CoChat

Neurocomputational mechanisms underlying subjective valuation of effort costs

Open access
Chong, Trevor T.-J., Apps, Matthew, Giehl, Kathrin, Sillence, Annie, Grima, Laura L., Husain, Masud · 2017 · PLOS Biology

The DOI matches its Crossref record; 80 of the 80 references it cites were found in a registry or an index; 32 of the 112 sources its claims cite were checked for support.

Abstract

In everyday life, we have to decide whether it is worth exerting effort to obtain rewards. Effort can be experienced in different domains, with some tasks requiring significant cognitive demand and others being more physically effortful. The motivation to exert effort for reward is highly subjective and varies considerably across the different domains of behaviour. However, very little is known about the computational or neural basis of how different effort costs are subjectively weighed against rewards. Is there a common, domain-general system of brain areas that evaluates all costs and benefits? Here, we used computational modelling and functional magnetic resonance imaging (fMRI) to examine the mechanisms underlying value processing in both the cognitive and physical domains. Participants were trained on two novel tasks that parametrically varied either cognitive or physical effort. During fMRI, participants indicated their preferences between a fixed low-effort/low-reward option and a variable higher-effort/higher-reward offer for each effort domain. Critically, reward devaluation by both cognitive and physical effort was subserved by a common network of areas, including the dorsomedial and dorsolateral prefrontal cortex, the intraparietal sulcus, and the anterior insula. Activity within these domain-general areas also covaried negatively with reward and positively with effort, suggesting an integration of these parameters within these areas. Additionally, the amygdala appeared to play a unique, domain-specific role in processing the value of rewards associated with cognitive effort. These results are the first to reveal the neurocomputational mechanisms underlying subjective cost-benefit valuation across different domains of effort and provide insight into the multidimensional nature of motivation.

Overview

CoChat Summary

TL;DR

Using computational modeling and fMRI, this study found that subjective devaluation of rewards by cognitive and physical effort engages largely overlapping domain-general valuation regions, but the right amygdala uniquely tracks cognitive-effort value. Cognitive and physical effort discounting were best described by different functions (hyperbolic vs. parabolic), suggesting partly separable valuation mechanisms.

Key findings

  • A domain-general valuation network—dorsal ACC/dmPFC, right dlPFC, intraparietal sulcus, and right anterior insula—tracked subjective value (SV) for both cognitive and physical effort in conjunction analyses (whole-brain FWE-corrected p < .05). Activity in many of these regions covaried negatively with reward and positively with effort.
  • The right amygdala showed domain-specific sensitivity to cognitive over physical effort valuation (MNI 22, −12, −10; Z = 4.11; cluster pFWE = .04); the reverse contrast (physical > cognitive) revealed no suprathreshold amygdala activity.
  • The winning computational model used separate discounting parameters and separate softmax βs for the two domains: cognitive effort was best fit by a hyperbolic function and physical effort by a parabolic function. Cognitive and physical discounting parameters were not significantly correlated (ρ = 0.22, p = .20), and softmax βs did not differ significantly (βcog = 10.29 [SE 3.87]; βphys = 6.82 [SE 3.04]; t(33) = 0.84, n.s.).
  • Effort negatively predicted choice (cognitive: t = −1.92, Z = −4.69, p < .0001; physical: t = −1.86, Z = −4.59, p < .0001), reward positively predicted choice (cognitive: t = 1.90, Z = 4.88, p < .0001; physical: t = 1.81, Z = 4.78, p < .0001), while training reinforcement rates did not (cognitive p = .28; physical p = .94).

