Task Performance and Analysis

Abstract

Task performance and analysis, which MeSH classifies under industrial psychology, is the family of methods for describing what a task requires of whoever performs it and for measuring how well they perform it. It joins two questions human-factors research treats as one: what is the structure of the work — its goals, subgoals, decisions, and actions — and what does that structure cost the operator in speed, accuracy, and mental effort? This article sets out the main techniques of task analysis, from the hierarchical decomposition of observable actions to the cognitive task analysis of the judgement behind them, and the main measures of performance and mental workload, from error rates to physiological indices. It shows why a task's demands cannot be read off its surface, and why two tasks that look equally hard can consume different attentional resources.

Keywords: task analysis, mental workload, dual-task interference

The problem is as old as organized work and as current as the cockpit warning that arrives one second too late. To design a job, a training programme, or an interface, one must first know what the task actually demands — not the idealized procedure in the manual, but the perception, decision, and action a competent person really carries out, and the load those place on finite attention. Task performance and analysis supplies both halves of that knowledge: a descriptive account of the work's structure and a quantitative account of its cost. The two are inseparable, because a change that simplifies a task's structure is justified only if it lowers the operator's load or raises their performance, and neither can be judged without measuring it.

What Task Performance and Analysis Is

Task performance and analysis is the study of the task as the unit of work: the goal-directed activity a person undertakes, described precisely enough to be designed for, trained for, and measured. In the MeSH vocabulary the heading covers both the analysis of tasks — breaking an activity into its component demands — and the measurement of how well those tasks are carried out. The two are complementary faces of a single enterprise. An analysis that never measures performance cannot say whether its account of the work is right or whether a proposed change helps; a measurement made without an analysis of the task cannot say why performance is what it is, or what to change to improve it.

The field grew from the scientific-management tradition of the early twentieth century, when Frederick Winslow Taylor and his followers first broke industrial jobs into timed elementary motions in the belief that there was one best way to perform any task (Taylor, 1911). That programme was narrow — it treated the worker as a motor system to be optimized — but it established the two ideas the modern field still rests on: that work has an analysable structure, and that performance can be quantified. What later research added was the recognition that the costly part of most tasks is not muscular but mental, so that a full account must reach past the observable motions to the perception, decision, and attentional load beneath them (Wickens et al., 2021).

Types of Task Performance and Analysis

MeSH places task performance and analysis beneath three broader descriptors — ergonomics, industrial psychology, and psychomotor performance — reflecting that the topic is at once a method of fitting work to people, a concern of the psychology of work, and a matter of skilled human performance. Beneath it in turn MeSH files three narrower descriptors, listed in Table 1. Such a classification is an indexing hierarchy rather than a theory of the field: it records how the literature is catalogued, and its three children are specialized branches rather than an exhaustive taxonomy of the many kinds of task analysis and performance measurement. The techniques that actually organize the field — hierarchical and cognitive task analysis, workload measurement, dual-task methods — cut across these categories and are treated in their own sections below. None of the three narrower descriptors has a dedicated article on this site yet.

Task Analysis: Decomposing Work

The descriptive half of the field is task analysis: the systematic decomposition of an activity into the elements a person must carry out to achieve its goal. Its most widely used form is hierarchical task analysis (HTA), developed by John Annett and colleagues in the late 1960s, which represents a task as a nested hierarchy of goals and subgoals, each broken down until the analyst reaches operations specific enough to be useful, and each governed by a plan that states the conditions under which its subgoals are carried out and in what order (Stanton, 2006). The power of HTA is that it is redescription rather than mere listing: the analyst stops decomposing a branch only when further detail would not change any decision the analysis is meant to inform, so the depth of the tree is governed by its purpose. Half a century after its introduction, HTA remains the common substrate on which more specialized methods are built, precisely because a good hierarchy of goals is neutral between the many uses — interface design, training, error prediction — to which it is put (Kirwan & Ainsworth, 1992).

HTA describes what an operator does, but for skilled work the harder question is what they must know and decide. Cognitive task analysis (CTA) answers it by recovering the perceptual cues, judgements, and strategies that expertise consists of — the parts of a task that are invisible because they happen in the practitioner's head (Crandall et al., 2006). Because experts are notoriously poor at reporting the knowledge they use fluently, CTA relies on structured interview and observation techniques that probe difficult cases and critical decisions rather than asking directly. One widely adopted toolkit, applied cognitive task analysis, packages these methods into a practitioner-usable sequence — a task diagram, a knowledge audit, a simulation interview, and a cognitive-demands table — so that a domain analyst without a psychology background can surface the cognitive demands of a job (Militello & Hutton, 1998). Together, HTA and CTA span the field's descriptive range: the observable structure of the work and the cognition that drives it.

