Active inference, sensory attenuation and illusions - PubMed Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
Review
. 2013 Nov;14(4):411-27.
doi: 10.1007/s10339-013-0571-3. Epub 2013 Jun 7.

Active inference, sensory attenuation and illusions

Harriet Brown  1 , Rick A Adams, Isabel Parees, Mark Edwards, Karl Friston
Affiliations
Review

Active inference, sensory attenuation and illusions

Harriet Brown et al. Cogn Process. 2013 Nov.

Abstract

Active inference provides a simple and neurobiologically plausible account of how action and perception are coupled in producing (Bayes) optimal behaviour. This can be seen most easily as minimising prediction error: we can either change our predictions to explain sensory input through perception. Alternatively, we can actively change sensory input to fulfil our predictions. In active inference, this action is mediated by classical reflex arcs that minimise proprioceptive prediction error created by descending proprioceptive predictions. However, this creates a conflict between action and perception; in that, self-generated movements require predictions to override the sensory evidence that one is not actually moving. However, ignoring sensory evidence means that externally generated sensations will not be perceived. Conversely, attending to (proprioceptive and somatosensory) sensations enables the detection of externally generated events but precludes generation of actions. This conflict can be resolved by attenuating the precision of sensory evidence during movement or, equivalently, attending away from the consequences of self-made acts. We propose that this Bayes optimal withdrawal of precise sensory evidence during movement is the cause of psychophysical sensory attenuation. Furthermore, it explains the force-matching illusion and reproduces empirical results almost exactly. Finally, if attenuation is removed, the force-matching illusion disappears and false (delusional) inferences about agency emerge. This is important, given the negative correlation between sensory attenuation and delusional beliefs in normal subjects--and the reduction in the magnitude of the illusion in schizophrenia. Active inference therefore links the neuromodulatory optimisation of precision to sensory attenuation and illusory phenomena during the attribution of agency in normal subjects. It also provides a functional account of deficits in syndromes characterised by false inference and impaired movement--like schizophrenia and Parkinsonism--syndromes that implicate abnormal modulatory neurotransmission.

