Efficiency

Abstract

Efficiency, which MeSH classifies under industrial psychology, is the ratio of what a cognitive system achieves to the resources it expends achieving it — performance divided by effort, time, or energy rather than performance alone. The concept lets psychology separate two people or two conditions that reach the same result, because the one that reaches it with less mental cost is the more efficient. Three research traditions give the idea empirical teeth: the neural-efficiency hypothesis, which finds that abler brains often expend less metabolic activity on a task; processing-efficiency theory, which shows that anxiety can drain efficiency while leaving raw performance intact; and instructional efficiency, which combines test performance with rated mental effort into a single computable score. Modern effort-based accounts recast efficiency as the outcome of a cost-benefit decision the brain makes about how much control to deploy.

Keywords: efficiency, mental effort, neural efficiency, cognitive load

Two students score the same on a test. One has spent the evening relaxed and unhurried; the other has strained to the edge of exhaustion. On the outcome measure they are identical, yet almost anyone would judge the first to have performed the better — not because the answers differ but because the cost of producing them does. This intuition is what the psychology of efficiency makes precise. Efficiency is not a measure of how well a task is done but of how much it costs to do it that well, and treating the two separately turns out to matter across the field: in how intelligence relates to brain activity, in how anxiety undermines thinking, in how instruction should be designed, and in how the brain decides whether a demanding task is worth the effort at all (Hoffman & Schraw, 2010).

What Efficiency Is

Efficiency is a relational quantity: it makes no sense to call a performance efficient in isolation, only efficient relative to the resources it consumed. In MeSH the descriptor Efficiency is defined as the ratio of output to effort or resources, and it is filed at once under the psychology of industry and under organization and administration, reflecting its double life as a psychological construct and a management one. In cognitive psychology the resource in the denominator is usually mental effort, time, or the metabolic cost of neural work, and the output is task performance. The single most important consequence of this framing is that outcome and cost can move independently: a manipulation can leave accuracy untouched while doubling the effort required, and only an efficiency measure will detect it.

The foundation for treating cognitive resources as limited and allocable was laid by the distinction between data-limited and resource-limited processes. A process is resource-limited when giving it more mental resources improves its performance, and data-limited when performance is capped by the quality of the input no matter how much effort is poured in (Norman & Bobrow, 1975). This distinction is what makes efficiency measurable: only in the resource-limited regime does the trade-off between effort and output exist to be quantified. It also carries a warning that recurs throughout the field — that a raw performance score confounds the difficulty of the data with the resources the person supplied, and disentangling them requires measuring the resources directly. Whether “efficiency” as inferred from brain imaging is even a coherent single construct has been questioned on exactly these grounds, since the same behavioral output can arise from very different underlying resource profiles (Poldrack, 2015).

Types of Efficiency

Efficiency is a broad descriptor, and MeSH places one narrower kind directly beneath it in the tree. That child sits under the same parent construct — efficiency as a ratio of output to resource — but specializes it to a particular domain. It is worth stressing that the MeSH hierarchy is an indexing classification built for retrieving literature, not a theory of the mind, so the placement records how the biomedical literature is organized rather than asserting a strict conceptual containment; the psychological senses of efficiency discussed in this article are largely orthogonal to this administrative subdivision. Table 1 lists the direct MeSH child of Efficiency.

Neural Efficiency

The most striking application of the efficiency idea to individual differences is the neural-efficiency hypothesis: the proposal that more intelligent brains work less hard to accomplish the same thing. The founding observation came from positron emission tomography, which measured cerebral glucose metabolism while participants worked on an abstract-reasoning task and found that higher-scoring individuals showed lower metabolic rates — the abler brain expended less energy for the better result (Haier et al., 1988). The finding inverted a natural assumption that smarter meant harder-working neural tissue, and it recast intelligence partly as a matter of doing more with less.

Two decades of replication and qualification followed, synthesized in a comprehensive review that confirmed the effect is real but conditional (Neubauer & Fink, 2009). Neural efficiency appears most reliably for tasks of low to moderate difficulty, in the frontal cortex, and it interacts with sex and with the nature of the task; it is not a blanket property of the abler brain. Crucially, the relationship reverses when a task becomes hard enough: on the most demanding problems, more able individuals invest more neural resources, not fewer, because they are still engaged where others have disengaged or failed. A direct manipulation of task demand confirmed this crossover, showing that whether higher ability predicts lower or higher activation depends systematically on how difficult the task is relative to the person's capacity (Dunst et al., 2014). Efficiency, in short, is not a fixed trait of a brain but a description of how that brain matches its resource investment to the demands it faces.

