirical tests of the efficient coding hypothesis in individual neurons and populations of neurons.

Friston, K. J. The labile brain. III. Transients and spatio-temporal receptive fields. Philos. Trans. R. Soc. Lond. B Biol. Sci. 355, 253–265 (2000).

Bialek, W., Nemenman, I. & Tishby, N. Predictability, complexity, and learning. Neural Comput. 13, 2409–2463 (2001).

Lewen, G. D., Bialek, W. & de Ruyter van Steveninck, R. R. Neural coding of naturalistic motion stimuli. Network 12, 317–329 (2001).

Laughlin, S. B. Efficiency and complexity in neural coding. Novartis Found. Symp. 239, 177–187 (2001).

Tipping, M. E. Sparse Bayesian learning and the Relevance Vector Machine. J. Machine Learn. Res. 1, 211–244 (2001).

Paus, T., Keshavan, M. & Giedd, J. N. Why do many psychiatric disorders emerge during adolescence? Nature Rev. Neurosci. 9, 947–957 (2008).

Gilestro, G. F., Tononi, G. & Cirelli, C. Widespread changes in synaptic markers as a function of sleep and wakefulness in Drosophila. Science 324, 109–112 (2009).

Roweis, S. & Ghahramani, Z. A unifying review of linear Gaussian models. Neural Comput. 11, 305–345 (1999).

Hebb, D. O. The Organization of Behaviour (Wiley, New York, 1949).

Paulsen, O. & Sejnowski, T. J. Natural patterns of activity and long-term synaptic plasticity. Curr. Opin. Neurobiol. 10, 172–179 (2000).

von der Malsburg, C. The Correlation Theory of Brain Function. Internal Report 81–82, Dept. Neurobiology, Max-Planck-Institute for Biophysical Chemistry (1981).

Singer, W. & Gray, C. M. Visual feature integration and the temporal correlation hypothesis. Annu. Rev. Neurosci. 18, 555–586 (1995).

Bienenstock, E. L., Cooper, L. N. & Munro, P. W. Theory for the development of neuron selectivity: orientation specificity and binocular interaction in visual cortex. J. Neurosci. 2, 32–48 (1982).

Abraham, W. C. & Bear, M. F. Metaplasticity: the plasticity of synaptic plasticity. Trends Neurosci. 19, 126–130 (1996).

Pareti, G. & De Palma, A. Does the brain oscillate? The dispute on neuronal synchronization. Neurol. Sci. 25, 41–47 (2004).

Leutgeb, S., Leutgeb, J. K., Moser, M. B. & Moser, E. I. Place cells, spatial maps and the population code for memory. Curr. Opin. Neurobiol. 15, 738–746 (2005).

Durstewitz, D. & Seamans, J. K. Beyond bistability: biophysics and temporal dynamics of working memory. Neuroscience 139, 119–133 (2006).

Anishchenko, A. & Treves, A. Autoassociative memory retrieval and spontaneous activity bumps in small-world networks of integrate-and-fire neurons. J. Physiol. Paris 100, 225–236 (2006).

Abbott, L. F., Varela, J. A., Sen, K. & Nelson, S. B. Synaptic depression and cortical gain control. Science 275, 220–224 (1997).

Yu, A. J. & Dayan, P. Uncertainty, neuromodulation and attention. Neuron 46, 681–692 (2005).

Doya, K. Metalearning and neuromodulation. Neural Netw. 15, 495–506 (2002).

Chawla, D., Lumer, E. D. & Friston, K. J. The relationship between synchronization among neuronal populations and their mean activity levels. Neural Comput. 11, 1389–1411 (1999).

Fries, P., Womelsdorf, T., Oostenveld, R. & Desimone, R. The effects of visual stimulation and selective visual attention on rhythmic neuronal synchronization in macaque area V4. J. Neurosci. 28, 4823–4835 (2008).

Womelsdorf, T. & Fries, P. Neuronal coherence during selective attentional processing and sensory-motor integration. J. Physiol. Paris 100, 182–193 (2006).

Desimone, R. Neural mechanisms for visual memory and their role in attention. Proc. Natl Acad. Sci. USA 93, 13494–13499 (1996). A nice review of mnemonic effects (such as repetition suppression) on neuronal responses and how they bias the competitive interactions between stimulus representations in the cortex. It provides a good perspective on attentional mechanisms in the visual system that is empirically grounded.

