Task uncertainty, not lost focus, drives multitasking errors

Clinical takeaway: Sequential work protects accuracy where multitasking degrades it. That's because the interference happens in perception, not motivation. 

A clinical day is built on divided attention: messages arriving mid-encounter, results to review between patients, a call cutting into documentation. The standing assumption is that the mistakes which follow are lapses of effort by people who should have focused harder. 

Cognitive science has long measured a performance cost when people switch tasks, usually blamed on attention slipping. What those accounts left unexplained is why errors multiply specifically when a person is unsure which task applies, a condition closer to real working life than tidy alternation between two known jobs. A new study locates the cost somewhere more fundamental: in how the brain represents information itself. 

Participants watched a striped circle appear on a screen, followed by a second image that had changed in one of two ways: the pattern had shifted position, or the stripes had changed in width. They then reported which of the two features they had been tracking and what change they saw. The catch was that no one was told which feature counted as the right one to track. Participants had to infer it from feedback on earlier attempts, and the correct answer periodically changed without warning. 

Accuracy fell precisely when that inference was shakiest. The errors were not the product of drifting attention. When people were unsure whether position or stripe width mattered, they held onto the feature they should have ignored, and that irrelevant visual information contaminated the perceptual judgment itself, a pattern the authors call feature interference. Participants recovered within a few attempts once feedback clarified which feature mattered. 

The contrast that sharpens the finding came from computational models. An artificial network trained simply to perform the same tasks correctly showed no accuracy cost under uncertainty. A second network, trained instead to mimic the choices participants actually made, developed the same interference. 

The researchers tested more than 200 human participants on the feature-tracking task, alongside recordings of neural activity in animal work and the paired network models. The study was funded through the Simons Foundation's Simons Collaboration on the Global Brain and NIH. 

The practical reading is that working conditions, not effort, set the ceiling. The authors suggest the mechanism may extend beyond vision. The same mixing could explain why reading fails when someone is talking nearby, though the experiments tested visual features only. For settings where interruptions are unavoidable, that points the leverage toward reducing ambiguity about what matters right now rather than exhorting anyone to focus. 

"The brain is good at keeping track of a lot of different things, but it becomes a big job to make sure they don't get in the way of each other," said Marlene Cohen, PhD, professor of neurobiology at the University of Chicago and senior author. "When you're under high cognitive load, you're trying to do too many things at once. Information gets mixed together in your brain even if you know it very well, and that turns out to be an explanation for this behavior." 

Source: Xue C, et al. (2026 Sep 10) Nat Neurosci. Feature interference underlies a neuronal basis for the behavioral cost of task uncertainty