Capacity allocation is dynamic — a harder sub-task takes a larger share at other tasks' expense

What it is

Rather than picturing attention as a filter fixed at one location, it can be pictured as a flexibly allocatable resource: the total is finite, but how it is divided among whatever tasks are currently underway can be re-adjusted at any moment. This is the capacity allocation model, systematically proposed by Kahneman: the harder a task is, and the more active effort it requires, the larger the share of resources it draws, leaving a correspondingly smaller share for other tasks.

This answers a different question from early/late selection theory — those theories debate at which stage filtering occurs; this model is concerned with how the total resource gets divided among simultaneous tasks and what determines that ratio. It is a separate main line of thinking alongside bottleneck theory rather than a competing answer to the same question.

Why it happens

The model treats available attentional resources as a pool whose total fluctuates with arousal level but is finite at any given moment: the total is larger at moderately elevated arousal, and shrinks when arousal is too low or too high — the same underlying logic as the inverted-U arousal-performance relationship.

The rule for allocation is not an even split; it is governed by an allocation policy that weighs several factors together: the current task's difficulty (harder tasks need more resources), the task's importance or how much effort someone is personally willing to invest, and fixed allocation tendencies formed by long-standing habit. Difficulty is the most direct driver — for the same task, as difficulty rises, the resources it needs to occupy rise monotonically, and once one task pushes its resource use high enough, the share left for other concurrent tasks is automatically compressed, showing up as a performance drop in those tasks even though their own difficulty hasn't changed.

This model explains something both early and late selection theory struggle with: the degree of dual-task interference is not fixed — it varies continuously with the difficulty of one of the tasks. The harder that task gets, the more visibly it crowds out the other — direct evidence for "resources being dynamically reallocated" rather than "a channel being fixed open or shut."

Studying it

A common approach manipulates the difficulty of one task and observes the crowding-out effect on a concurrent task: participants perform two tasks simultaneously, and the difficulty of one is gradually increased (more computation, a shorter response window, harder discrimination), while the resulting drop in the other task's performance is measured.

Common independent variables: graded levels of primary-task difficulty, and whether the secondary task itself requires active effort. Common dependent variables: the slope of the secondary task's performance decline as primary-task difficulty rises, and the correlation between physiological arousal measures (such as pupil diameter or heart-rate variability) and task difficulty.

This paradigm is commonly used to quantify the relationship between task difficulty and resource consumption — going beyond a simple "is there interference or not" to answer the finer question of "how does the degree of interference change with difficulty," making it a standard tool for evaluating how a high-load task affects a secondary one.

A methodological caution: the model's operationalization of "effort" and "resources" (commonly using pupil diameter as a physiological marker of arousal/effort) itself rests on an assumption — that pupil dilation genuinely reflects invested cognitive effort rather than some other physiological process — and this assumption needs careful handling when interpreting specific data.

Where it stops holding

• The allocation policy itself can be overridden by subjective willingness. For tasks of equal objective difficulty, a participant who subjectively wants to do well will invest more resources — meaning objective difficulty alone cannot fully predict the outcome of allocation.
• The total is regulated by arousal level, not a constant. Fatigue, stress, and emotional state all first change the available total before affecting the allocation outcome, so the same person doing the same combination of tasks can show different degrees of interference depending on their state at the time.
• The model does not explain why interference between certain task pairs is especially small (a visual task paired with a spoken one, for instance) — this kind of finding was later attributed to the tasks drawing on different types of resource, not merely a matter of quantity allocation. That belongs to a different theoretical framework and falls outside what this model explains.
• This describes dynamic allocation within a single pool of total capacity; it does not address the deeper question of whether the resource is a single homogeneous pool at all.

Related

• Same group: A5.10.1 Limited processing resources mean simultaneously presented information necessarily involves trade-offs · A5.10.2 Early selection theory holds that filtering happens before semantic analysis · A5.10.3 Late selection theory holds that all channels are processed to the semantic level, with selection happening afterward · A5.10.5 The bottleneck's location shifts with task type; there is no single fixed filtering stage
• Nearby: A5.02 Divided attention and dual-tasking · A9.09 Physiological measures of cognitive load
• Search terms: Kahneman capacity model · attentional resource allocation · arousal and performance · dual-task interference