Treisman, A. Feature binding, attention and object perception. Philos. Trans. R. Soc. Lond. B Biol. Sci. 353, 1295–1306 (1998).

Maunsell, J. H. & Treue, S. Feature-based attention in visual cortex. Trends Neurosci. 29, 317–322 (2006).

Spratling, M. W. Predictive-coding as a model of biased competition in visual attention. Vision Res. 48, 1391–1408 (2008).

Reynolds, J. H. & Heeger, D. J. The normalization model of attention. Neuron 61, 168–185 (2009).

Schroeder, C. E., Mehta, A. D. & Foxe, J. J. Determinants and mechanisms of attentional modulation of neural processing. Front. Biosci. 6, D672–D684 (2001).

Hirayama, J., Yoshimoto, J. & Ishii, S. Bayesian representation learning in the cortex regulated by acetylcholine. Neural Netw. 17, 1391–1400 (2004).

Edelman, G. M. Neural Darwinism: selection and reentrant signaling in higher brain function. Neuron 10, 115–125 (1993).

Knobloch, F. Altruism and the hypothesis of meta-selection in human evolution. J. Am. Acad. Psychoanal. 29, 339–354 (2001).

Friston, K. J., Tononi, G., Reeke, G. N. Jr, Sporns, O. & Edelman, G. M. Value-dependent selection in the brain: simulation in a synthetic neural model. Neuroscience 59, 229–243 (1994).

Sutton, R. S. & Barto, A. G. Toward a modern theory of adaptive networks: expectation and prediction. Psychol. Rev. 88, 135–170 (1981).

Montague, P. R., Dayan, P., Person, C. & Sejnowski, T. J. Bee foraging in uncertain environments using predictive Hebbian learning. Nature 377, 725–728 (1995). A computational treatment of behaviour that combines ideas from optimal control theory and dynamic programming with the neurobiology of reward. This provided an early example of value learning in the brain.

Schultz, W. Predictive reward signal of dopamine neurons. J. Neurophysiol. 80, 1–27 (1998).

Daw, N. D. & Doya, K. The computational neurobiology of learning and reward. Curr. Opin. Neurobiol. 16, 199–204 (2006).

Redgrave, P. & Gurney, K. The short-latency dopamine signal: a role in discovering novel actions? Nature Rev. Neurosci. 7, 967–975 (2006).

Berridge, K. C. The debate over dopamine's role in reward: the case for incentive salience. Psychopharmacology (Berl.) 191, 391–431 (2007).

Sella, G. & Hirsh, A. E. The application of statistical physics to evolutionary biology. Proc. Natl Acad. Sci. USA 102, 9541–9546 (2005).

Rescorla, R. A. & Wagner, A. R. in Classical Conditioning II: Current Research and Theory (eds Black, A. H. & Prokasy, W. F.) 64–99 (Appleton Century Crofts, New York, 1972).

Bellman, R. On the Theory of Dynamic Programming. Proc. Natl Acad. Sci. USA 38, 716–719 (1952).

Watkins, C. J. C. H. & Dayan, P. Q-learning. Mach. Learn. 8, 279–292 (1992).

Todorov, E. in Advances in Neural Information Processing Systems (eds Scholkopf, B., Platt, J. & Hofmann T.) 19, 1369–1376 (MIT Press, 2006).

Camerer, C. F. Behavioural studies of strategic thinking in games. Trends Cogn. Sci. 7, 225–231 (2003).

Smith, J. M. & Price, G. R. The logic of animal conflict. Nature 246, 15–18 (1973).

Nash, J. Equilibrium points in n-person games. Proc. Natl Acad. Sci. USA 36, 48–49 (1950).

Wolpert, D. M. & Miall, R. C. Forward models for physiological motor control. Neural Netw. 9, 1265–1279 (1996).

Todorov, E. & Jordan, M. I. Smoothness maximization along a predefined path accurately predicts the speed profiles of complex arm movements. J. Neurophysiol. 80, 696–714 (1998).

Tseng, Y. W., Diedrichsen, J., Krakauer, J. W., Shadmehr, R. & Bastian, A. J. Sensory prediction-errors drive cerebellum-dependent adaptation of reaching. J. Neurophysiol. 98, 54–62 (2007).

