Multichannel Multitasking

Multichannel Multitasking

Multichannel multitasking is Nir Eyal's label, from Indistractable in 2019, for a carve-out from the general finding that multitasking is costly: two tasks competing for the same channel interfere, while two tasks using different channels can sometimes run together. The underlying science is Christopher Wickens's multiple resource theory, and the carve-out is real but narrower than the label suggests. Cross-modal pairs still compete for central control, so the right test is not whether the senses differ but whether one of the tasks is automatic.

What it is

Eyal concedes the standard finding before making his exception, which is what makes the exception credible. Two podcasts, one in each ear, is impossible. An audiobook while walking is not.

The mechanism he invokes is sensory: different channels, less collision. The better statement of it is Wickens's multiple resource theory, published in Theoretical Issues in Ergonomics Science in 2002, which holds that tasks drawing on different modalities interfere less than tasks sharing one. That part is well supported.

The limit is where the popular version goes wrong. Adam Gazzaley and Larry Rosen concede the same thing Eyal does, that genuinely parallel processing happens and automated actions can run alongside a task at no cost, and then draw the boundary more sharply: the competition is not for a sensory channel, it is for cognitive control. Two tasks that both need the prefrontal cortex compete even when they arrive through different senses, which is why the brain switches networks rather than running both.

In effect

That trims the carve-out considerably. An audiobook while walking is safe because walking is automated, not because hearing and legs are different channels. An audiobook while doing anything requiring judgement is not covered by the exception at all.

David Strayer's workload framing is the usable rule. Non-interactive audio is cognitively cheap, conversation is expensive, and the gap holds well beyond driving. It is the same reason a passenger's conversation does not impair driving the way a phone call does, while a podcast sits somewhere in between. Strayer and William Johnston showed in 2001 that phone conversation impairs driving despite using different senses from the visual task, which is the cleanest demonstration that different modalities are not sufficient.

So the stacking Eyal recommends, pairing an audiobook or a saved article with exercise, is fine where one side is automatic. It is not a general licence, and the test is not whether the senses differ. It is whether the second thing requires a decision.

What it does not say

It does not say that different senses means no interference. Interference falls with modality separation and rarely disappears.

It does not licence stacking two demanding tasks. Anything needing judgement on both sides is outside the exception.

It does not overturn the switching literature. The costs of switching between tasks that share control are unaffected by this carve-out.

And the term is a label rather than a discovery. The mechanism belongs to Wickens's multiple resource theory, and the credit for it should go there.

Sources

  1. Eyal, N. (2019). Indistractable. Ch. 20, pp. 143-145. The label and the carve-out. The walking-and-creativity study behind it is footnoted with no researcher named; it is most likely Oppezzo, M., & Schwartz, D. L. (2014), Journal of Experimental Psychology: Learning, Memory, and Cognition, 40(4), 1142-1152, which should be confirmed against the endnote before citing.
  2. Wickens, C. D. (2002). Multiple resource theory. Theoretical Issues in Ergonomics Science, 3(2), 159-177. The mechanism.
  3. Strayer, D. L., & Johnston, W. A. (2001). Psychological Science, 12(6), 462-466. Phone conversation impairs driving despite different senses.
  4. Gazzaley, A., & Rosen, L. D. The Distracted Mind. The correction that competition is for cognitive control rather than for a sensory channel.
  5. Evidence grade: mixed. The modality part is well supported; the carve-out holds only where one task is automatic, and the vault records that revised version as the one to publish.