Methods

  • 38 healthy right-handed participants were recruited; 4 were excluded (2 for missing >9% of scanned trials, 2 for head motion >5 mm), leaving 34 participants (23 females; mean age 24 y, range 19–39).
  • Participants were trained outside the scanner on a cognitive RSVP task with 1–6 attentional switches and a physical dynamometer task at 8%, 13%, 18%, 23%, 28%, and 33% of individually calibrated maximal voluntary contraction. During fMRI, they chose between a fixed low-effort/low-reward baseline and a variable higher-effort/higher-reward offer (2, 4, 6, 8, or 10 credits; effort levels 2–6) separately for each domain; choices were not executed in the scanner.
  • Model-based fMRI fitted linear, hyperbolic, parabolic, and exponential discount functions using softmax and maximum likelihood estimation, compared 36 models with AIC/BIC, and used the SV difference between chosen offer and baseline as a parametric regressor at offer cue onset in SPM8 GLMs; whole-brain FWE-corrected p < .05 and an amygdala small-volume correction were used.

Limitations

  • The paper states that the amygdala domain-specificity finding deserves replication and would be more compelling if demonstrable at a whole-brain level; it also states that future studies should verify the conclusions using cognitive and physical effort tasks best described by identical discounting functions, and that the exploratory single-function analyses were based on models that may not have been the most appropriate fits.
  • Evident from the text: the final sample was limited to young, healthy, right-handed adults (n = 34; mean age 24; 23 females), so generalizability to other populations is unclear.
  • Evident from the text: participants did not execute their chosen effort while being scanned, and the specific tasks (RSVP attention switching and grip-force exertion) may limit generalizability to other cognitive or physical effort domains.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1371/journal.pbio.1002598

Cites

80 of 80 references verified0 citation errors · 0 not checkable

Statistics

12 tests recomputed · 0 with the reported and recomputed p on opposite sides of .05