Bays, P. M. & Wolpert, D. M. Computational principles of sensorimotor control that minimize uncertainty and variability. J. Physiol. 578, 387–396 (2007). A nice overview of computational principles in motor control. Its focus is on representing uncertainty and optimal estimation when extracting the sensory information required for motor planning.

Shadmehr, R. & Krakauer, J. W. A computational neuroanatomy for motor control. Exp. Brain Res. 185, 359–381 (2008).

Verschure, P. F., Voegtlin, T. & Douglas, R. J. Environmentally mediated synergy between perception and behaviour in mobile robots. Nature 425, 620–624 (2003).

Cohen, J. D., McClure, S. M. & Yu, A. J. Should I stay or should I go? How the human brain manages the trade-off between exploitation and exploration. Philos. Trans. R. Soc. Lond. B Biol. Sci. 362, 933–942 (2007).

Ishii, S., Yoshida, W. & Yoshimoto, J. Control of exploitation-exploration meta-parameter in reinforcement learning. Neural Netw. 15, 665–687 (2002).

Usher, M., Cohen, J. D., Servan-Schreiber, D., Rajkowski, J. & Aston-Jones, G. The role of locus coeruleus in the regulation of cognitive performance. Science 283, 549–554 (1999).

Voigt, C. A., Kauffman, S. & Wang, Z. G. Rational evolutionary design: the theory of in vitro protein evolution. Adv. Protein Chem. 55, 79–160 (2000).

Freeman, W. J. Characterization of state transitions in spatially distributed, chaotic, nonlinear, dynamical systems in cerebral cortex. Integr. Physiol. Behav. Sci. 29, 294–306 (1994).

Tsuda, I. Toward an interpretation of dynamic neural activity in terms of chaotic dynamical systems. Behav. Brain Sci. 24, 793–810 (2001).

Jirsa, V. K., Friedrich, R., Haken, H. & Kelso, J. A. A theoretical model of phase transitions in the human brain. Biol. Cybern. 71, 27–35 (1994). This paper develops a theoretical model (based on synergetics and nonlinear oscillator theory) that reproduces observed dynamics and suggests a formulation of biophysical coupling among brain systems.

Breakspear, M. & Stam, C. J. Dynamics of a neural system with a multiscale architecture. Philos. Trans. R. Soc. Lond. B Biol. Sci. 360, 1051–1074 (2005).

Bressler, S. L. & Tognoli, E. Operational principles of neurocognitive networks. Int. J. Psychophysiol. 60, 139–148 (2006).

Werner, G. Brain dynamics across levels of organization. J. Physiol. Paris 101, 273–279 (2007).

Pasquale, V., Massobrio, P., Bologna, L. L., Chiappalone, M. & Martinoia, S. Self-organization and neuronal avalanches in networks of dissociated cortical neurons. Neuroscience 153, 1354–1369 (2008).

Kitzbichler, M. G., Smith, M. L., Christensen, S. R. & Bullmore, E. Broadband criticality of human brain network synchronization. PLoS Comput. Biol. 5, e1000314 (2009).

Rabinovich, M., Huerta, R. & Laurent, G. Transient dynamics for neural processing. Science 321 48–50 (2008).

Tschacher, W. & Hake, H. Intentionality in non-equilibrium systems? The functional aspects of self-organised pattern formation. New Ideas Psychol. 25, 1–15 (2007).

Maturana, H. R. & Varela, F. De máquinas y seres vivos (Editorial Universitaria, Santiago, 1972). English translation available in Maturana, H. R. & Varela, F. in Autopoiesis and Cognition (Reidel, Dordrecht, 1980).

Fiorillo, C. D., Tobler, P. N. & Schultz, W. Discrete coding of reward probability and uncertainty by dopamine neurons. Science 299, 1898–1902 (2003).

Niv, Y., Duff, M. O. & Dayan, P. Dopamine, uncertainty and TD learning. Behav. Brain Funct. 1, 6 (2005).

Fletcher, P. C. & Frith, C. D. Perceiving is believing: a Bayesian approach to explaining the positive symptoms of schizophrenia. Nature Rev. Neurosci. 10, 48–58 (2009).

Phillips, W. A. & Silverstein, S. M. Convergence of biological and psychological perspectives on cognitive coordination in schizophrenia. Behav. Brain Sci. 26, 65–82 (2003).

