You Don't Have a Spotlight in Your Head
Attention isn't a beam that illuminates the world. It's a volume knob that controls which prediction errors your brain takes seriously.
The most famous psychology experiment of all time involves a gorilla.
In 1999, Daniel Simons and Christopher Chabris at Harvard asked people to watch a video of two teams passing a basketball. Count the passes made by the team in white. Simple enough. Halfway through, a person in a gorilla suit walks into the frame, beats their chest, and walks off.
About half the participants didn't see it.
Not "didn't notice" in the vague sense. They genuinely did not perceive a gorilla standing in the middle of their visual field for nine full seconds. When told about it afterward, some accused the researchers of switching the video.
The standard explanation is that attention works like a spotlight. You point it at the basketball passes, and the gorilla falls outside the beam. Clean metaphor. Easy to picture.
It's also wrong.
Precision, Not Spotlights
The predictive processing framework gives us a better answer. Your brain isn't shining a light on some things and leaving others in darkness. It's adjusting the volume on different channels of prediction error.
Harriet Feldman and Karl Friston laid this out in their 2010 paper "Attention, Uncertainty, and Free-Energy" in Frontiers in Human Neuroscience. Their argument: attention is the process of optimizing precision-weighting on prediction errors. When you "pay attention" to something, you're turning up the gain on errors from that source. Those errors hit harder. They update your model more aggressively.
When you're not paying attention to something, the errors from that source still exist. They're still being generated. Your brain just assigns them low precision, so they barely register.
This is a fundamentally different picture than the spotlight model. The gorilla data was hitting your retina the whole time. Your visual cortex was generating prediction errors about it. But because your precision-weighting was cranked up on "white shirts passing a basketball" and cranked down on everything else, the gorilla errors never reached the threshold for conscious updating.
The data was there. The weight wasn't.
Andy Clark describes this in "The Experience Machine" as one of the most counterintuitive implications of the predictive brain. You don't miss things because they're filtered out before they reach you. You miss them because your brain decides, in real time, how seriously to take each stream of incoming error. Attention is not selection. It's prioritization.
Your Name in a Crowded Room
Now flip it. You're at a loud party. Multiple conversations happening. Music playing. You're deep in a discussion about something and paying zero attention to the group behind you.
Then someone in that group says your name.
You hear it instantly. Like someone flipped a switch.
Colin Moray demonstrated this in 1959 at the Quarterly Journal of Experimental Psychology. He had people listen to different audio in each ear and focus on only one channel. About 33% of participants detected their own name in the unattended channel. Almost no one detected any other word.
The spotlight model has no good answer for this. If attention is a beam pointed at one ear, how does information from the other ear get through?
Precision-weighting handles it cleanly. Your brain maintains certain priors at permanently high precision. Your name is one of them. It's been reinforced tens of thousands of times since childhood. So even when you're allocating low precision to background noise, name-related prediction errors carry enough baseline weight to break through.
You didn't "hear through the filter." There was no filter. Your brain just has a few channels that never fully turn down.
The Volume Knob Theory of Focus
Think of your sensory experience as a massive mixing board. Hundreds of channels, all live. Visual data from every part of your field of vision. Sounds from every direction. Touch, temperature, proprioception. It's all coming in, all the time.
Attention is you (or more accurately, your brain's automatic precision-weighting system) adjusting the faders on that board. Some channels get boosted. Their prediction errors drive real updates to your world model. Other channels get pushed to near-zero. Their errors are technically there but functionally irrelevant.
This is what Jakob Hohwy describes in "The Predictive Mind" as the brain's solution to an impossible problem. You can't process everything at full resolution. The computational cost would be absurd. So instead of filtering out data, you adjust how much each stream of data matters. Precision-weighting is computationally cheaper than full processing, and it's flexible. You can shift the weights in milliseconds.
Anil Seth extends this in "Being You." He argues that precision-weighting doesn't just control what you notice. It controls what you consciously experience. Consciousness itself might be a function of which prediction errors are assigned high enough precision to update your self-model. The gorilla isn't just unnoticed. For all practical purposes, it doesn't exist in your experience.
ADHD Is a Precision Problem
This framework completely reframes attention disorders.
ADHD isn't a broken spotlight. It's dysregulated precision-weighting. The brain has trouble maintaining high precision on task-relevant prediction errors and keeps inappropriately boosting precision on task-irrelevant ones.
You're trying to read a document. Your brain should be assigning high precision to visual prediction errors from the text and low precision to the ambient sounds in the room, the feeling of the chair, the notification that just popped up on your phone.
