Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

The DOI matches its Crossref record; 102 of the 102 references it cites were found in a registry or an index; 10 of the 63 sources its claims cite were checked for support.

Action and behavior: a free-energy formulation

Open access
Friston, Karl J., Daunizeau, Jean, Kilner, James, Kiebel, Stefan J. · 2010 · Biological Cybernetics

Abstract

We have previously tried to explain perceptual inference and learning under a free-energy principle that pursues Helmholtz's agenda to understand the brain in terms of energy minimization. It is fairly easy to show that making inferences about the causes of sensory data can be cast as the minimization of a free-energy bound on the likelihood of sensory inputs, given an internal model of how they were caused. In this article, we consider what would happen if the data themselves were sampled to minimize this bound. It transpires that the ensuing active sampling or inference is mandated by ergodic arguments based on the very existence of adaptive agents. Furthermore, it accounts for many aspects of motor behavior; from retinal stabilization to goal-seeking. In particular, it suggests that motor control can be understood as fulfilling prior expectations about proprioceptive sensations. This formulation can explain why adaptive behavior emerges in biological agents and suggests a simple alternative to optimal control theory. We illustrate these points using simulations of oculomotor control and then apply to same principles to cued and goal-directed movements. In short, the free-energy formulation may provide an alternative perspective on the motor control that places it in an intimate relationship with perception.

Overview

CoChat Summary

## TL;DR

This paper argues that action, like perception, can be understood as minimizing free energy: agents do not select actions by reward maximization, but by fulfilling prior expectations about the sensory consequences of movement. Through simple simulations, it shows that this framework can produce retinal stabilization, intended movements, multimodal reaching, goal-directed behavior, and learned autonomous trajectories.

## Key findings

  • Action can be cast as the suppression of sensory prediction error: the paper states that “the only thing that action can affect is the prediction error at the sensory level,” and that movement is driven by sensory prediction errors rather than explicit motor commands.
  • In oculomotor simulations, tight priors on perturbations led to action that “countered almost immediately” the stimulus motion, while flat priors produced perceived motion; the paper reports a 6×6 = 36-channel sensory array in one retinal simulation.
  • In the cued reaching simulations, behavior was robust to sensory noise: proprioceptive noise alone could be compensated by vision, visual noise alone by proprioception, but when both modalities were noisy, reaching was compromised.
  • In the goal-directed “mountain-car” simulation, the agent learned the controlled dynamics over 16 trials, and after learning it produced the required circuitous trajectory to the target; the paper notes that after 16 trials the parameters “converged roughly” to the control environment.
  • In the autonomous-behavior simulation, after 16 trials of 512 time-bins, the agent learned to expect a Lorenz attractor and then continued to follow that learned trajectory even under perturbations, showing robustness to exogenous forces.

## Methods

  • The paper is a theoretical and simulation study framed around the free-energy principle and active inference, with action and perception implemented through gradient descent on free energy under Gaussian/Laplace assumptions.
  • It uses several toy dynamical systems: retinal stabilization/oculomotor control, two-joint arm reaching, the mountain-car problem, and a Lorenz-attractor learning task; one learning example uses 16 trials of 32 s time-bins, and another uses 16 trials of 512 time-bins.
  • Analysis is based on generalized coordinates of motion, prediction errors, and variational (free-energy) updates of conditional expectations and precisions; the paper reports that all simulations were implemented with a single Matlab routine, `spm_ADEM.m`.

## Limitations

  • The paper itself states that the simulations are simple and that many motor-control and sensorimotor issues are only addressed “superficially,” with the examples used mainly to “impart the intuitions behind active inference.”
  • The text repeatedly says that some claims rely on simplifying assumptions, including Gaussian/Laplace approximations, generalized-state truncation, and assumptions about diffeomorphic sensory mappings (stated in the paper).
  • Evident from the text: the results are based on idealized simulations and toy environments rather than empirical experiments on biological agents or neural recordings.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1007/s00422-010-0364-z

Cites

102 of 102 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Verification

Claims the paper makes

10 Valid · 3 Not valid · 32 Not checked · 265 Own results and interpretations

181 more claims

Create a free account to see all 310 claims and every source verdict.

See all claims — free account
  • Probably not a substantial claim. 67
  • It may state two separate claims as one. 67
  • The quote only partly states this claim. 2
  • Borderline: one of the tests for a substantial claim is unclear. 1

Create a free account to see these claims.

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

3 more claims about this paper

Create a free account to see every claim other papers make about this one.

See all claims — free account

Citations

Cites

Show all 100 references

Research works printed but not linked: 2

  1. MacKay DJC (1995) Free-energy minimization algorithm for decoding and cryptoanalysis. Electron Lett 31: 445–447
  2. Watkins CJCH, Dayan P (1992) Q-learning. Mach Learn 8: 279–292

Cited by

See all 866

Metadata

Authors
Friston, K. J., Daunizeau, J., Kilner, J., & Kiebel, S. J.
Affiliations
Wellcome Centre for Human Neuroimaging (GB)
Corresponding author
Friston, Karl J.
Venue
Biological Cybernetics
Published
2010-02-11
Volume
102
Issue
3
Pages
227-260
ISSN
  • 0340-1200 (ISSN-L)
  • 1432-0770
OpenAlex
W2130195834
Cited by
866
References
100
Citations by year
2026: 32 · 2025: 65 · 2024: 64 · 2023: 77 · 2022: 63 · 2021: 79 · 2020: 67 · 2019: 56 · 2018: 67 · 2017: 51 · 2016: 36 · 2015: 39 · 2014: 34 · 2013: 33 · 2012: 29
FWCI
18.62
Type
journal article
Language
en
Open access
Licence
cc-by-nc
Locations
  • Biological Cybernetics (publishedVersion, cc-by-nc)
  • PubMed (publishedVersion)
  • UCL Discovery (University College London) (submittedVersion)
  • Max Planck Digital Library (submittedVersion)
  • MPG.PuRe (Max Planck Society) (submittedVersion)
Topics
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Embodied and Extended Cognition
  • Social Sciences › Psychology › Social Psychology › Action Observation and Synchronization
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Motor Control and Adaptation
Keywords
free energy principle, active sampling, perceptual inference, motor control, oculomotor control, goal-directed movement, proprioception, perceptual learning
MeSH terms
Afferent Pathways, Animals, Bayes Theorem, Behavior, Brain, Computer Simulation, Humans, Learning, Models, Neurological

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Action and behavior: a free-energy formulation. https://app.cochat.ai/papers/10.1007/s00422-010-0364-z
The paper
Friston, K. J., Daunizeau, J., Kilner, J., & Kiebel, S. J. (2010). Action and behavior: a free-energy formulation. Biological Cybernetics. https://doi.org/10.1007/s00422-010-0364-z

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.