What Is the Subjective Cost of Cognitive Effort? Load, Trait, and Aging Effects Revealed by Economic Preference

If cognitive effort feels genuinely costly — similar to a long commute or a painful injection — then people should be willing to accept less money to avoid it. If that's true, you can measure how much less. And if you can measure that, you have something the field has never had before: an objective price tag on mental work. Andrew Westbrook, Daria Kester, and Todd Braver found that this price tag varies enormously across people, and that variation has been quietly contaminating decades of cognitive research. Here's the problem they were trying to solve. Cognitive science has long assumed that people treat mental effort as costly — psychologists call us "cognitive misers." But that cost is subjective and differs across individuals in ways that almost no study controls for. When someone performs poorly on a demanding task, we tend to conclude they have less cognitive ability. But what if they simply find the task more aversive? A person who looks impaired might just be unmotivated. Someone who looks unusually capable might simply be unusually willing to engage. Westbrook and colleagues argue this isn't a marginal concern — it's a central gap. For clinical groups marked by low motivation, or for older adults, the confound is especially dangerous. The field has tried to handle this with self-report scales like the NASA Task Load Index or the Need for Cognition Scale. But self-reports are indirect. Asking someone what their effort is "worth" doesn't prove they'll actually trade that effort for money when it counts. So Westbrook and colleagues borrowed a tool from economics. The logic is called revealed preferences: instead of asking people what they value, you infer it from the choices they actually make. Their paradigm, which they call Cognitive Effort Discounting, or COG-ED, pits a low-effort task paired with a smaller monetary reward against a higher-effort task paired with a larger reward. The low-effort offer is adjusted trial by trial until the participant is indifferent between the two options. That indifference point — the exact dollar amount where hard and easy feel equivalent — is your measure. It's a continuous, incentivized, behaviorally grounded number. The high-effort task was the N-back, a working memory probe where you monitor a sequence of items and respond whenever the current item matches the one N steps back. Westbrook and colleagues varied N to parametrically increase cognitive load. In the main version of the paradigm, participants chose between the one-back for a smaller, variable reward and harder N-back levels for a fixed two-dollar offer. After six choices per level, with adjustments halving each time, the final indifference point was precise to about a cent and a half. Payoffs were real money — participants were paid for a randomly selected choice, which they then completed four additional times. The within-person results were clean. As N-back load increased, the subjective value of the harder option dropped in a roughly linear fashion. For younger adults, the effect was large — a partial eta-squared of 0.46. For older adults, it was smaller but still reliable. Put concretely: for a two-dollar offer at the four-back, younger adults' mean subjective value was about 98 cents. Older adults' mean was about 40 cents. A younger adult needed roughly an extra dollar to choose the four-back over the one-back; an older adult needed an extra dollar sixty. What makes this finding sharp is that the load effect held even after statistically controlling for task performance. When multilevel models included both objective N-back load and signal-detection accuracy — known as d-prime — load remained a reliable negative predictor of subjective value, while better performance predicted higher willingness to choose the harder task. That's an important separation. Higher N-back levels aren't just harder to do; they're subjectively costlier, above and beyond errors or response slowing. The aversiveness of mental effort is partly independent of whether you're any good at it. Now to the between-person story, which is where things get consequential. Westbrook and colleagues tested two groups in two experiments: younger adults with a mean age around 22, and older adults with a mean age around 75 to 78. Across both experiments, older adults discounted cognitive effort more steeply. Area under the curve — a summary metric where higher values mean less discounting — averaged around 0.77 to 0.80 for younger adults and 0.43 to 0.51 for older adults, depending on reward amount. The age difference held even when controlling for performance. The authors provide a telling comparison: older adults performing the three-back achieved a mean d-prime of 1.74, statistically indistinguishable from younger adults' performance on the four-back, at 1.65. Yet subjective value was reliably lower for the older adults on that matched-difficulty comparison. Same objective competence, higher subjective cost. Trait cognitive engagement also mattered. People who score higher on the Need for Cognition Scale — who report genuinely enjoying mental challenge — discounted effort less, and their COG-ED scores predicted their Need for Cognition ratings. But here's the telling detail: when Need for Cognition was statistically controlled, age still explained a significant and large additional chunk of variance in effort discounting — a delta R-squared of 0.27. So trait engagement doesn't account for the age effect. Older adults aren't just less curious; something more specific is raising the subjective cost of thinking as people age. And critically, COG-ED outperformed self-report in capturing these differences. Effort discounting area under the curve predicted age group with a p-value below 0.01. The NASA Task Load Index did not predict age at all, with a p-value of 0.35. That's the methodological case for the paradigm made in a single comparison. Westbrook and colleagues then ran a benchmark experiment, asking how COG-ED relates to delay discounting — the canonical behavioral-economic measure of how much people devalue rewards that arrive later. The two paradigms shared a structure, and the team administered both to the same participants. The results showed convergence and dissociation. On the convergence side, both measures showed the same classic amount effect: larger rewards were discounted less. For delay discounting, twenty-five-thousand-dollar offers were discounted less than one-thousand-dollar offers in both age groups. For cognitive effort, five-dollar offers were discounted less than one-dollar offers, reliably so in older adults. The two measures also correlated: across 33 participants, the relationship had a coefficient of 0.30 and an R-squared of 0.14. Modest shared variance — they're related, but not the same thing. The dissociation is equally informative. Income predicted delay discounting but not effort discounting. Age predicted effort discounting but not delay discounting once income was accounted for. A significant age-by-domain interaction emerged when income was included. These measures overlap but are not interchangeable — COG-ED captures domain-specific variance that standard delay discounting misses, particularly the variance tied to aging. That's what validates it as a distinct tool rather than a repackaging of something already in the literature. What changes if you adopt this framework? The ability-motivation confound becomes measurable rather than assumed. Instead of inferring capacity from performance alone, researchers can measure subjective effort cost and treat it as an explicit variable. Westbrook and colleagues point to age-related changes in dopaminergic function as one plausible mechanism — computational models link tonic dopamine to perceived average reward and vigor, and both animal and human work implicates dopamine in effort-based decisions. Whether reduced dopaminergic signaling or greater compensatory recruitment underlies higher subjective cost in older adults is now a testable question, not speculation. The authors also flag implications for clinical populations marked by anergia or avolition — depression being the obvious example — where apparent cognitive deficits might reflect motivational rather than capacity differences. They note a convergent early signal: informal post-task reports of "fair pay" averaged $1.89 for more demanding conditions and $0.89 for less demanding ones, landing squarely in the range COG-ED now measures formally. The core contribution is precise: a continuous, incentivized, behaviorally grounded metric for something researchers have long had to assume. The price of thinking varies. Now you can measure it. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.