Processing Efficiency and Anxiety

If neural efficiency concerns how much resource a brain spends, processing-efficiency theory concerns what happens when something siphons that resource away. Its central and much-replicated claim is that anxiety impairs the efficiency of cognitive processing — the relationship between performance and the effort required to achieve it — more, and earlier, than it impairs effectiveness, the quality of performance itself (Eysenck & Calvo, 1992). An anxious person burdened with worrying thoughts can often still reach the correct answer, but only by recruiting extra effort and compensatory strategies to offset the worry consuming their working memory. Measured on outcome alone, they look unimpaired; measured on efficiency, the cost of the worry becomes visible. This distinction between effectiveness and efficiency is the theory's enduring contribution, and it is precisely the outcome-versus-cost separation that defines the study of efficiency generally.

The theory was later extended and sharpened into attentional control theory, which specifies the mechanism: anxiety disrupts the balance between a goal-directed attentional system and a stimulus-driven one, impairing the central-executive functions of inhibition and shifting (Eysenck et al., 2007). Because these executive functions are what allocate and protect working memory resources, anxiety's drain on efficiency is traced to a specific failure of attentional control rather than a general dulling of ability. The account explains why anxious individuals are especially impaired on tasks that stress inhibition and task-switching, and why they compensate through increased effort when they can — a direct, mechanistic story about where cognitive efficiency is lost.

Automaticity and the Limits of Resources

A second route to efficiency is not to have more resources but to need fewer, and the classic demonstration of this is the shift from controlled to automatic processing. In a landmark program of work, consistent practice on a search task — where a target was always a target and never appeared as a distractor — produced processing that became fast, effortless, and largely immune to the number of items to be searched, whereas inconsistent mapping kept processing slow, effortful, and capacity-limited (Schneider & Shiffrin, 1977). Automatic processing is the limiting case of efficiency: a skill practiced to automaticity consumes almost no central resources, runs in parallel with other tasks, and no longer competes for the limited pool that effortful processing draws on. The empirical signature of this transition is the power law of practice — the time to perform a task falls as a power function of the number of practice trials, with large early gains that taper as the skill nears its efficient limit. An influential account traces this to a change of mechanism rather than mere speedup: automatization is the replacement of a slow, resource-hungry algorithm by fast, direct retrieval of past solutions from memory, so that performance accelerates precisely because it stops computing and starts remembering (Logan, 1988).

This is why efficiency and practice are inseparable. The resources freed by automating a component skill become available for higher-level work, which is why an expert reader can attend to meaning while decoding runs itself, and why a novice, spending resources on the mechanics, cannot. The trade-off between speed and accuracy that governs reaction time is itself an efficiency frontier: within a task, one can convert time into accuracy and back, and the shape of that curve reflects how efficiently evidence is being accumulated toward a decision (Heitz, 2014). Automatization moves the whole frontier outward, delivering more accuracy per unit of time and effort, and selective allocation of the freed resources is the business of selective attention.

Instructional Efficiency and Cognitive Load

The efficiency idea found its most practical home in education, through the marriage of cognitive-load theory and a computable efficiency measure. Cognitive-load theory holds that working memory is severely limited and that instruction succeeds or fails by how well it manages the load it imposes: learning is best when instruction minimizes load extraneous to the material and directs the learner's limited resources toward building schemas (Sweller, 1988). Two instructional methods might produce identical test scores while imposing very different loads, and the one that teaches as well with less effort is plainly the better method — but only an efficiency measure can tell them apart. This connects the design of instruction directly to learning and to problem solving, where load management is decisive.

The tool that made this comparison routine combines standardized test performance with standardized rated mental effort into a single instructional-efficiency score, defined so that high performance achieved with low effort counts as high efficiency and low performance bought with high effort counts as low (Paas & Van Merriënboer, 1993). Placing conditions in a performance-by-effort space, the measure reads efficiency as the perpendicular distance from the line where effort and performance are balanced. The construct was later revisited and refined — its assumptions about how to standardize and combine the two measures examined critically, and alternatives proposed for when effort should be measured and how it should be interpreted (Van Gog & Paas, 2008). The broader lesson, drawn across learning and problem solving, is that efficiency is not one thing but a family of ratios whose right definition depends on which resource is scarce and which output is valued (Hoffman & Schraw, 2010).

