Predictive Processing and Praxis: Rethinking Sensory Processing
This blog is part of a series exploring the implications of predictive processing for clinical practice.
Several people have asked us how predictive processing relates to praxis. Before discussing praxis specifically, it is helpful to understand that predictive processing proposes a fundamentally different model of how the brain processes sensory information. This blog compares the traditional detector model of sensory processing with the predictive processing model and considers what this means for occupational therapy interventions.
Two Models of Sensory Processing
There are two broad ways of understanding how the brain processes sensory information:
The classic detector model of sensory processing – the traditional model that has informed sensory processing theory and much of clinical practice for many years.
The predictive processing model of sensory processing – the leading model in contemporary neuroscience, although it is less familiar in many clinical settings.
Importantly, predictive processing does not suggest that previous interventions are ineffective. Rather, it provides a different explanation for why they work and may lead to a different emphasis in clinical reasoning.
The Classic Detector Model of Sensory Processing
In the classic detector model, sensory information from the environment (bottom-up input) is detected by the sensory systems, processed, integrated with other sensory information and then used to generate an appropriate motor response. Top-down influences such as learning, expectations and automatic motor programmes are recognised, but sensory processing is generally viewed as beginning with incoming sensory signals.
Some sensory input becomes habituated over time and therefore requires less conscious attention. Through repeated practice, movements become increasingly automatic. For example, when swinging a cricket bat or tennis racquet, the eyes detect the approaching ball, proprioceptive information provides information about body position, this information is integrated, a motor plan is generated, and the movement is executed. With experience, this sequence becomes increasingly automatic.
The Predictive Processing Model of Sensory Processing
Predictive processing proposes almost the opposite sequence.
Rather than waiting for sensory information before deciding what to do, the brain is continuously generating predictions about the body, the environment and the sensory information it expects to receive. These predictions are generated automatically and outside conscious awareness.
Incoming sensory information is then compared with these predictions.
When the sensory information matches what was predicted, very little prediction error is generated. As a result, there is little need to update the brain's internal model and the activity proceeds smoothly and automatically.
When sensory information differs from what was predicted, a prediction error occurs. Despite its name, a prediction error does not mean the brain has made a mistake or that the prediction was incorrect. It simply means that the incoming sensory information differs from what the brain expected.
The brain must then determine whether the prediction should be updated or whether the unexpected sensory information should be treated as unimportant or "noise". This process is known as precision weighting, where the brain estimates how reliable both the incoming sensory information and its existing prediction are.
From this perspective, the goal of the nervous system is not to detect sensory information, but to minimise prediction errors while maintaining an accurate model of the world.
An Example: Swinging a Bat
Imagine swinging a bat to hit a ball.
Before the movement even begins, the brain has already generated predictions about the position of the body, the weight of the bat, the expected movement, and the sensory consequences of the swing.
If events unfold exactly as predicted, very little prediction error is generated and the movement feels smooth, coordinated and automatic.
However, if the ball unexpectedly changes direction, prediction errors increase. The brain rapidly updates its predictions and the body adjusts the movement to respond to the new information.
Rather than first detecting sensory information and then deciding what to do, predictive processing suggests that movement itself is often generated to fulfil the brain's predictions, with sensory information continually checking whether those predictions remain accurate.
Walking Down Stairs
A similar process occurs when walking down a familiar staircase.
The brain already has well-established predictions about the height of each step, the force required and the coordinated body movements needed to descend safely. As long as the stairs behave as expected, very little prediction error is generated and walking feels effortless.
If one step is unexpectedly loose or unstable, the sensory information no longer matches the brain's prediction. Prediction errors increase, posture and movement are adjusted almost instantly, and the brain updates its model so future predictions become more accurate.
What Does This Mean for Praxis?
This difference in how movement is understood has important implications for praxis.
Traditionally, praxis has been viewed as the ability to formulate and execute motor plans based on sensory information.
From a predictive processing perspective, praxis can instead be viewed as the ability to generate accurate predictions about intended actions and their expected sensory consequences, while continuously refining those predictions through prediction errors.
Difficulties with praxis may therefore arise from challenges in forming stable predictions about actions, weighting incoming sensory information, or effectively updating predictions from experience.
This offers an understanding of why some individuals find it difficult to acquire new motor skills or to generalise movements across different contexts.
Implications for Sensory Processing Differences
One influential predictive processing account proposes that autistic people experience more prediction errors because the brain gives greater weight to incoming sensory information and less weight to prior expectations. This means that events may feel less predictable and that experiences which neurotypical individuals readily generalise may instead be processed as more unique or novel.
As a result, more sensory information requires active processing, increasing cognitive and sensory load. When prediction errors accumulate faster than they can be resolved, this may contribute to sensory overload.
It is important to note that this is one of several current predictive processing accounts of autism and continues to be an active area of research.
Implications for Occupational Therapy Interventions
One of the key messages of predictive processing is that interventions which currently work are likely to continue working. What changes is our understanding of why they work.
For Autistic children and by extension other people who also have sensory processing difficulties, there is an increased number of ‘prediction errors’ due to not finding the actions predictable and not generalising the actions. Eg seeing that some actions are more unique than they are. This increases the amount/load of sensory input which can then lead to sensory overload if there is too much of it. It may also be useful to practice body movements in the context that they are being used repetitively so that the brain can learn the prediction.