What feels like multitasking on cognitively demanding tasks is, for almost everyone, actually rapid task-switching — the brain handles one task at a time and toggles between them, paying a real cost (called switch cost) each time: a brief period of reduced performance and slower reaction time while attention reorients to the new task. Research measuring this consistently finds that people switching between two demanding tasks are slower and make more errors on both than people who do the tasks sequentially, even though the switchers usually feel like they're being more efficient.
The exception is genuinely automatic tasks — walking while talking works fine because walking requires minimal conscious attention for most people. The failure mode is specifically pairing two tasks that both require conscious, effortful attention (writing an email while in a meeting, texting while driving), where the felt sense of managing both is not matched by actual performance on either.
Rubinstein and colleagues had participants alternate between different cognitively demanding tasks (like solving math problems and classifying geometric shapes) and precisely measured the time cost each time attention had to reorient from one task to the other. Even for simple, well-practiced tasks, switching produced a measurable delay — time the brain spent disengaging the rules of one task and activating the rules of the next — and the cost grew with task complexity. Across their experiments, people switching between two demanding tasks were consistently both slower and more error-prone on both tasks than people who completed the same two tasks sequentially, despite the switchers often reporting they felt they were being efficient.