In ADHD, those weights keep shifting. The notification gets a sudden precision boost. The chair feeling gets a precision boost. The conversation in the next room gets a precision boost. Not because the person isn't trying to focus. Because the precision-weighting system isn't holding its settings.
Rebecca Lawson, Christoph Mathys, and Geraint Rees published computational evidence for atypical precision-weighting in a 2017 Nature Neuroscience paper. Their study focused on autism, showing that autistic adults overestimate the volatility of their sensory environment, leading to atypical precision assignments. But the same hierarchical Bayesian framework has been applied to ADHD, where the problem is essentially inverted. Not overly rigid precision, but overly volatile precision.
Here's where it gets practical. The dopaminergic system, the brain's dopamine machinery, is thought to be a key neural substrate for precision-weighting. Wolfram Schultz's landmark 1997 paper in Science showed that dopamine neurons don't just signal reward. They signal prediction error. Dopamine encodes the difference between what you expected and what happened.
If dopamine is the currency of precision-weighting, then ADHD, which involves disrupted dopamine signaling, is literally a disorder of how the brain assigns importance to prediction errors.
This is why stimulant medications work. Ritalin and Adderall boost dopamine availability. In the predictive processing framework, they're not "helping you focus" in some vague way. They're restoring the brain's ability to maintain appropriate precision-weighting. Keep task-relevant errors loud. Keep task-irrelevant errors quiet.
Meditation as Precision Training
Karl Friston's free-energy principle (from his 2005 paper "A Theory of Cortical Responses") predicts that any system minimizing surprise should be able to improve its precision-weighting with practice. And that's exactly what meditation research shows.
Norman Farb and colleagues found in a 2007 Social Cognitive and Affective Neuroscience paper that mindfulness meditation shifted activity between two distinct neural networks for self-reference. Meditators could decouple from the "narrative self" network (the one that worries about the past and future) and engage the "experiential self" network (the one that processes present-moment sensory data).
In precision-weighting terms: meditation trains you to manually adjust which prediction errors get high precision. You practice boosting precision on breath-related sensory errors and reducing precision on thought-related errors. Over and over. The mental equivalent of practicing scales.
Antoine Lutz and colleagues confirmed this in a 2008 Trends in Cognitive Sciences review. Long-term meditators showed enhanced attentional regulation, which the authors framed as improved ability to allocate and sustain precision on chosen targets.
I've noticed this in my own practice. Meditation doesn't empty your mind. The background channels don't go silent. You just get better at keeping the faders where you want them. Thoughts still come. You just stop turning up their volume.
What This Changes
The spotlight metaphor made attention feel mysterious. Some force in your head shining a beam around, and you either had control of it or you didn't.
Precision-weighting makes it mechanical. Your brain is running a mixing board. Every channel is always live. Attention is just the gain settings. Some of those settings are automatic (your name), some are trained (meditation), some are chemically regulated (dopamine), and some are disordered (ADHD, anxiety).
Understanding the mechanism doesn't give you instant control. But it does tell you where the knobs are. And that your brain isn't broken when the gorilla walks by unnoticed. It's just running the mix you asked for.
Sources
- Gorillas in Our Midst: Sustained Inattentional Blindness for Dynamic Events (Simons & Chabris, 1999, Perception)
- Attention, Uncertainty, and Free-Energy (Feldman & Friston, 2010, Frontiers in Human Neuroscience)
- A Theory of Cortical Responses (Friston, 2005, Philosophical Transactions of the Royal Society B)
- Adults with Autism Overestimate the Volatility of the Sensory Environment (Lawson, Mathys & Rees, 2017, Nature Neuroscience)
- A Neural Substrate of Prediction and Reward (Schultz, 1997, Science)
- Attending to the Present: Mindfulness Meditation Reveals Distinct Neural Modes of Self-Reference (Farb et al., 2007, Social Cognitive and Affective Neuroscience)
- Attention Regulation and Monitoring in Meditation (Lutz et al., 2008, Trends in Cognitive Sciences)
- The Experience Machine: How Our Minds Predict and Shape Reality (Clark, 2023, Pantheon)
- The Predictive Mind (Hohwy, 2013, Oxford University Press)
- Being You: A New Science of Consciousness (Seth, 2021, Dutton)
- Some Effects of Unattended Messages: A Revision (Moray, 1959, Quarterly Journal of Experimental Psychology)
Part of the Prediction Machine series. Previous: Your Brain on Psychedelics Is Your Brain with the Filter Off. Next: You Are a Hallucination You Tell Yourself.