Measuring Performance and Mental Workload

The quantitative half of the field measures how well a task is performed and what it costs. Performance itself is not one quantity but several, because an operator can trade speed for accuracy: the same person can respond faster at the price of more errors or more carefully at the price of time, so speed and accuracy must be reported together, and Fitts's information-theoretic measure of aimed movement was an early attempt to combine them into a single index of the effective rate at which a task can be done (Fitts, 1954). But two operators can reach the same performance while one is coasting and the other is at the limit of their capacity, and that difference — the spare capacity a task leaves unused — is mental workload.

Workload is measured three ways. Subjective scales ask the operator to rate the demand they felt; the most widely used, the NASA Task Load Index, has the operator rate six dimensions — mental, physical, and temporal demand, performance, effort, and frustration — and combines them into a weighted score, and it has remained a standard of the field for decades because it is quick, sensitive, and diagnostic of which demand dominates (Hart, 2006). Secondary-task methods measure the spare capacity directly, by loading the operator with a second task and seeing how much its performance suffers. And physiological measures — heart-rate variability, pupil diameter, and brain activity among them — track the bodily correlates of effort without interrupting the primary task, an approach whose promise and limits are the subject of active review (Charles & Nixon, 2019). No single measure is decisive; because workload is a construct rather than a directly observable quantity, converging evidence from the three families is what a careful assessment seeks (Longo et al., 2022).

Attentional Limits: Dual-Task Interference and Multiple Resources

Why does a second task cost anything at all? The answer defines the field's theoretical core. Kahneman's influential capacity model treated attention as a single limited pool of effort that any task draws on, so that two tasks interfere whenever their combined demand exceeds the supply (Kahneman, 1973). This single-resource view explained much, but not the striking fact that some task pairs interfere far more than their difficulty predicts while others barely interfere at all. Norman and Bobrow sharpened the picture by distinguishing resource-limited processes, whose performance improves as more attention is devoted to them, from data-limited processes, whose performance is capped by the quality of the input no matter how much effort is applied — a distinction that showed why simply measuring dual-task decrement can mislead (Norman & Bobrow, 1975).

Wickens resolved the puzzle with multiple-resource theory: attentional capacity is not one pool but several, organized along dimensions such as processing stage, perceptual modality, and the verbal-versus-spatial code of the material, so that two tasks interfere in proportion to the resources they share (Wickens, 2002). A visual-spatial task and an auditory-verbal task can be combined with little cost because they draw on separate resources; two visual-spatial tasks compete for the same one and interfere severely. The theory is quantitative enough to predict which pairings will overload an operator and which will not, and it underlies the design principle of offloading a task to an unused modality (Wickens, 2008). At the extreme, when two tasks demand the same central stage of processing, they cannot proceed in parallel at all: one must wait for the other, producing the strict bottleneck measured in the psychological refractory period — the delay in responding to a second stimulus that arrives while the response to the first is still being selected (Pashler, 1994). Neuroimaging has since localized part of this limit to a shared frontal-parietal network that is recruited by disparate tasks and can process only one demanding decision at a time (Marois & Ivanoff, 2005).

Interactive demonstrations

The three demonstrations below let the reader work with the field's core ideas. The first builds a hierarchical task analysis of an everyday task, showing how goals redescend into subgoals and operations under plans; the second computes the NASA-TLX weighted workload score from six ratings and their pairwise weights; the third pits two tasks against each other in the multiple-resource model and shows how interference rises with the resources they share.

Worked Example

Consider the workload assessment in the second demonstration, which makes the NASA-TLX concrete. The index has an operator rate a task on six subscales, each from 0 to 100, and then weight the subscales by their relevance to that task through fifteen pairwise comparisons: each of the six subscales is pitted against every other once, and a subscale's weight is the number of comparisons in which the operator judged it the more important source of workload. The fifteen comparisons distribute a total of fifteen weight points, so the weights of the six subscales always sum to 15 (Hart, 2006). The overall workload is the weighted average of the ratings, W = Σ(r_i × w_i) / 15.

Take a driver using an unfamiliar satellite-navigation system in city traffic. Suppose the ratings are mental demand 80, temporal demand 70, effort 65, performance 40, frustration 55, and physical demand 20, and that the pairwise comparisons give weights of 5, 4, 3, 2, 1, and 0 to those same six subscales — mental demand chosen every time it appeared, physical demand never. The weighted sum is 80 × 5 + 70 × 4 + 65 × 3 + 40 × 2 + 55 × 1 + 20 × 0 = 400 + 280 + 195 + 80 + 55 + 0 = 1010, so the overall workload is 1010 / 15 = 67.3. The unweighted average of the same six ratings is only (80 + 70 + 65 + 40 + 55 + 20) / 6 = 55.0.

