Trends in Neurosciences
Volume 27, Issue 12, December 2004, Pages 712-719
Journal home page for Trends in Neurosciences

The Bayesian brain: the role of uncertainty in neural coding and computation

https://doi.org/10.1016/j.tins.2004.10.007Get rights and content

To use sensory information efficiently to make judgments and guide action in the world, the brain must represent and use information about uncertainty in its computations for perception and action. Bayesian methods have proven successful in building computational theories for perception and sensorimotor control, and psychophysics is providing a growing body of evidence that human perceptual computations are ‘Bayes' optimal’. This leads to the ‘Bayesian coding hypothesis’: that the brain represents sensory information probabilistically, in the form of probability distributions. Several computational schemes have recently been proposed for how this might be achieved in populations of neurons. Neurophysiological data on the hypothesis, however, is almost non-existent. A major challenge for neuroscientists is to test these ideas experimentally, and so determine whether and how neurons code information about sensory uncertainty.

Section snippets

Bayesian inference and the Bayesian coding hypothesis

The fundamental concept behind the Bayesian approach to perceptual computations is that the information provided by a set of sensory data about the world is represented by a conditional probability density function over the set of unknown variables – the posterior density function. A Bayesian perceptual system, therefore, would represent the perceived depth of an object, for example, not as a single number Z but as a conditional probability density function p(Z/I), where I is the available

Are human observers Bayes' optimal?

What does it mean to say that an observer is ‘Bayes' optimal’? Humans are clearly not optimal in the sense that they achieve the level of performance afforded by the uncertainty in the physical stimulus. Absolute efficiencies (a measure of performance relative to a Bayes' optimal observer) for performing high-level perceptual tasks are generally low and vary widely across tasks. In some cases, this inefficiency is entirely due to uncertainty in the coding of sensory primitives that serve as

Neural representations of uncertainty

The notion that neural computations take into account the uncertainty of the sensory and motor variables raises two important questions: (i) how do neurons, or rather populations of neurons, represent uncertainty, and (ii) what is the neural basis of statistical inferences? Several schemes have been proposed over the past few years, which we now briefly review.

Discussion

We have described psychophysical evidence that shows human observers to behave in a variety of ways like optimal Bayesian observers. The most compelling features of these data in regard to the Bayesian coding hypothesis are: (i) that subjects implicitly ‘adjust’ cue weights in a Bayes' optimal way based on stimulus and viewing parameters; (ii) that perceptual and motor behavior reflect a system that takes into account the uncertainty of both sensory and motor signals; (iii) that humans behave

References (58)

  • D.C. Knill

    Mixture models and the probabilistic structure of depth cues

    Vision Res.

    (2003)
  • J.I. Gold et al.

    Neural computations that underlie decisions about sensory stimuli

    Trends Cogn. Sci.

    (2001)
  • D. Tolhurst

    The statistical reliability of signals in single neurons in cat and monkey visual cortex

    Vision Res.

    (1983)
  • P.E. Latham

    Optimal computation with attractor networks

    J. Physiol. Paris

    (2003)
  • E. Mach

    Contributions to the Analysis of the Sensations (C. M. Williams, Trans.)

    (1980)
  • Helmholtz, H. (1925) Physiological Optics, Vol. III: The perceptions of Vision (J. P. Southall, Trans.), Optical...
  • D.M. Wolpert

    An internal model for sensorimotor integration

    Science

    (1995)
  • C.M. Harris et al.

    Signal-dependent noise determines motor planning

    Nature

    (1998)
  • R.J. van Beers

    Sensorimotor integration compensates for visual localization errors during smooth pursuit eye movements

    J. Neurophysiol.

    (2001)
  • R.J. van Beers

    Role of uncertainty in sensorimotor control

    Philos. Trans. R. Soc. Lond. B Biol. Sci.

    (2002)
  • S.S. Shimozaki

    An ideal observer with channels versus feature-independent processing of spatial frequency and orientation in visual search performance

    J. Opt. Soc. Am. A Opt. Image Sci. Vis.

    (2003)
  • B.R. Beutter

    Saccadic and perceptual performance in visual search tasks. I. Contrast detection and discrimination

    J. Opt. Soc. Am. A Opt. Image Sci. Vis.

    (2003)
  • R.F. Murray

    Saccadic and perceptual. performance in visual search tasks. II. Letter discrimination

    J. Opt. Soc. Am. A Opt. Image Sci. Vis.

    (2003)
  • J.A. Saunders et al.

    Visual feedback control of hand movements

    J. Neurosci.

    (2004)
  • Y. Weiss

    Motion illusions as optimal percepts

    Nat. Neurosci.

    (2002)
  • R. van Ee

    Bayesian modeling of cue interaction: bistability in stereoscopic slant perception

    J. Opt. Soc. Am. A Opt. Image Sci. Vis.

    (2003)
  • J. Trommershauser

    Statistical decision theory and trade-offs in the control of motor response

    Spat. Vis.

    (2003)
  • J. Trommershauser

    Statistical decision theory and the selection of rapid, goal-directed movements

    J. Opt. Soc. Am. A Opt. Image Sci. Vis.

    (2003)
  • J. Pearl

    Probabilistic Reasoning in Intelligent Systems: Networks of Plausible Inference

    (1988)
  • Cited by (1713)

    • The Human Affectome

      2024, Neuroscience and Biobehavioral Reviews
    View all citing articles on Scopus
    View full text