Predictions in the dark

I awoke one night to murmuring sounds. No longer asleep, not yet awake I tried to drift back into sleep again, but the sounds persisted and morphed into a high-pitched squealing or screaming. Suddenly I realised with absolute clarity: a woman is being murdered outside my house. I pretended to not hear anything and tried to go back to sleep, but not helping was not an option: I had just read about the murder of Kitty Genovese in the 1960s in New York City, where she had been stabbed repeatedly over the course of 35 minutes, with 38 potential witnesses from nearby apartments not helping her, not even calling the police. I had no way to fight the murderer myself, but the least I could do was call the police.

I got up and hesitated briefly before parting the curtains. The yellow streetlights illuminated an empty road. I looked to the left, to the right, but there was no one. The sound had also stopped until suddenly there was a new moan from right in front of me. Our front garden was separated from the pavement via a hip-high wall and in the obscuring shadow of that wall someone moved, crouching above someone else.

I backed up into the safety of my room’s darkness. Someone was actually being murdered in my front garden; I hadn't just imagined it. I needed to call the police. I picked up my mobile phone and realised that I hadn't actually seen the guy stab the woman. Maybe I was confused and it was just a drunk guy who’d fallen into our front garden? No, those screams were female, and of great pain and despair. Nevertheless, I needed to double-check. And so I parted the curtains once more and saw the shapes moving in the shadow. I was about to dial 911, but held my gaze for another second when my entire perception changed within an instant: what I saw was not a man crouching over a woman, but a fox standing over a large bird or small mammal.

How was this possible? How could I mistake the sound of a bird dying for a woman being murdered, and see a man instead of a fox? To provide an answer to this question it’s important to consider what I knew and what I didn't know.

A few days before these events, I moved from a practically crime-free place to one of the areas with the highest crime rates in London. I was wary of crime around me, almost expected it to happen. Additionally, I didn't know there were foxes in big cities such as London. Most crucially though, I had just read about the bystander effect, that was investigated after two scientists read the above-mentioned story about Kitty Genovese. Is it any surprise then that I interpreted these events as a man murdering a woman rather than a fox killing a bird? I had no reason to assume the latter would ever happen in London, but every reason to assume that someone being murdered would happen in London. But still – I had seen a fox as a human and heard a bird as a woman. How could something as fundamental as perception fail because of a story I’d heard?

To provide a more detailed explanation of how perception itself could be influenced by such knowledge, it is necessary to consider one of the most popular current theories of perception: predictive coding. According to predictive coding, perception is not a passive process by which sensory information is absorbed as the data hits the senses; instead, the brain creates models of the world and tests these with sensory information. Any error between the prediction and the actual sensory information then updates and improves the model.We don’t perceive directly what’s out there, we perceive what we predict, unless the information from our senses contradicts these predictions. This also means that we are also more likely to misperceive something in situations with imprecise data, such as in the dark or when a sound is fairly quiet. Sensory data only changes our perception when it deviates significantly from the model, so if the sensory data is not very precise, it is unlikely to provide enough evidence against a model to change it.

This can be used to explain my error: because of what I knew and didn't know about life in London, I had a prior belief that 1) people murder each other in London, and 2) there are no foxes (or other larger carnivorous mammals). Additionally, the story of a woman being murdered in a large city was freshly on my mind from having read the story. So when I heard the quiet, ambiguous sounds, the sensory evidence from the dying bird was not sufficient to change my brain’s prior hypothesis that this was a woman being murdered. Additionally, since this sensory evidence didn't disagree with my hypothesis, it strengthened my prior belief that someone was being murdered. When I then briefly saw the fox and the bird in the shadows at night, I couldn't actually see all that much. My prior belief was fairly strong towards one hypothesis, and the new sensory evidence wasn't precise, so it again just acted to reinforce what I already believed. But when I finally looked at the scene for another 2-3 seconds (and my eyes had a chance to adjust to the darkness, which improved the quality of the data), the difference between what my brain predicted to receive in sensory information, and what it then actually received was strong enough to signal that the model was completely wrong and that a different model, namely a fox killing a bird, was the better fit for the data. And so my perception changed from ‘a man murdering a woman’ to ‘a fox killing a bird’.



Hohwy, J. (2013). The predictive mind. Oxford University Press.

Latané, B., & Darley, J. M. (1970). The unresponsive bystander: Why doesn't he help?. Prentice Hall.

Slater, L. (2005). Opening Skinner's box: Great psychological experiments of the twentieth century. WW Norton & Company.

Future reality

Do-it-yourself brain stimulation-kit for home-enthusiasts