The gap between the two numbers is the point of the weighting. Because the subscales that this driver rated highest — mental and temporal demand — are also the ones they judged most relevant to driving, weighting pulls the score up from 55.0 to 67.3: the workload is concentrated in exactly the dimensions that matter for this task. Had the high ratings fallen on subscales the operator deemed irrelevant, weighting would have pulled the score the other way. The weighted score is thus not merely more precise but more diagnostic, and the bars of Figure 1 show where the load actually comes from — a profile a single number could never convey. These values are illustrative of the procedure, not measurements from any one study.

Figure 1

Discussion

The recurring lesson of task performance and analysis is that the demands of a task are not visible on its surface. A job that looks simple can be attentionally punishing, and two tasks that take the same time can leave wholly different amounts of spare capacity, so an analysis confined to the observable procedure — the descendant of Taylor's timed motions — systematically misses what makes modern work hard (Taylor, 1911). The field's development has been a steady move inward, from the motions of the body to the structure of goals, and from there to the knowledge, decisions, and attentional load that a competent performance actually consumes (Crandall et al., 2006).

The two halves of the field need each other. A task analysis that is never validated against measured performance is a hypothesis, not a finding; a workload number reported without an analysis of the task that produced it cannot be acted on, because it does not say which demand to relieve. The strongest applications join them: an HTA or CTA identifies where the cognitive load concentrates, a workload measure confirms it, and a redesign — reallocating a subtask, offloading it to an unused modality, or simplifying the goal structure — is then justified by the improvement it produces (Wickens et al., 2021). Multiple-resource theory gives this practice its sharpest tool, because it predicts in advance which combinations of demands an operator can sustain and which will break down (Wickens, 2008).

Current Directions

Three lines of work are currently active. The first is the physiological measurement of workload. Heart-rate variability, pupillometry, electrodermal activity, and EEG promise a continuous, unobtrusive index of an operator's load, and systematic reviews are now mapping which signals track which aspects of demand and how reliably — a necessary consolidation, because the individual studies use incompatible measures and the signals are confounded by stress, physical effort, and the environment (Tao et al., 2019). The reviews converge on the view that no single physiological measure is a clean workload meter, and that the practical path is a combination of signals matched to the task (Charles & Nixon, 2019).

The second is the definition of mental workload itself, long used as though its meaning were settled. Recent work argues that the construct has been measured more than it has been defined, and proposes inclusive, testable definitions intended to make workload comparable across the subjective, performance-based, and physiological traditions that currently talk past one another (Longo et al., 2022). The third is the extension of task analysis from individuals to whole sociotechnical systems. As the units of analysis become teams, organizations, and human-automation ensembles, researchers ask whether methods built to describe one person's task are fit to describe a system's, and are adapting and testing them against a systems-thinking standard (Salmon et al., 2017). Across all three, the open problem is the field's founding one, made harder by complex modern work: to measure what a task truly demands, precisely enough to design against it.

Common Misconceptions

“A task analysis is just a list of the steps in a procedure.”

Hierarchical task analysis is a redescription of goals into subgoals under plans, not a flat list; and cognitive task analysis recovers the knowledge and judgement that never appear as steps at all (Crandall et al., 2006).

“How hard a task feels can be read from how long it takes.”

Two tasks of equal duration can impose very different mental workloads; time on task measures performance, not the spare attentional capacity the task leaves, which is why workload is assessed separately (Hart, 2006).

“People can multitask freely if they try hard enough.”

Two tasks that compete for the same processing resource interfere severely regardless of effort, and when they demand the same central stage they cannot run in parallel at all — a structural bottleneck, not a lack of will (Pashler, 1994).

“A physiological signal gives an objective measure of workload.”

No single physiological measure is a clean index of workload; the signals are confounded by stress and physical effort, and reliable assessment combines converging measures rather than trusting one (Tao et al., 2019).

Glossary

Cognitive task analysis.

A family of methods for recovering the perceptual cues, knowledge, and decisions that underlie skilled performance — the cognitive part of a task that observation of actions alone cannot reveal.

Data-limited process.

A process whose performance is capped by the quality of its input rather than by attention, so that devoting more effort to it yields no improvement.

Dual-task interference.

The loss of performance when two tasks are carried out at once, used experimentally to measure the attentional demand each task imposes.

Hierarchical task analysis.

The decomposition of a task into a nested hierarchy of goals and subgoals, each governed by a plan specifying when and in what order its subgoals are performed.

Index of difficulty.

In Fitts's law, a logarithmic combination of a movement's amplitude and its target's width that predicts the time an aimed movement takes.

Mental workload.

The demand a task places on an operator's limited attentional resources; assessed by subjective, secondary-task, and physiological measures, and distinct from the task's performance.

Multiple-resource theory.

