Multiple Resource Theory¶

Status: emerging
Last updated: 2026-06-01
Sources: 14639220210123806.Pdf
Tags: [multiple-resource-theory, mental-workload, attention, time-sharing, dual-task, multitasking, performance-prediction, cognitive-foundations]

Summary¶

Multiple resource theory (MRT) explains why people can sometimes perform two tasks at once with little interference and sometimes cannot. Wickens (2002) sets out the theory and its four-dimensional model, which holds that two tasks interfere more to the extent that they draw on the same processing resources, defined along four dimensions: processing stages (perceptual/cognitive vs response), perceptual modalities (auditory vs visual), visual channels (focal vs ambient), and processing codes (spatial vs verbal). The theory supports a computational model that predicts the relative interference of task pairs from their demand levels and the conflict between the resources they share. MRT is the mechanism behind much of the dual-task account in Information Processing and the resource view of demand in Mental Workload.

Body¶

Context¶

Wickens (2002), in Multiple resources and performance prediction, describes the origins and structure of multiple resource theory and presents the four-dimensional model he first proposed in 1980. The paper traces MRT from the early "single-channel bottleneck" conception of human information processing to a model in which several relatively independent resource pools support concurrent performance. Within this knowledge base the article isolates a mechanism that two existing articles depend on but do not develop: it gives the resource account underlying the multitasking section of Information Processing and the limited-resource premise of Mental Workload (where MRT is already noted via Wickens 1984, 2008), and it connects to the time-sharing concerns of Situation Awareness and Aviation Human Factors. It is held at emerging status as a single-source article.

Key Points¶

Wickens (2002) frames MRT as a theory of multiple-task performance whose value lies in the overload situation, where an operator must perform two or more tasks at once. He distinguishes the two parts of the name: "resources" connotes something limited and allocatable, so that a task left with insufficient residual resources performs worse, while "multiple" denotes the existence of several separate pools rather than one. Workload aligns with the resource aspect but not the multiple aspect; MRT is therefore not purely a theory of attention nor of workload (PDF pp. 1–2, orig. pp. 159–160). He situates resources on a continuum from resource-limited to data-limited processing (Norman and Bobrow 1975) and traces the theory's origin to the relaxation of the single-channel bottleneck (PDF pp. 2–3, orig. pp. 160–161).

The core of the paper is the four-dimensional model. Two tasks conflict more when they share levels on any of four dimensions: processing stages, separating perceptual and cognitive activity from response selection and execution; perceptual modalities, where cross-modal pairings (auditory with visual) time-share better than intra-modal pairings (PDF pp. 5–7, orig. pp. 163–165); visual channels, where focal vision (foveal, for fine detail and pattern recognition) and ambient vision (peripheral, for orientation and movement) behave as separate resources, as when a driver keeps the car in lane by ambient vision while reading a sign with focal vision (PDF pp. 7–9, orig. pp. 165–167); and processing codes, separating spatial from verbal/linguistic processing, which predicts, for example, the danger of manual phone dialling while steering because both compete for spatial code (PDF pp. 9–11, orig. pp. 167–169).

Wickens (2002) then renders the model computationally. Predicted interference for a task pair combines two components: a demand component that grows with the difficulty (resource demand) of the time-shared tasks, and a conflict component that penalises the pair in proportion to the degree of resource overlap read from a conflict matrix, with overlap greatest when tasks share dimensions and both demand focal processing (PDF pp. 12–14, orig. pp. 170–172). The model emphasises the multiple aspect over precise quantitative demand values, so its output is meaningful as a comparison across candidate task combinations rather than as a single absolute number (PDF pp. 13–14, orig. pp. 171–172). He closes by outlining three challenges to the model — coding the demand of a task, representing resource allocation and prioritisation policy between tasks, and resolving whether focal and ambient vision are genuinely separate resources or whether ambient processing is effectively preattentive (PDF pp. 7–8, 15–16, orig. pp. 165–166, 173–174).

Conclusion¶

Wickens (2002) concludes that multiple resource theory predicts dual-task interference from the overlap of the resources two tasks demand, and that its four-dimensional model is most useful for the comparative judgement a designer needs: which pairing of tasks, displays, or modalities will interfere least. The computational rendering makes those comparisons explicit while the author cautions against over-reading any single numerical output. The remaining challenges — demand coding, allocation policy, and the status of ambient vision — mark where the theory's predictions are least settled.

References¶

Wickens, C.D. (2002) 'Multiple resources and performance prediction', Theoretical Issues in Ergonomics Science, 3(2), pp. 159–177. doi: 10.1080/14639220210123806. wickens2002multipleresources

Norman, D.A. & Bobrow, D.G. (1975) 'On data-limited and resource-limited processes', Cognitive Psychology, 7(1), pp. 44–64. To be validated.

Open Questions¶

Are focal and ambient vision genuinely separate resources, or is ambient processing effectively preattentive and resource-free (Wickens, 2002)?
How should resource allocation and task-prioritisation policy be represented when predicting interference for tasks of unequal priority?
How well do the model's relative interference predictions hold for three or more concurrent tasks, as in supervisory monitoring of multiple automated systems (see Mental Workload, Supervisory Control Of Automation)?