Is My Brain Really Predicting the Next Moment?

Have you ever asked yourself, while reading about predictive processing, whether it can really be true that your brain is running predictions of the next moment right now, as you read this sentence? It is a fair question, and one of the pleasures of the field is that you can run small experiments on yourself and catch the prediction machinery in the act.

The short answer is yes. The nervous system is predicting continuously and mostly below awareness. Some of those predictions are large — what your body will do next, what someone will say next. Most are tiny — what the next syllable of a sentence will sound like, where the next foot will land, what the coffee cup will feel like when you pick it up (Clark, 2013) [Supported Finding]. The predictions run whether or not you attend to them.

A few small experiments make the machinery visible.

Read this sentence: The chef prepared a delicious ______. Whatever word appeared in your mind just now — meal, dish, sauce — is a prediction the language system generated before you consciously chose to generate one. Your nervous system was already narrowing the space of possibilities. If the actual next word had been "algorithm," you would have felt the prediction error as a small mental hiccup. That hiccup is what prediction error feels like at low intensity.

Another small demonstration: pick up a familiar object — a cup, a phone — and pay attention to what you expected the weight to feel like before you lifted it. You almost certainly had a prediction. Most of the time it is correct and you notice nothing. The moment you notice is the moment the prediction was wrong. If you have ever reached for what looked like a full cup and found it empty, the shock in your arm was your motor system's prediction error arriving with full physiological force.

A slightly more subtle experiment: close your eyes and imagine a room you know well. Notice how detailed the imagined room is even though no sensory data is arriving. That richness is the same predictive model that ordinarily runs against real sensory input, running without the correction. Under normal conditions, the model and the correction combine to produce what feels like seeing. Alone, the model produces what feels like remembering.

What these small experiments suggest is that prediction is not an exotic laboratory finding. It is what most of ordinary experience is quietly made of. Once you know to look for it, you can catch small predictions all day — the sound the door was going to make, the tone of voice you expected from a colleague, the second bite of a familiar meal that tasted different because your prior was slightly off.

This is one of the more useful things to take from the field. Perception feels like receiving. It is actually mostly predicting, corrected. Learning to notice the prediction in real time is one of the small durable practices that quietly loosens the grip of any specific prediction on any specific moment.

For the underlying model, see The Brain Is a Prediction Machine. For where the specific predictions come from, Priors — The Brain's Working Assumptions is the companion piece.