Friston, K. & Kiebel, S. Cortical circuits for perceptual inference. Neural Netw. 22, 1093–1104 (2009).


Does signal reduction imply predictive coding in models of spoken word recognition?

Pervasive behavioral and neural evidence for predictive processing has led to claims that language processing depends upon predictive coding. Formally, predictive coding is a computational mechanism where only deviations from top-down expectations are passed between levels of representation. In many cognitive neuroscience studies, a reduction of signal for expected inputs is taken as being diagnostic of predictive coding. In the present work, we show that despite not explicitly implementing prediction, the TRACE model of speech perception exhibits this putative hallmark of predictive coding, with reductions in total lexical activation, total lexical feedback, and total phoneme activation when the input conforms to expectations. These findings may indicate that interactive activation is functionally equivalent or approximant to predictive coding or that caution is warranted in interpreting neural signal reduction as diagnostic of predictive coding.


Author Summary

Perception inevitably depends on combining sensory input with prior expectations. This is particularly critical for identifying degraded input. However, the underlying neural mechanism by which expectations influence sensory processing is unclear. Predictive Coding theories suggest that the brain passes forward the unexpected part of the sensory input while expected properties are suppressed (i.e., Prediction Error). However, evidence to rule out the opposite mechanism in which the expected part of the sensory input is enhanced or sharpened (i.e., Sharpening) has been lacking. In this study, we investigate the neural mechanisms by which sensory clarity and prior knowledge influence the perception of degraded speech. A univariate measure of brain activity obtained from functional magnetic resonance imaging (fMRI) is in line with both neural mechanisms (Prediction Error and Sharpening). However, combining multivariate fMRI measures with computational simulations allows us to determine the underlying mechanism. Our key finding was an interaction between sensory input and prior expectations: for unexpected speech, increasing speech clarity increases the amount of information represented in sensory brain areas. In contrast, for speech that matches prior expectations, increasing speech clarity reduces the amount of this information. Our observations are uniquely simulated by a model of speech perception that includes Prediction Errors.

Citation: Blank H, Davis MH (2016) Prediction Errors but Not Sharpened Signals Simulate Multivoxel fMRI Patterns during Speech Perception. PLoS Biol 14(11): e1002577. https://doi.org/10.1371/journal.pbio.1002577

Academic Editor: Robert Zatorre, McGill University, CANADA

Received: May 10, 2016 Accepted: October 19, 2016 Published: November 15, 2016

Copyright: © 2016 Blank, Davis. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Data Availability: The data is available at https://osf.io/2ze9n/ doi: 10.17605/OSF.IO/2ZE9N

Funding: This research was supported by UK Medical Research Council (MRC) funding of the Cognition and Brain Sciences Unit MC-A060-5PQ80 (to MHD). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Competing interests: The authors have declared that no competing interests exist.

Abbreviations: fMRI, functional magnetic resonance imaging IFG, inferior frontal gyrus RDM, representational dissimilarity matrix ROI, region of interest RSA, representational similarity analysis STG, superior temporal gyrus STS, superior temporal sulcus


The Prediction Error Account of Art – Current Formulation and its Contradiction

The basic tenet of Van de Cruys and Wagemans (2011) model is that a temporary state of unpredictability (or prediction error) is important for the emergence of perceptual pleasure vis-a-vis a work of art. Understanding perception in terms of predictions means that it is possible for perceptual configuration to induce different sequences of affect and to do so partly independently of the particular content of perception (TPEA, 1040). Accordingly, artists are supposed to intentionally create incongruities (perceived as prediction errors) that may not be possible in a natural visual environment, and viewers are able to tolerate and even enjoy the unpredictability because they expect to be surprised in their encounters with art (TPEA, 1041). By delaying prediction confirmation, artists create a positive affect: the viewer quickly runs into incongruities, which presumably generate an arousal aimed at reducing prediction errors. It is this incompatibility (or prediction error) that is the source of some of the emotionality of a work of art. In other words, artists intuitively attempt to strike the optimal balance between predictability and surprise. The mental effort required of a person in order to cope with the prediction error is a condition sine qua non for registering the perceptual pleasure of a Gestalt formation (prediction error reduction). According to Van de Cruys and Wagemans (2011): “Only by using minimal prediction errors painters can ensure that viewers will obtain their reward and not give up prematurely. Final gratification postponed as long as the artist has hidden in the painting enough micro reward the viewer can discover. ” (TPEA, 1050).