Interactive demonstrations

The three demonstrations below make the core ideas manipulable. The first builds the neural-efficiency relationship, letting ability and task difficulty vary while the predicted brain activation responds — including the crossover where the abler brain begins to invest more, not less. The second is the instructional-efficiency measure itself: place a condition in the performance-by-effort plane and read its efficiency as distance from the balance line. The third traces the power law of practice, showing how reaction time falls with repetition as a skill automatizes and its resource cost collapses.

Neural efficiency and its crossover

The neural-efficiency hypothesis holds that abler brains use less activation for a given task — but only up to a point. Raise the task difficulty past the midpoint and the relationship reverses. Set ability and difficulty and watch the predicted activation.

Note. A stylized model of the demand-dependent neural-efficiency effect: below the difficulty midpoint, activation falls as ability rises; above it, the slope reverses. Illustrative of the mechanism, not calibrated to a dataset. Deterministic; no random sampling.

The instructional-efficiency measure

Paas and Van Merriënboer combined standardized performance and standardized mental effort into one score, E = (z_P − z_E)/√2, the perpendicular distance from the balance line. Set a condition and read its efficiency.

Note. Points above the dashed line (performance exceeding its effort cost) are efficient; points below it are inefficient. The worked example in the text sets z_P = 0.8 with z_E = −0.5 (E = 0.92) and z_E = 0.7 (E = 0.07). Deterministic; no random sampling.

The power law of practice

As a skill is practiced it automatizes, and its time cost falls as a power function of the number of trials, RT = a · n^−b. Move through the trials and watch reaction time drop steeply at first, then level off.

Note. With a = 800 ms and b = 0.3, reaction time at trial 1 is 800 ms — a 0% reduction from the first trial as the freed resources make the skill effortless. Deterministic; no random sampling.

Worked Example

The instructional-efficiency measure shows most clearly why efficiency must be computed rather than read off performance. Suppose an experiment compares two ways of teaching the same material and, after training, measures both test performance and self-rated mental effort. To combine them, each raw score is converted to a z-score against the pooled mean and standard deviation of the whole sample, so that performance and effort are on a common standardized scale. Efficiency is then defined as E = (z_P − z_E) / √2, the signed perpendicular distance of the point (z_E, z_P) from the diagonal line z_P = z_E, where performance exactly matches its effort cost.

Consider two conditions that produce identical standardized performance, z_P = 0.8. In condition A the learners reported low effort, z_E = −0.5; in condition B they reported high effort, z_E = 0.7. Their efficiencies are E_A = (0.8 − (−0.5)) / √2 = 1.3 / 1.4142 = 0.919, and E_B = (0.8 − 0.7) / √2 = 0.1 / 1.4142 = 0.071. The two conditions are tied on the outcome measure that a test score alone would report, yet condition A is more than ten times as efficient, a gap of 0.849 in efficiency units, because it bought the same performance far more cheaply. An instructor comparing only test scores would call the methods equivalent; the efficiency measure reveals that one imposes far less cost for the same result. The figure below plots both conditions in the performance-by-effort plane, and the values are re-derived in code and match the second demonstration.

Discussion

Across these traditions a single logic holds: efficiency separates what a system achieves from what the achievement costs, and that separation repeatedly reveals structure that outcome measures hide. Neural-efficiency research found that ability shows up in the metabolic cost of thinking, not only in its success (Haier et al., 1988); (Neubauer & Fink, 2009). Processing-efficiency theory found that anxiety attacks cost before it attacks accuracy (Eysenck & Calvo, 1992). Cognitive-load research found that two equally effective lessons can differ enormously in the effort they demand, and built a measure to catch the difference (Paas & Van Merriënboer, 1993). In each case the payoff of measuring efficiency rather than performance is the same: a finer-grained, more actionable picture of the mind at work.

