With this I mean the notion of humans being able to, for example, look at a painting and tell that something doesn't belong in there. For example sun glasses on the Mona Lisa, without prior knowledge of the painting. I just find it easier to spot or add things to something that are wrong than to add something that fits in perfectly.

I found some pointers to Gestalt Theory but I couldn't find anything there yet. I'll keep looking but maybe someone here knows what I'm trying to convey. Thanks.


The predictive coding model of brain function suggests that the entire function of the brain is to update an internal model of the world.

This process occurs by generating predictions based on the current model, and comparing these predictions to sensory information. The difference between the predictions and the actual information is the prediction error: this is the stuff that the brain has to work harder on, because it is these errors that require further action to update the models and make better predictions in the future.

Based on this model, unexpected things are highly salient, because they generate large prediction errors. In contrast, expected things are boring, and can mostly be ignored. This is a mechanism to filter the massive amount of sensory information we are capable of taking in at any given moment.

As of now, predictive coding is still just a model, and there is a lot of work being done towards finding evidence for predictive coding in the brain. That said, it explains the phenomenon you are describing fairly well.

Some references for further reading:

Bastos, A. M., Usrey, W. M., Adams, R. A., Mangun, G. R., Fries, P., & Friston, K. J. (2012). Canonical microcircuits for predictive coding. Neuron, 76(4), 695-711.

Kok, P., & de Lange, F. P. (2015). Predictive coding in sensory cortex. In An introduction to model-based cognitive neuroscience (pp. 221-244). Springer, New York, NY.

Melloni, L., van Leeuwen, S., Alink, A., & Müller, N. G. (2012). Interaction between bottom-up saliency and top-down control: how saliency maps are created in the human brain. Cerebral cortex, 22(12), 2943-2952.

Rao, R. P., & Ballard, D. H. (1999). Predictive coding in the visual cortex: a functional interpretation of some extra-classical receptive-field effects. Nature neuroscience, 2(1), 79-87.

Spratling, M. W. (2012). Predictive coding as a model of the V1 saliency map hypothesis. Neural Networks, 26, 7-28.

Yuille, A., & Kersten, D. (2006). Vision as Bayesian inference: analysis by synthesis?. Trends in cognitive sciences, 10(7), 301-308.

  • $\begingroup$ Thanks, I accepted your answer. But do you maybe have a reference to a paragraph or similar where your statement written in bold comes from? I couldn't read it out directly in the Wikipedia article. $\endgroup$ – Charles Nough Feb 26 '20 at 22:55
  • $\begingroup$ @CharlesNough It's not a quote of anything in particular, but I've included some other more scientific references to follow up on. The Spratling paper is probably the one closest to focusing on that particular phrase. The idea itself is fairly central to predictive coding as a framework, though, because the signal propagated up the hierarchy is the error, which is the unexpected part. $\endgroup$ – Bryan Krause Feb 26 '20 at 23:06

In fact one might say that pretty much most of human cognition is detecting when things diverge from the norm.

Some simple examples include:

Movement detection is nothing more than detecting when the input in our eyes diverges from the norma of nothing moving. Equally the same can be said with detection of sudden sounds.

Of course, once a sudden sound like a washing machine turning on might continue for so long it becomes the norm, then we may then not even consciously notice it.

Remembering faces, is a process of coding how each face diverges from the norm.

McCarthy (inventor of Lisp) proposed that human knowledge should be encoded as "exceptions". e.g. "a penguin is a bird except it can't fly".

Neural networks, also, first converge to the norm, and then refine themselves by the various exceptions to that norm.

  • $\begingroup$ Thanks for your examples, can't give you any points unfortunately, since I'm not ranked high enough. But where would you categorise this feeling? Bryan Krause put in to the model of predictive coding, is this where you put it as well or would you assign it to some other direction? $\endgroup$ – Charles Nough Feb 27 '20 at 9:39

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.