This concept is appealing for a number of reasons. First, it offers a much stronger explanatory framework than the neuroscientific and psychological models of art experience, which focus exclusively on a bottom–up account of visual processing (Zeki, 1999 Shimamura, 2013) and neglect or downplay the role of the top𠄽own, inferential activity of brain/mind. Second, although not specifically stated by the authors, the model responds to the so-called �rk room problem,” the apparent paradox that, in order to minimize surprise, agents should avoid sensory stimulation altogether and should proceed directly to the least stimulating environment and stay there they should take up a position in the nearest �rk room” and never move again. Neatly summarized: avoid surprises and you will last longer. Predictive coding theorists offer a simple solution to the dark room scenario: prior beliefs render dark rooms surprising. That is, agents that predict rich stimulating environments will find the �rk room” surprising and will leave at the earliest opportunity. The postulate of surprise minimization therefore by no means inhibits subjects from active, exploratory behavior and novelty-seeking, including presumably an aesthetic experience (Friston et al., 2012b Schwartenbeck et al., 2013).

Third, the model is apparently compatible with some well-established and respected art-historical theories, notably Gombrich’s (1960) theories of the prognostic character of the perception of pictures, the role of the beholder’s share and the viewer’s inferences in perception, and his notion of the artist working through a cycle of scheme and correction. Indeed, Gombrich’s (1960) famous maxim that “[t]o read the artist’s picture is to mobilize our memories and experience of the visible world and to test his image through tentative projections” leaves the door open to the Bayesian brain perspective an interesting challenge would be to recast Gombrich’s account in the explanatory terms of predictive coding, but this will not be pursued here. Furthermore, it provides scientific footing to some philosophical interpretations of aesthetic experience, most notably Gadamer’s (1975/2004) hermeneutical scenario, which highlights the activity of the perceiving subject vis-à-vis the aesthetic object. Gadamer describes the nature of this exchange as ongoing and dynamic, suggesting that understanding is an open-ended or at least an extended process that does not end the moment the representational content is identified or the information embedded in the work of art obtained, but also includes a more complex response and understanding: 𠇊ll encounter with the language of art is an encounter with an unfinished event and is itself part of this event. There is no absolute progress and no final exhaustion of what lies in work of art” (Gadamer, 1975/2004, p. 85).


Reinforcement learning framework

Reinforcement learning (RL) is perhaps the most influential framework developed to describe how an agent learns by interacting with its environment. RL is derived from the behaviorist view of animal behavior, in which an organism’s knowledge of the world is exclusively modeled based on its behavior. Crucially, RL theories focus on mechanistic accounts for behaviors based on several learning-related parameters established from empirical sources.

Both humans and nonhuman animals are excellent models for a variety of learning and decision-making tasks that are grounded on RL theories. Describing learning and learned outcomes through mathematical models is a powerful way to make explicit and testable predictions about how an organism will behave in a particular context and how they will make decisions that take into account internal states, such as motivation and subjective value. 27 The RL framework can capture seemingly complex behaviors with relatively simple yet elegant rules, as in the famous Rescorla–Wagner model. 28 Although various RL models differ in how they describe different cognitive phenomena, they share several core elements, such as the rate of learning or the salience of stimuli, to fit the specifics of learning and decision-making processes.

RL has its roots and applications in both engineering and psychology. RL has its core foundations in the work of Richard Bellman, most famous for developing the Bellman optimality equation and dynamic programming. The more widely appreciated root of RL is conceptualizing how organisms gather information from their environment to learn and make decisions. RL requires an agent that moves through different states, or contexts, in a given environment. Other necessary components include a reward signal, a value function, and a policy. Reward outcome is central to all forms of RL and consists of a quantity the agent gets as a result of its actions within the environment. The agent then computes a value function using that reward outcome that calculates the expected value of certain states/contexts as well as the conjunction of specific states and actions. The agent uses these value functions to develop a set of preferred actions, known as a policy. A model of the environment is an optional component of RL that can provide the organism with guidance on how to move from state to state.