Two cautions temper the picture. The first is conceptual: efficiency inferred from brain activity is not guaranteed to be a single, well-defined quantity, because identical behavior can rest on different resource profiles and lower activation need not mean better functioning (Poldrack, 2015). The second is methodological: because efficiency is a ratio of two measured things, it inherits the measurement error of both, and how the numerator and denominator are standardized and combined can change the conclusion (Van Gog & Paas, 2008). Efficiency is therefore best treated not as a property to be read off a single number but as a relationship to be modelled, with the resource in the denominator named explicitly.

Current Directions

The most active current work reframes efficiency as the product of a decision rather than a fixed capacity. Neuroeconomic accounts treat mental effort as a cost the brain weighs against expected reward, so that how much effort a person deploys — and hence how efficient they appear — reflects a rational cost-benefit computation about whether the demanding option is worth it (Westbrook & Braver, 2015). This idea is formalized in the expected value of control, which holds that the brain sets its level of cognitive control to maximize the expected payoff net of the intrinsic cost of control itself, giving a principled reason why effort is conserved and efficiency pursued (Shenhav et al., 2017). On this view the sensation that mental work is costly is not an incidental limitation but a signal that keeps a limited resource from being squandered (Kool & Botvinick, 2018).

A parallel line asks why the resource is limited at all. Resource-rational analysis proposes that human cognition is best understood as making optimal use of finite computational resources, so that many apparent shortcomings are the efficient response of a bounded system rather than failures of it (Lieder & Griffiths, 2020). Complementary work argues that the very constraints on the capacity for cognitive control may themselves be adaptive — that a system able to pursue too many goals at once would interfere with itself, so limited capacity is a rational design feature that protects efficient, non-conflicting performance (Musslick & Cohen, 2021). Integrative models now attempt to unite the physiological, motivational, and computational sides of effort into a single account of how the brain regulates its own expenditure (Andre et al., 2019).

Common Misconceptions

“Efficiency is the same as good performance.”

Efficiency is performance divided by cost, so two people with identical scores can differ greatly in efficiency; measuring outcome alone hides exactly the cost difference that efficiency is meant to capture (Eysenck & Calvo, 1992).

“A more able brain always works less hard.”

Neural efficiency holds mainly for easy-to-moderate tasks; when demands are high the relationship reverses, and abler individuals invest more neural resources, not fewer (Dunst et al., 2014).

“Lower brain activation means better cognition.”

Efficiency inferred from imaging is not a guaranteed single quantity; the same behavior can rest on different resource profiles, so lower activation cannot simply be equated with superior functioning (Poldrack, 2015).

“Feeling that thinking is effortful is just a limitation.”

Modern accounts treat the cost of mental effort as a functional signal in a cost-benefit computation, protecting a limited resource rather than merely marking a deficiency (Kool & Botvinick, 2018).

Glossary

Attentional control theory.

The successor to processing-efficiency theory, holding that anxiety impairs the goal-directed attentional system and the executive functions of inhibition and shifting, which is where its drain on cognitive efficiency arises.

Automatic processing.

Fast, effortless processing developed through consistent practice that consumes almost no central resources and runs in parallel with other tasks; the limiting case of cognitive efficiency.

Cognitive load.

The demand an activity places on the limited capacity of working memory; instruction is efficient when it minimizes load extraneous to the material to be learned.

Controlled processing.

Slow, effortful, capacity-limited processing that requires attention and competes for a limited resource pool; the effortful counterpart to automatic processing.

Data-limited process.

A process whose performance is capped by the quality of the input, so that adding mental resources does not improve it; contrasted with a resource-limited process.

Effectiveness.

The quality of task performance considered on its own, without reference to the resources spent; distinguished from efficiency, which relates performance to its cost.

Efficiency.

The ratio of output to the effort, time, or resources expended producing it; the central construct relating what a cognitive system achieves to what the achievement costs.

Expected value of control.

A theory that the brain sets its level of cognitive control to maximize expected reward net of the intrinsic cost of control, giving a principled account of when effort is worth deploying.

Instructional efficiency.

A computed score combining standardized test performance with standardized rated mental effort, so that a method teaching as well with less effort registers as more efficient.

Mental effort.

The amount of cognitive resource actually allocated to a task, often measured by self-report or physiological indices; the denominator in most cognitive efficiency measures.

Neural efficiency.

The hypothesis that more able individuals expend less brain activation for a given task, reliably so on easy-to-moderate tasks but reversing when demands are high.

Power law of practice.