PubMed Disclaimer

Figures

Fig. 1
Fig. 1
Generative model: This figure shows the generative process and model used in these simulations. The generative process (left) models real-world states and causes, while the generative model (right) is used by the subject to make inferences about causes of its sensations. In the real world, the hidden state x i models self-generated forces that are sensed by both somatosensory s s and proprioceptive s p input channels. External forces are modelled with the hidden cause ν e and are sensed only by the somatosensory input channel. Action causes the self-generated force to increase and is modified by a sigmoid squashing function σ (a hyperbolic tangent function). The hidden state decays slowly over four time bins. In the generative model, causes of sensory data are divided into internal causes ν i and external causes ν e. The hidden cause excites dynamics in hidden states x i and x e which decay slowly over time as above. Internal force is perceived by both proprioceptive and somatosensory receptors, while external force is perceived only by somatosensory receptors. Crucially, the precision of the sensory prediction error π is influenced by the level of internal force, again modulated by a squashing function, and controlled by a parameter γ which governs the level of attenuation of precision. The pink circles highlight this state-dependent precision, which effectively controls the influence of sensory prediction errors during active inference
Fig. 2
Fig. 2
Functional anatomy: Speculative mapping of Eq. (3) onto neuroanatomy. Somatosensory and proprioceptive prediction errors are generated by the thalamus, while conditional expectations and prediction errors about hidden states (circles) (the forces) are placed in sensorimotor cortex. The expectations and prediction errors about the hidden causes of forces (triangles) have been placed in the prefrontal cortex. In active inference, proprioceptive predictions descend to the spinal cord and elicit output from alpha motor neurons (playing the role of proprioceptive prediction error units) via a classical reflex arc. Red connections originate from prediction error units (ξ cells) and can be regarded as intrinsic connections or ascending (forward) extrinsic connections from superficial principal cells. Conversely, the black connections represent intrinsic connections and descending (backward) efferents from (deep) principal cells encoding conditional expectations (μ∼ cells). The cyan connections denote descending neuromodulatory effects that mediate sensory attenuation. The crucial point to take from this schematic is that conditional expectations of sensory states (encoded in the pyramidal cell μ∼x) can either be fulfilled by descending proprioceptive predictions (that recruit classical reflex arcs), or they can be corrected by ascending sensory prediction errors. In order for descending motor efferents to prevail, the precision of the sensory prediction errors must be attenuated
Fig. 3
Fig. 3
Sensory attenuation and action: simulation results illustrating the permissive effect of sensory attenuation in movement. The model was supplied with a prior belief about the hidden cause of internally generated movement, while sensory attention was high (γ=6). This prior expectation was a simple Gaussian function of time (blue line in the lower left panel) and engenders beliefs about forces (upper right panel), which produce proprioceptive predictions (upper left panel). Action is enslaved to fulfil these predictions (lower right panel). Note the confidence interval around the external cause temporarily inflates during action (lower left panel), reflecting the attenuation of sensory precision
Fig. 4
Fig. 4
A simulation of akinesia: This figure uses the same format as previous figure but reports the results of simulations when sensory attenuation is much lower (γ = 2). In this case, bottom-up prediction errors retain a higher precision than descending predictions during movement. Conditional expectations that are updated by ascending prediction errors (upper right panel) overwhelm prediction errors based upon top-down predictions, and consequently infer that there is no change in the state of the world. This means that proprioceptive prediction errors are not produced (upper left panel) and action is profoundly suppressed (lower right panel)
Fig. 5
Fig. 5
Movement and precision: True internally generated force x i and perceived internally generated force (lower 90 % confidence interval of x i) simulated over a range of sensory attenuations, where γ={6,…−4}. Confident movement gradually emerges as the prior precision increases in relation to sensory precision, with movement being around half its maximum amplitude when prior and sensory precisions are balanced (γ = 2, vertical line). Force on the y axis is measured in arbitrary units
Fig. 6
Fig. 6
Simulation of the force-matching task. In the first part of this simulation (left hand panels), an internal force is generated (from a prior belief about the hidden cause ν i), followed by the presentation of an external force. The estimates of the hidden states (upper right panel) are similar, but the confidence interval around the force for the internally generated state is much broader. If perceptual inference is associated with the lower 90 % confidence bound of the estimate of the hidden state, it will be lower when the force is self-generated (double-headed arrow, upper right panel). This is demonstrated in the right-hand panels. This is a simulation the force-matching paradigm where the external force is matched to the lower bound of the 90 % confidence interval of the internal force. This means that internally generated force is now greater than the externally applied force (double-headed arrow, upper left panel)
Fig. 7
Fig. 7
Sensory attenuation in schizophrenia: Left panel results of the force-matching simulation repeated under different levels of self-generated force. For normal levels of sensory attenuation (blue circles), internally produced force is higher than externally generated force at all levels of force, consistent with published data. Force-matching typical of schizophrenia (red circles) was simulated by reducing sensory attenuation and increasing the precision of prediction errors at higher levels of the hierarchy. This resulted in a more veridical perception of internally generated force (small circles). Right panel empirical results using the same format adapted (with permission) from (Shergill et al. 2003, 2005)
Fig. 8
Fig. 8
Pathology of sensory attenuation. To simulate the force-matching results seen in schizophrenia, sensory attenuation was reduced and precision at non-sensory levels of the hierarchy increased to allow movement. This results in a precise and accurate perception of internally and externally generated sensations (upper left panel). However, the causes of sensory data are not accurately inferred: an illusory cause (circled response in the lower left panel) is perceived during internally generated movement that is antagonistic to the movement. This is because the proprioceptive prediction errors driving action are rendered overly precise, meaning higher levels of the hierarchy must be harnessed to explain them, resulting in a ‘delusion’
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

See all "Cited by" articles

References

    1. Abrams J, Barbot A, Carrasco M. Voluntary attention increases perceived spatial frequency. Atten Percept Psychophys. 2010;72:1510–1521. doi: 10.3758/APP.72.6.1510. - DOI - PMC - PubMed
    1. Adams RA, Shipp S, Friston KJ. (2012). Predictions not commands: active inference in the motor system. Brain Struct Funct. doi:10.1007/s00429-012-0475-5 - PMC - PubMed
    1. Aliu SO, Houde JF, Nagarajan SS. Motor-induced suppression of the auditory cortex. J Cogn Neurosci. 2009;2009(21):791–802. doi: 10.1162/jocn.2009.21055. - DOI - PMC - PubMed
    1. Baess P, Jacobsen T, Schröger E. Suppression of the auditory N1 event-related potential component with unpredictable self-initiated tones: evidence for internal forward models with dynamic stimulation. Int J Psychophysiol. 2008;70:137–143. doi: 10.1016/j.ijpsycho.2008.06.005. - DOI - PubMed
    1. Baess P, Widmann A, Roye A, Schröger E, Jacobsen T. Attenuated human auditory middle latency response and evoked 40-Hz response to self-initiated sounds. Eur J Neurosci. 2009;29:1514–1521. doi: 10.1111/j.1460-9568.2009.06683.x. - DOI - PubMed

Publication types

LinkOut - more resources