Wickens's account that attention comprises several distinct resources defined by processing stage, modality, and code, so that two tasks interfere in proportion to the resources they share.

NASA Task Load Index.

A subjective workload scale on which an operator rates six dimensions of demand and weights them by pairwise comparison, yielding a weighted overall workload score (the NASA-TLX).

Plan.

In hierarchical task analysis, the statement attached to a goal that specifies the conditions under which its subgoals are carried out and the order in which they occur.

Psychological refractory period.

The delay in responding to a second stimulus that arrives while the response to a first is still being selected, evidence of a central processing bottleneck.

Resource-limited process.

A process whose performance improves as more attentional resources are devoted to it, in contrast to a data-limited process.

Speed–accuracy trade-off.

The lawful exchange by which an operator can respond faster only at the cost of more errors, or more accurately only at the cost of time, so that neither measures performance alone.

Task analysis.

The systematic decomposition of an activity into the goals, subgoals, decisions, and actions a person must carry out to achieve it; the descriptive half of the field.

Time and motion study.

The timing and recording of the elementary motions of a job, the founding technique of scientific management and the historical root of task analysis.

Key Researchers

Paul M. Fitts (1912–1965). American psychologist who founded the engineering-psychology study of the human motor system; Fitts's law quantifies the speed–accuracy trade-off in aimed movement. Wikipedia · Wikidata

Daniel Kahneman (1934–2024). Israeli-American psychologist and Nobel laureate whose Attention and Effort (1973) framed attention as a single limited capacity, the model against which multiple-resource theory later defined itself. Wikipedia · Wikidata

Gary A. Klein (b. 1944). American research psychologist who pioneered cognitive task analysis and naturalistic decision making, developing methods to recover the judgement behind expert performance. Wikipedia · Personal site

Harold Pashler. American cognitive psychologist whose work on dual-task interference and the psychological refractory period established the case for a central processing bottleneck. ORCID · Wikipedia · Wikidata

Neville A. Stanton. British human-factors researcher whose work developed and extended hierarchical task analysis and systems-ergonomics methods. ORCID · Wikipedia · Wikidata

Frederick Winslow Taylor (1856–1915). American engineer who founded scientific management and, with it, the timing and analysis of work; time-and-motion study is the historical root of task analysis. Wikipedia · Wikidata

Christopher D. Wickens. American engineering psychologist whose multiple-resource theory of attention and work on mental workload are central to the analysis of task demand. ORCID · Faculty page

Frequently Asked Questions

What is task analysis in simple terms?
It is a systematic way of describing what a task really requires — its goals, the subgoals they break into, and the decisions and actions a competent person carries out to achieve them. The most common form, hierarchical task analysis, draws the task as a tree of goals and subgoals, each with a plan for when its parts are done (Stanton, 2006).

How is cognitive task analysis different from ordinary task analysis?
Ordinary task analysis captures what an operator does; cognitive task analysis captures what they must know and decide — the cues they notice, the judgements they make, and the strategies they use. Because experts cannot easily report this knowledge, cognitive task analysis uses structured interviews built around difficult cases (Crandall et al., 2006).

What is mental workload?
It is the demand a task places on a person's limited attention — roughly, how much of their mental capacity the task uses up. Two tasks can take the same time yet impose very different workloads, which is why workload is measured separately from performance (Hart, 2006).

How is mental workload measured?
Three ways: subjective rating scales such as the NASA-TLX, on which the operator rates the demand they felt; secondary-task methods, which load the operator with a second task to reveal spare capacity; and physiological measures such as heart-rate variability and pupil size. Because none is decisive alone, careful assessment combines them (Longo et al., 2022).

What is the NASA-TLX?
It is the most widely used subjective workload scale. The operator rates a task on six dimensions — mental, physical, and temporal demand, performance, effort, and frustration — and weights them by pairwise comparison, giving a weighted overall score that also shows which demand dominated (Hart, 2006).

Why can't people truly multitask?
Because two tasks that draw on the same mental resource interfere with each other, and two that need the same central decision stage cannot run at once at all — one must wait for the other. This bottleneck is structural, not a matter of effort or practice (Pashler, 1994).

Why do some pairs of tasks combine more easily than others?
Multiple-resource theory holds that attention is several resources, not one, defined by processing stage, sensory modality, and whether the material is verbal or spatial. Tasks that use different resources — a spoken conversation and a visual search, say — combine with little cost, while two that use the same resource interfere badly (Wickens, 2002).

What does time-and-motion study have to do with modern task analysis?
It is its ancestor. Taylor's scientific management first broke jobs into timed elementary motions, establishing that work has an analysable structure and that performance can be quantified — the two premises the modern field still rests on, even as it moved from muscular motions to mental demands (Taylor, 1911).

References