Multitasking Practice Doesn't Erase the Mental Bottleneck

Can Practice Make You Handle Two Tasks at Once?

You check Slack while cleaning up meeting notes. You skim the deck for your next meeting while drafting an email. The more familiar you become with this kind of juggling, the faster you get at running several tasks together. Cognitive psychology experiments have shown the same pattern: with enough repetition, the performance cost of doing two tasks together shrinks considerably.

So can practice eventually get you to a point where two tasks simply stop interfering with each other? A study first published online in November 2025 says no — even after up to 12 days of training, interference remained at a specific processing stage. Let’s look at what the experiment actually involved.

Two Tasks, Practiced for Up to 12 Days

The study was conducted jointly by Torsten Schubert of Martin Luther University Halle-Wittenberg in Germany, Roman Liepelt of FernUniversität in Hagen, and Tilo Strobach of Medical School Hamburg. The paper appeared in the Quarterly Journal of Experimental Psychology, and the university publicized the findings in March 2026.

Across three experiments, the researchers gave participants two tasks — one visual, one auditory. In one, participants used their right hand to indicate the size of a circle that flashed briefly on screen. In the other, they had to say whether a tone they heard was high, medium, or low.

Participants practiced the tasks — sometimes one at a time, sometimes both together — for up to 12 days. As expected, they got faster and made fewer errors with practice. Earlier studies had observed the same improvement, and some interpreted it to mean that with enough skill, the two tasks would barely interfere with each other at all.

This time, the researchers manipulated how long it took to select a response after training. That’s the stage where, after recognizing the circle or the tone, a person decides which response to give according to a fixed rule. The team wanted to know: if this stage takes longer for one task, does it slow down the other task too?

When they lengthened the response-selection time for the relatively short visual task, the response to the longer auditory task was delayed as well. But when they lengthened the response-selection time for the longer task, the shorter task showed no such effect. The researchers concluded that this pattern is consistent with the two tasks’ response selection happening one after another, in sequence.

Practice improved overall performance, but changing the conditions revealed interference between the two tasks again. The fact that people perform quickly under familiar conditions doesn’t mean every processing stage is running in parallel.

Getting Better Isn’t the Same as Interference Disappearing

The explanation the researchers tested is called the “latent bottleneck model.” While one task is selecting its response, the other task has to wait its turn to enter that same stage. The point where processing gets held up like this is called the bottleneck.

Under this model, not everything happens strictly one step at a time. Stages like taking in sensory information or executing a motor response can overlap. The stage considered limited is specifically the one where each task decides which response to give.

Practice can improve both how fast a task is processed and how well two tasks are coordinated. But that improvement doesn’t mean the underlying capacity limit has disappeared. This study measured reaction times and errors, so its conclusions shouldn’t be stretched to cover every kind of everyday work or every length of training.

What Happens in a Workplace That Uses AI Tools

Beyond the lab, actual workplace activity data is worth a look too. ActivTrak, a workplace analytics software company, analyzed digital work activity from 777 client organizations and 218,900 employees between 2022 and 2024 in a 2025 report.

In this sample, 58% of employees used AI tools — up 107% from 2022. This should be read as a change observed among this company’s specific client base, not a universal statistic.

The same report found that collaboration time rose 27% and multitasking time rose 5%. Productive time spent in work apps and sites increased 2%, but the share of “focus time” within total recorded time fell from 65% to 62%, and the average length of a single focus session shortened by 8%. Here, “focus” is a category the software assigns based on digital activity patterns — it isn’t a direct measure of what’s happening in an employee’s head.

The report shows that AI adoption and changes in work patterns are occurring together. It’s not an experiment that proves AI causes reduced focus. It’s also worth considering that people with heavier workloads or different job roles may simply have been more likely to adopt AI in the first place.

There’s nothing inherently wrong with a setup where AI drafts something while a person handles other work — software and human working in parallel is fine on its own. What needs scrutiny is what comes next: the burden created when a person has to review multiple drafts in rotation, respond to messages, and make decisions, all around the same time.

In the university’s press release, Professor Strobach raised the safety implications for situations that require juggling multiple tasks at once — like talking on the phone while driving, or air traffic control. To be clear, this experiment didn’t directly measure performance in those professions or among AI users specifically.

Oswarld’s Lens

I’ve watched skilled teams struggle after being handed two projects at once. At first, reports get delivered, meetings happen, schedules move forward. But then a client makes an unexpected demand, the market shifts suddenly, or one team member leaves — and both projects’ performance can decline together.

A team isn’t a single brain — work can be split among several people. Still, reading this study, I was reminded of situations where things seemed to be going fine under familiar conditions, and that was taken as proof there was plenty of slack in the system. What this research made me think is that you also need to know, in advance, who takes on the review and decision-making when the workload suddenly spikes.

What I’m paying attention to is how human work changes after AI gets introduced. Even if AI cuts down drafting time, if that saved time just gets funneled into piling more review and decision-making onto one person, the burden can grow rather than shrink. You need to look at the number and difficulty of tasks someone is juggling at once, and how often they have to switch between them.

When companies bring in AI, I’d like them to also protect blocks of time where a person can focus on one thing. Beyond counting how many extra drafts got produced, I mean tracking whether reviews got finished on time, and whether missed details or rework actually went down.

Check This Before You Add More Work

This experiment made one thing clear: repeated practice genuinely helps. But once conditions changed, interference between the two tasks reappeared. So even when handing more work to a skilled person, I’d caution against judging based purely on how fast they normally handle things. You also need to decide, in advance, what gets triaged first when an unexpected request comes in, and whether existing work can be paused to make room for it.

If you’re juggling work that demands instant replies to messages alongside a proposal that needs deep focus, in the same time slot, it might help to first define how much time can actually be split between the two. And it’s worth checking whether the time AI saves you actually turns into real, protected focus time — or just gets absorbed elsewhere.

References & Further Reading

• Schubert, T., Liepelt, R., & Strobach, T., Evidence for a Latent Bottleneck After Extensive Dual-Task Practice of a Visual-Manual and an Auditory-Verbal Task, first published online November 2025. A study that manipulated response-selection time in two tasks to examine interference.
• Martin Luther University Halle-Wittenberg, Psychology: Study shows limits of multitasking, March 11, 2026. A university press release covering the experimental procedure and the researchers’ explanation.
• Schubert, T., Kübler, S., & Strobach, T., A mechanism underlying improved dual-task performance after practice: Reviewing evidence for the memory hypothesis, 2024. A review examining how practice improves task-coordination ability.
• ActivTrak, 2025 State of the Workplace. A report analyzing client organizations’ digital work activity from 2022 to 2024.

The author is Oswarld (Kwangseob Ahn). Current roles: Adjunct Professor at Sejong University, Strategy Consultant at INLEVEL9. Career, research, books, and recent work are kept current on the About page. Latest · July 2026: HEMA-2: A Consolidation-Aware Tri-Memory Architecture with Multi-Channel Scheduling for Lifelong Conversational AI.