Multiple resource theory explains why dual-task interference can't be predicted from task difficulty alone

What it is

The most important thing multiple resource theory explains is why whether two tasks interfere can't be predicted from how hard each one is alone: two task pairs with identical combined difficulty can differ several-fold in actual interference, purely because of whether the resource dimensions they occupy overlap. A model that predicts interference from total difficulty alone, drawing on a single resource, will systematically underestimate interference for "both tasks easy but same dimension" pairs, and overestimate it for "both tasks hard but different dimensions" pairs.

Why it happens

Earlier single-resource models assumed one unified attention capacity, allocated so that harder tasks draw more of it, meaning interference should depend only on whether the two tasks' combined resource demand exceeds total capacity. But a large body of dual-task research finds that task difficulty and resource overlap contribute to interference as two relatively independent factors. Multiple resource theory splits total capacity across several dimensions while still keeping the single-resource logic within each dimension — the harder the task, the more of that dimension's capacity it takes. That combination lets the model explain both the difficulty-driven interference seen within one dimension (single-resource logic still applies there) and why cross-dimension combinations show markedly less interference than same-dimension ones (dimensions don't share supply). That's the extra layer of explanatory power multiple resource theory adds over a single-resource model.

Studying it

The evidence here is comparative: hold task difficulty fixed while varying channel overlap, or hold channel overlap fixed while varying task difficulty, and observe the resulting change in dual-task interference in each case; if the change driven by channel overlap is clearly independent of the change driven by raising difficulty alone, that supports multiple resource theory having stronger explanatory power than a single-resource model. Methodological note: this comparison requires manipulating both difficulty and channel overlap within the same experiment — manipulating only one factor can show that factor matters, but can't show which model explains the data better.

Where it stops holding

Multiple resource theory doesn't deny that a single-resource model still holds within one resource dimension — as long as two tasks sit on the same dimension, combined difficulty still predicts interference reasonably well, and single-resource logic applies within that dimension; it just can't be applied across dimensions using overall difficulty. When a task is complex enough to require frequent central decision-making and priority coordination, the resources that coordination consumes don't belong cleanly to any one dimension, and plain difficulty-based prediction regains some of its explanatory power.

Applying it

When running a dual-task usability assessment, don't use "how hard is each task" as the sole risk predictor — also assess how much the two tasks' resource dimensions overlap; scoring difficulty and dimensional overlap together predicts which task combinations will cause problems more accurately than difficulty assessment alone. Verification: rank a batch of task combinations two ways — by difficulty alone, and by difficulty combined with dimensional overlap — then check both rankings against the actual measured drop in dual-task performance; if the ranking that includes dimensional overlap tracks the measured results noticeably better, the situation genuinely calls for a multiple-resource framework rather than a difficulty-only metric.

Related

• Same group: A9.10.1 attention resources aren't one pool — they split into pools by channel, stage, and response · A9.10.2 tasks sharing a resource pool interfere more than tasks using different pools · A9.10.3 vision and hearing are separate perceptual resources, and speech and manual response are separate response resources · A9.10.5 designing parallel tasks should deliberately assign them to different resource pools to reduce interference
• Nearby: A9.06 intrinsic, extraneous and germane load (the shared-capacity model within a single task, a different question from the cross-task prediction problem discussed here)
• Search terms: single-resource model · multiple resource theory · task difficulty prediction