In dynamic programming, developed by Bellman for engineering applications, a complete model of the environment is required. This idea requires the action of an agent to be guided by the expected payoff of the action in addition to the total expected payoff of potential actions in hypothetical future states. 29 The same principle applies to temporal discounting (TD) models, the predominant form of RL model applied in psychological studies of humans and other animals. 30 TD learning notably differs from dynamic programming, as it does not require any model of the environment. Instead, learning is accomplished by comparing expected reward to actual reward after a certain transition in time. This difference is the reward prediction error. This prediction error is used to update the value function and, ultimately, the policy of an agent interacting with its environment. Prediction error signaling is indeed the fundamental attribute of the original models of learning. 28 In simple terms, a prediction error calculates the difference between what the animal expects to have happen and what actually happens to the animal on a given event or trial. 31 This can also be described as an error signal. 32


Why visual perception is a decision process

Summary: Prediction errors play a role in the context of dynamic perceptual events that take place within fractions of a second. Findings support the hypothesis that visual perception occurs as a result of a decision process.

Source: RUB

Neuroscientists at the Ruhr-Universität Bochum (RUB) together with colleagues at the Freiburg University show that this is not strictly the case. Instead, they show that prediction errors can occasionally appear as visual illusion when viewing rapid image sequences. Thus, rather than being explained away prediction errors remain in fact accessible at final processing stages forming perception. Previous theories of predictive coding need therefore to be revised. The study is reported in Plos One on 4. May 2020.

Our visual system starts making predictions within a few milliseconds

To fixate objects in the outside world, our eyes perform far more than one hundred thousand of rapid movements per day called saccades. However, as soon as our eyes rest about 100 milliseconds, the brain starts making predictions. Differences between previous and current image contents are then forwarded to subsequent processing stages as prediction errors. The advantage to deal with differences instead of complete image information is obvious: similar to video compression techniques the data volume is drastically reduced. Another advantage turns up literally only at second sight: statistically, there is a high probability that the next saccade lands on locations where differences to previous image contents are largest. Thus, calculating potential changes of image content as the differences to previous contents prepares the visual system early on for new input.

To test whether the brain uses indeed such a strategy, the authors presented rapid sequences of two images to human volunteers. In the first image two gratings were superimposed, in the second image only one of the gratings was present. The task was to report the orientation of the last seen single grating. In most cases, the participants correctly reported the orientation of the present orientation, as expected. However, surprisingly, in some cases an orientation was perceived that was exactly orthogonal to the present orientation. That is, participants saw sometimes the difference between the previous superimposed gratings and the present single grating. „Seeing the difference instead of the real current input is here a visual illusion that can be interpreted as directly seeing the prediction error,” says Robert Staadt from the Institute of Neural Computation of the RUB, first author of the study.

Avoiding the pigeonhole benefits flexibility

“Within the framework of the predictive coding theory, prediction errors are mostly conceived in the context of higher cognitive functions that are coupled to conscious expectations. However, we demonstrate that prediction errors also play a role in the context of highly dynamic perceptual events that take place within fractions of a second,” explains Dr. Dirk Jancke, head of the Optical Imaging Group at the Institute of Neural Computation. The present study reveals that the visual system simultaneously keeps up information about past, current, and possible future image contents. Such a strategy allows both stability and flexibility when viewing rapid image sequences. “Altogether, our results support hypotheses that consider perception as a result of a decision process,” says Jancke. Hence, prediction errors should not be sorted out too early, as they might become relevant for following events.

Besides straightforward physical parameters like stimulus duration, brightness, and contrast, other, more elusive factors that characterize psychological features might be involved. Image is in the public domain.

Visual perception underlies decision making

In next studies the scientists will scrutinize the sets of parameters that drive the perceptual illusion most effectively. Besides straightforward physical parameters like stimulus duration, brightness, and contrast, other, more elusive factors that characterize psychological features might be involved. The authors’ long-term perspective is the development of practical visual tests that can be used for an early diagnosis of cognitive disorders connected to rapid perceptual decision processes.

Funding: The study was partly financed through grants of the Collaborative Research Centre (CRC) 874 at RUB, which is supported by the German Research Foundation since 2010. The CRC “Integration and representation of sensory processes” investigates how sensory signals generate neuronal maps, and result in complex behavior and memory formation.