Figures

Fig 1 Cognitive and physical effort tasks. Upper Panel. Participants were first trained on the cognitive and physical tasks. Each trial commenced with a pie chart indicating the upcoming effort level. (A) The cognitive effort task utilised an RSVP paradigm. The main task was to detect a target “7” in one of two letter streams at either side of fixation (here initially indicated by “F” and “Q”). Each target stream was surrounded by three distractor streams. An arrowhead at the beginning of each trial indicated the initial target stream. During the trial, a central “3” was a cue to switch attention to the opposite target stream. Cognitive effort was manipulated as the number of attentional switches per trial (1–6). (B) The physical effort task required participants to maintain a sustained grip on a handheld dynamometer at one of six levels of force, as a function of their individually calibrated maximal voluntary contraction (MVC). Lower Panel. (C) After training, participants chose between a fixed low-effort/low-reward baseline and a variable high-effort/high-reward offer. Choices were made while being scanned with fMRI and were made separately for the cognitive and physical domains.
Fig 1 Cognitive and physical effort tasks. Upper Panel. Participants were first trained on the cognitive and physical tasks. Each trial commenced with a pie chart indicating the upcoming effort level. (A) The cognitive effort task utilised an RSVP paradigm. The main task was to detect a target “7” in one of two letter streams at either side of fixation (here initially indicated by “F” and “Q”). Each target stream was surrounded by three distractor streams. An arrowhead at the beginning of each trial indicated the initial target stream. During the trial, a central “3” was a cue to switch attention to the opposite target stream. Cognitive effort was manipulated as the number of attentional switches per trial (1–6). (B) The physical effort task required participants to maintain a sustained grip on a handheld dynamometer at one of six levels of force, as a function of their individually calibrated maximal voluntary contraction (MVC). Lower Panel. (C) After training, participants chose between a fixed low-effort/low-reward baseline and a variable high-effort/high-reward offer. Choices were made while being scanned with fMRI and were made separately for the cognitive and physical domains.CC BY · PubMed Central
Fig 2 Group data reveal similar choice behaviour for cognitive and physical effort. Proportion of accepted offers (±1 standard error of the mean [SEM]) as a function of (A) effort and (B) reward for the cognitive and physical effort tasks. Underlying data for panels A–B can be found in S1 Data.
Fig 2 Group data reveal similar choice behaviour for cognitive and physical effort. Proportion of accepted offers (±1 standard error of the mean [SEM]) as a function of (A) effort and (B) reward for the cognitive and physical effort tasks. Underlying data for panels A–B can be found in S1 Data.CC BY · PubMed Central
Fig 3 Computational modelling revealed that reward devaluation by cognitive and physical effort was best described by different functions. The winning model indicated that cognitive valuation was best fitted by a hyperbolic function and physical valuation by a parabolic function. Three classes of model were compared, based on whether cognitive and physical effort were assumed to discount reward with single or separate functions, or single or separate softmax βs. Models were fitted using maximum likelihood estimation and compared with an Akaike Information Criterion (AIC) (illustrated here). Additional model comparisons using a Bayesian Information Criterion (BIC) revealed the identical pattern of results (S5 Fig). SV(t) represents the subjective value of the offer on trial t, R is the reward in credits (2, 4, 6, 8, or 10), E is the effort level (0.2, 0.4, 0.6, 0.8, or 1.0), and k is a subject-specific discounting parameter. Underlying data can be found in S1 Data.
Fig 3 Computational modelling revealed that reward devaluation by cognitive and physical effort was best described by different functions. The winning model indicated that cognitive valuation was best fitted by a hyperbolic function and physical valuation by a parabolic function. Three classes of model were compared, based on whether cognitive and physical effort were assumed to discount reward with single or separate functions, or single or separate softmax βs. Models were fitted using maximum likelihood estimation and compared with an Akaike Information Criterion (AIC) (illustrated here). Additional model comparisons using a Bayesian Information Criterion (BIC) revealed the identical pattern of results (S5 Fig). SV(t) represents the subjective value of the offer on trial t, R is the reward in credits (2, 4, 6, 8, or 10), E is the effort level (0.2, 0.4, 0.6, 0.8, or 1.0), and k is a subject-specific discounting parameter. Underlying data can be found in S1 Data.CC BY · PubMed Central
Fig 4 A network of domain-general areas was involved in the valuation of cognitive and physical effort. Conjunction analyses demonstrated that the activity of these areas covaried with SV parameters, independent of the type of effort cost. (A) Whole-brain render, showing foci in the right intraparietal sulcus (IPS) and the right dorsolateral prefrontal cortex (dlPFC), comprising the middle frontal gyrus (MFG) and adjacent inferior frontal sulcus (IFS). (B) Coronal section showing activity in the anterior cingulate cortex (ACC) and adjacent dorsomedial prefrontal cortex (dmPFC), IFS, and right insula. (C) Sagittal sections showing ACC/dmPFC, IPS, and dlPFC activity. (D) Parameter estimates for domain-general clusters. Underlying data for panel D can be found in S1 Data.
Fig 4 A network of domain-general areas was involved in the valuation of cognitive and physical effort. Conjunction analyses demonstrated that the activity of these areas covaried with SV parameters, independent of the type of effort cost. (A) Whole-brain render, showing foci in the right intraparietal sulcus (IPS) and the right dorsolateral prefrontal cortex (dlPFC), comprising the middle frontal gyrus (MFG) and adjacent inferior frontal sulcus (IFS). (B) Coronal section showing activity in the anterior cingulate cortex (ACC) and adjacent dorsomedial prefrontal cortex (dmPFC), IFS, and right insula. (C) Sagittal sections showing ACC/dmPFC, IPS, and dlPFC activity. (D) Parameter estimates for domain-general clusters. Underlying data for panel D can be found in S1 Data.CC BY · PubMed Central
Fig 5 The right amygdala uniquely tracked the SV of rewards associated with cognitive costs. Coronal section showing activity within the right amygdala that parametrically varied with SV of the cognitive (but not physical) offer. Underlying data be found in S1 Data.
Fig 5 The right amygdala uniquely tracked the SV of rewards associated with cognitive costs. Coronal section showing activity within the right amygdala that parametrically varied with SV of the cognitive (but not physical) offer. Underlying data be found in S1 Data.CC BY · PubMed Central

Verification

Claims the paper makes

11 Valid · 1 Not valid · 39 Not checked · 119 Own results and interpretations

105 more claims · 1 with their cited sources checked

Create a free account to see all 170 claims and to verify the 80 references still open.