The regularity that the time to perform a task falls as a power function of the number of practice trials, describing how a skill grows more efficient as it automatizes.

Processing-efficiency theory.

Eysenck and Calvo's account that anxiety impairs the efficiency of processing — the effort needed to perform — more, and earlier, than it impairs the effectiveness of performance itself.

Resource-limited process.

A process whose performance improves as more mental resources are allocated to it; the regime in which the trade-off between effort and output, and hence efficiency, can be measured.

Resource-rational analysis.

The framework treating cognition as the optimal use of finite computational resources, so that many apparent shortcomings are the efficient response of a bounded system.

Speed-accuracy trade-off.

The within-task exchange of response time for correctness; its shape reflects how efficiently evidence is accumulated toward a decision, and practice moves the whole frontier outward.

Key Researchers

Michael W. Eysenck. British psychologist who, with Manuel Calvo, originated processing-efficiency theory and later attentional control theory, establishing the effectiveness-versus-efficiency distinction for anxiety and performance. Wikipedia

Richard J. Haier. American neuroscientist whose 1988 positron-emission-tomography study launched the neural-efficiency hypothesis of intelligence. Faculty page · Wikipedia

Daniel Kahneman (1934–2024). Nobel-laureate psychologist whose two-systems account frames the trade-off between effortless automatic processing and costly controlled processing that underlies cognitive efficiency. Wikipedia · Wikidata

Aljoscha C. Neubauer. Austrian differential psychologist and leading synthesist of the neural-efficiency hypothesis, mapping how it depends on task demand, ability, and sex. ORCID · Faculty page · Wikipedia

Fred Paas. Dutch educational psychologist who, with Jeroen van Merriënboer, devised the computational instructional-efficiency measure combining mental effort with performance. ORCID · Google Scholar · Faculty page

Amitai Shenhav. American cognitive neuroscientist whose expected-value-of-control theory gives a rational account of how the brain decides how much costly cognitive control to deploy. Faculty page

John Sweller. Australian educational psychologist who originated cognitive-load theory, the framework in which working-memory efficiency during learning is analyzed. ORCID · Faculty page · Wikipedia

Frederick Winslow Taylor (1856–1915). American engineer and founder of scientific management, the historical root of efficiency as a measurable industrial construct and the branch under which MeSH files the descriptor. Wikipedia · Wikidata

Frequently Asked Questions

What is efficiency in cognitive psychology?
Efficiency is the ratio of what a cognitive system achieves to the resources — effort, time, or energy — it spends achieving it. Two performances equal on outcome can differ in efficiency because the more efficient one costs less to produce (Hoffman & Schraw, 2010).

How does efficiency differ from effectiveness?
Effectiveness is the quality of performance considered alone; efficiency relates that performance to its cost. Anxiety, for example, can harm the efficiency of thinking — the effort a result demands — well before it harms the result itself (Eysenck & Calvo, 1992).

What is the neural-efficiency hypothesis?
It is the finding that more able individuals often use less brain activation for a given task. It holds reliably on easy-to-moderate tasks but reverses on very hard ones, where abler people invest more neural resources (Haier et al., 1988); (Dunst et al., 2014).

How is instructional efficiency measured?
Test performance and rated mental effort are each standardized and combined into one score, so a method that teaches as well while demanding less effort registers as more efficient than an equally effective but costlier method (Paas & Van Merriënboer, 1993).

Why does practice make a task more efficient?
Consistent practice shifts a task from controlled to automatic processing, which is fast, effortless, and largely free of capacity limits. The resources this frees become available for higher-level work (Schneider & Shiffrin, 1977).

Is a less active brain always a better one?
No. Efficiency inferred from imaging is not a guaranteed single quantity, because identical behavior can rest on different resource profiles, so lower activation cannot simply be equated with superior cognition (Poldrack, 2015).

Why does mental effort feel costly?
Modern accounts treat the felt cost of effort as a functional signal in a cost-benefit computation: the brain weighs the value of a demanding task against the intrinsic cost of the control it requires, conserving a limited resource (Shenhav et al., 2017); (Westbrook & Braver, 2015).

Why is cognitive capacity limited in the first place?
One influential answer is that the limit is adaptive: a system able to pursue unlimited goals at once would interfere with itself, so bounded capacity is a rational design feature that protects efficient, non-conflicting performance (Musslick & Cohen, 2021).

References