10 Jul 2019: Smout CA, Tang MF, Garrido MI, Mattingley JB (2019) Correction: Attention promotes the neural encoding of prediction errors. PLOS Biology 17(7): e3000368. https://doi.org/10.1371/journal.pbio.3000368 View correction

The encoding of sensory information in the human brain is thought to be optimised by two principal processes: ‘prediction’ uses stored information to guide the interpretation of forthcoming sensory events, and ‘attention’ prioritizes these events according to their behavioural relevance. Despite the ubiquitous contributions of attention and prediction to various aspects of perception and cognition, it remains unknown how they interact to modulate information processing in the brain. A recent extension of predictive coding theory suggests that attention optimises the expected precision of predictions by modulating the synaptic gain of prediction error units. Because prediction errors code for the difference between predictions and sensory signals, this model would suggest that attention increases the selectivity for mismatch information in the neural response to a surprising stimulus. Alternative predictive coding models propose that attention increases the activity of prediction (or ‘representation’) neurons and would therefore suggest that attention and prediction synergistically modulate selectivity for ‘feature information’ in the brain. Here, we applied forward encoding models to neural activity recorded via electroencephalography (EEG) as human observers performed a simple visual task to test for the effect of attention on both mismatch and feature information in the neural response to surprising stimuli. Participants attended or ignored a periodic stream of gratings, the orientations of which could be either predictable, surprising, or unpredictable. We found that surprising stimuli evoked neural responses that were encoded according to the difference between predicted and observed stimulus features, and that attention facilitated the encoding of this type of information in the brain. These findings advance our understanding of how attention and prediction modulate information processing in the brain, as well as support the theory that attention optimises precision expectations during hierarchical inference by increasing the gain of prediction errors.


An alternative to the two-factor model of delusions

Fletcher & Frith ( Reference Fletcher and Frith 2009) have argued that a two-stage theory of psychosis in schizophrenia is unnecessary, as both perception and belief rely on predictions and updating those predictions in the light of evidence (so-called prediction error). Frith has shown the importance in health of being able to monitor one's own thoughts and actions subconsciously, and of being able to predict the consequences of those actions using the so-called efference-copy or corollary discharge mechanism. He and others have shown that patients with schizophrenia do indeed have problems in predictive coding and monitoring of their actions (Feinberg, Reference Feinberg 1978 McGuire et al. Reference McGuire, Silbersweig, Wright, Murray, David, Frackowiak and Frith 1995 Shergill et al. Reference Shergill, Samson, Bays, Frith and Wolpert 2005 Mathalon & Ford, Reference Mathalon and Ford 2008). By linking this theory of abnormal predictive coding in psychosis to evidence that dopamine-driven prediction error signalling is also abnormal in psychosis, Fletcher and Frith argue that abnormal predictive coding at multiple levels of brain complexity could lead to both false perceptions and false beliefs in schizophrenia. This provides a unitary psychological process that could underlie all positive psychotic symptoms. This simplicity and parsimony is a very attractive aspect of the theory a disadvantage is that this parsimony may not be reflected at a biological level. For example, it is unlikely that a single neurotransmitter system could be responsible for predictive coding throughout the brain. Whilst predictive coding for rewards may be intricately linked to dopamine, it is unlikely that predictive coding for visual perception will have the same biological basis. Empirical work combining pharmacological, imaging and psychological studies is required to test this theory.


We thank Scott Brincat for assistance with surgeries and data preprocessing and Morteza Moazami and Jefferson Roy for assistance with surgeries and animal training. We also thank the MIT veterinary staff and animal caretakers for their excellent support. We also thank Jaan Aru and Bruno Gomes for comments on the manuscript and Miles Whittington for many useful conversations. This work was supported by National Institutes of Mental Health Grant R37MH087027 and 5K99MH116100-02, Office of Naval Research Multidisciplinary University Research Initiatives Grant N00014-16-1-2832, and the MIT Picower Institute Faculty Innovation Fund.

↵ 2 Present address: Department of Psychology and Vanderbilt Brain Institute, Vanderbilt University, Nashville, TN 37240.


Watch the video: ROC and AUC, Clearly Explained! (June 2022).


Comments:

  1. Adley

    I am assured, what is it - a false way.

  2. Ottah

    Thanks for the kind society.

  3. Harel

    Delightfully

  4. Arridano

    Talently...

  5. Thompson

    Incomparable topic, it is very interesting to me))))



Write a message