See all claims — free account
  • Probably not a substantial claim. 37
  • It may state two separate claims as one. 27
  • The quote only partly states this claim. 3

Create a free account to see these claims.

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Statistics that don’t add up

Each reported p-value recomputed from its test statistic and degrees of freedom.

1 of 1

Statistics recomputed

Every reported test result, recomputed from its test statistic and degrees of freedom. A result shows as not adding up only once an independent check has confirmed it.

  • Doesn’t add upz = 0.54reported p = .28 · recomputed p = .589

    “Z = 0.54, p = .28”

  • ConsistentF(1.7, 57.2) = 7.48reported p < .005 · recomputed p = .001

    “F(1.7, 57.2) = 7.48, p < .005”

  • ConsistentF(2.05, 67.6) = 89.1reported p < .001 · recomputed p < .001

    “F(2.05, 67.6) = 89.1, p < .001”

  • ConsistentF(2.04, 67.3) = 93.1reported p < .001 · recomputed p < .001

    “F(2.04, 67.3) = 93.1, p < .001”

  • Consistentz = -4.69reported p < .0001 · recomputed p < .001

    “Z = -4.69, p < .0001”

5 of 12 · the rest aren’t loaded

7 more recomputed tests

Create a free account to see every test the Stats check recomputed.

See all tests — free account

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

4 of 4

Citations

Cites

Show all 80 references

Cited by

Items 301–400 of 406

Metadata

Authors
Chong, T. T.-J., Apps, M., Giehl, K., Sillence, A., Grima, L. L., & Husain, M.
Affiliations
  • John Radcliffe Hospital (GB)
  • University of Oxford (GB)
  • University of Cologne (DE)
Corresponding author
  • Chong, Trevor T.-J.
  • Apps, Matthew
Venue
PLOS Biology
Published
2017-02-24
Volume
15
Issue
2
Pages
e1002598
ISSN
  • 1544-9173 (ISSN-L)
  • 1545-7885
OpenAlex
W2590825925
Cited by
415
References
80
Citations by year
2026: 34 · 2025: 42 · 2024: 46 · 2023: 120 · 2022: 35 · 2021: 36 · 2020: 32 · 2019: 20 · 2018: 33 · 2017: 15
FWCI
15.70
Type
journal article
Language
en
Open access
Locations
  • PLoS Biology (publishedVersion, cc-by)
  • Europe PMC (PubMed Central) (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • Monash University Research Portal (Monash University) (publishedVersion, other-oa)
  • PubMed (publishedVersion)
  • Kölner Universitäts PublikationsServer (Universität zu Köln) (publishedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
  • Figshare (submittedVersion, cc-by)
  • University of Birmingham Research Portal (University of Birmingham) (submittedVersion, cc-by)
Funders
  • National Health and Medical Research Council (1053226)
  • Biotechnology and Biological Sciences Research Council (BB/M013596/1)
  • Wellcome Trust (098282)
Topics
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Neural and Behavioral Psychology Studies
  • Life Sciences › Neuroscience › Cellular and Molecular Neuroscience › Neurotransmitter Receptor Influence on Behavior
  • Social Sciences › Decision Sciences › General Decision Sciences › Decision-Making and Behavioral Economics
Keywords
reward devaluation, cost-benefit valuation, cognitive effort, physical exertion, motivation, dorsomedial prefrontal cortex, dorsolateral prefrontal cortex, intraparietal sulcus, anterior insula, amygdala, neural basis of decision making
MeSH terms
Adult, Amygdala, Brain, Choice Behavior, Cognition, Computer Simulation, Physical Exertion, Female, Humans, Magnetic Resonance Imaging, Male, Models, Neurological, Reaction Time, Reward, Risk, Task Performance and Analysis, Logistic Models, Young Adult

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-07, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-07 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (qwen3.8-omni-flash and gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.