The Octopus Illusion: Why Multitasking Feels Like a Superpower and Isn't

An octopus has a nervous system that genuinely distributes control across its body: roughly two-thirds of its neurons live in its arms, each of which can process sensory information and execute basic movements without waiting for instruction from the central brain. When an octopus reaches into a crevice with one arm while using two others to anchor itself and a fourth to manipulate something at close range, that really is something close to parallel processing — a kind of attention distributed across multiple semi-autonomous systems in a way mammalian brains don’t work.

This is worth stating upfront because the human version of “I can multitask” is doing something entirely different, and the octopus comparison is not flattering to us.

What the brain is actually doing

When you switch between a document and a notification and back, what’s happening neurologically isn’t parallel processing — it’s serial attention switching with a residual activation cost. The research on this is extensive enough to have a name: task-switching costs, or in more informal literature, attention residue. The core finding, which has held across many replication attempts and experimental designs, is this: after redirecting attention from task A to task B, a measurable portion of cognitive resources remains occupied with task A for some time afterward. You’re reading task B’s content, but part of your working memory is still running a process related to task A. The completeness of any given moment of “focus” is lower than it feels like it is.

This matters more at the margin than it sounds. A single switch isn’t ruinous. The cost of one brief notification check is small enough that you genuinely won’t notice it in the output of any given paragraph. What compounds is the frequency of switching across a session. Knowledge work that requires holding a complex model in working memory — debugging, writing, designing, reading to understand something subtle — degrades in measurable ways under conditions of frequent interruption, even interruptions that feel brief and controlled. Research by Gloria Mark at UC Irvine found that the average knowledge worker was interrupted or self-interrupted roughly every 40–50 seconds of desk work. Recovery to the depth of focus present before the interruption took an average of 23 minutes — not because the task itself was that complex to re-enter, but because of the cascading series of “just one more thing” decisions that typically followed each interruption.

The self-assessment problem

The most counterintuitive finding in multitasking research is who reports being best at it. A well-cited study by Ophir, Nass, and Wagner (2009) identified a population they called heavy media multitaskers — people who habitually consume or switch between multiple information streams at once — and tested them on a battery of attention and task-switching measures against lighter multitaskers. The prediction was that frequent multitaskers would be better at it through practice.

They weren’t. Heavy multitaskers performed worse on tasks requiring them to filter out irrelevant information, switch deliberately between tasks, and manage working memory. They were more distractible and less able to suppress task-irrelevant stimuli than the group that multitasked less. There’s since been follow-up work complicating this result, but the core finding — that self-reported skill at multitasking and measured performance on attention tasks are weakly or negatively correlated — has survived reasonably well.

This creates a calibration problem that’s easier to state than to solve: the experience of multitasking, especially with familiar apps and content, feels smooth and controlled in a way that doesn’t map to what’s actually happening in working memory. The switching cost is invisible in the moment. The output cost tends to appear later, in revision time, in missed details, in the vague sense that something took longer than it should have — without any obvious candidate for why.

What digital context makes worse

The octopus manages its arms through genuine parallel nervous systems. Human “multitasking” between digital apps is switching between interfaces each of which has been designed to recapture your attention as efficiently as possible. These are not symmetric players in this dynamic.

An application designed for attention recapture — notifications, counts, new-content indicators, presence pings — will interrupt more often than any explicit decision to check it would, because the environmental design is built around manufacturing the impulse to check before a conscious decision is made. The switching cost appears to be small because each individual switch is small. The accumulated cost across a session doesn’t present as a single obvious event; it presents as work that took longer than expected and left you more tired than the actual content difficulty warranted.

Working with what you have

Human attention isn’t fixable. The task-switching cost isn’t going to be trained away; it’s structural. What’s malleable is the frequency with which switching is triggered, and the intentionality of the switches that do happen.

A few patterns that consistently help:

Separate the modes. Inbox-open, notifications-on work produces different output than inbox-closed, notifications-off work for tasks requiring depth. These aren’t compatible environments; choosing between them upfront is more effective than trying to manage the intrusion in real time.

Use time boundaries, not willpower. Deciding at the start of a session that you’ll check messages at a specific time — not “when I feel like it” — removes the decision cost of every subsequent urge to check. The cost of maintaining the decision is lower than the cost of making it repeatedly.

Notice the impulse before it becomes a switch. There’s a brief interval between “the urge to check something” and “the phone is already in hand.” Getting familiar with that interval, without necessarily acting on it, is what building real pause between impulse and action looks like in practice. It doesn’t require willpower in the conventional sense; it requires noticing a moment that previously passed too fast to register.

Set the environment, not the standard. Trying to concentrate harder in an interrupt-rich environment is mostly a way to be harder on yourself without changing the output. Changing what’s in reach changes the behavior more reliably than changing the intention.

Eight arms, one brain

The octopus does something genuinely unusual with distributed nervous system architecture that took 300 million years of evolution to produce. Using twelve open apps and calling it multitasking is not that. The brain you have is good at depth, sequencing, and sustained attention on one thing at a time. It’s not bad at attention management — it’s bad at the specific kind of environment that modern software was designed to create. Knowing the difference is most of the practical value of the octopus comparison.