1. Neuroscience

Modulating task-outcome value to mitigate real-world procrastination via noninvasive brain stimulation

  1. Zhiyi Chen  Is a corresponding author
  2. Zhilin Ren
  3. Wei Li
  4. Zhenzhen Huo
  5. Zhuanzheng Wang
  6. Ye Liu
  7. Bowen Hu
  8. Wanting Chen
  9. Ting Xu
  10. Leonov Artemiy
  11. Chenyan Zhang
  12. Bernhard Hommel
  13. Tingyong Feng  Is a corresponding author
  1. Faculty of Psychology, Southwest University, China
  2. Key Laboratory of Cognition and Personality, Ministry of Education, China
  3. Experimental Research Center for Medical and Psychological Sciences, School of Psychology, Third Military Medical University, China
  4. The Clinical Hospital of the Chengdu Brain Science Institute, China
  5. Key Laboratory for Neuroinformation, University of Electronic Science and Technology of China, China
  6. School of Psychology, Clark University, United States
  7. Institute for Psychological Research, Leiden University, Netherlands
https://doi.org/10.7554/eLife.108241.5
Share this article
Cite this article
  1. Zhiyi Chen
  2. Zhilin Ren
  3. Wei Li
  4. Zhenzhen Huo
  5. Zhuanzheng Wang
  6. Ye Liu
  7. Bowen Hu
  8. Wanting Chen
  9. Ting Xu
  10. Leonov Artemiy
  11. Chenyan Zhang
  12. Bernhard Hommel
  13. Tingyong Feng
(2026)
Modulating task-outcome value to mitigate real-world procrastination via noninvasive brain stimulation
eLife 14:RP108241.
https://doi.org/10.7554/eLife.108241.5
  2. Download BibTeX
  3. Download .RIS
6 figures, 13 tables and 1 additional file

Figures

Figure 1
Download asset Open asset
Experimental diagram of this study.

The upper sub-graph illustrates the whole multiple sessions tDCS neuromodulation pipeline, including seven sessions (days) and eight task-demanded days. Here, the ‘flash’ icon indicates conducting tDCS neuromodulation (active anodal stimulation for active NM group and sham stimulation for sham NM group). This ‘+’ label means a task-demanded day, where no stimulation is required for participants, but all the covariates of interest should be measured by experience sampling methods. The bottom sub-graph reflects the specific pipeline in task-demanded days. Participants were required to provide responses at five progressive time moments nearing the deadline for task aversiveness and outcome value in task-demanded days. In this diagram, the icon of ‘clock’ symbolizes ecological momentary assessment for measuring instant task willingness and outcome value. In addition, twice tests for daily emotions (labeled by ‘+’) were added for participants at 10:00 and 16:00 as covariates of no interests to be adjusted.

Figure 2 with 1 supplement
Download asset Open asset
Flow diagram of CONSORT (A) and partial details of randomized groups (B–D) and neural locations of electric pole (E).

(B) plots the distribution of all the participants' procrastination scores (GPS = General Procrastination Scale). (C) detailed what the full randomized block design is. (D) shows the comparison between the active neuromodulation group and sham control for procrastination scores. Each bar indicates the mean value, and the error bars reflect the standard deviation (SD). n.s. indicates no statistically significant difference. (E) indicates the pipeline to determine the location of the electric pole. The 10–20 EEG standard lead is used to locate the left dorsolateral prefrontal cortex (DLPFC) initially, and the neuronavigation is further utilized to locate the exact location of this targeting region (i.e., left DLPFC).

Figure 2—figure supplement 1
Download asset Open asset
Statistical power estimation by using ANOVA model and power contours according to experimental design.
Figure 3 with 1 supplement
Download asset Open asset
Results of neuromodulation effects to task-execution willingness and procrastination rates (PRs).

(A) shows the effects of neuromodulation to increase task-execution willingness for both active group and sham control across sessions that included in formal analysis (sessions 0 (baseline), 2, 3, 5, 6, and 7). (B) illustrates the effects of whole neuromodulation round to task-execution willingness for both groups. (C) plots the changes of task-execution willingness for both groups after neuromodulation. (D) provides a line chart to show the changes of task-execution willingness across each session that included in formal analysis. (E) shows the effects of neuromodulation to reduce PR for both active group and sham control across sessions that included in formal analysis. (F) illustrates the effects of whole neuromodulation round to task-completion rate for both groups. (G) plots the absolute changes of PR for both groups after neuromodulation. (H) provides a line chart to show the changes of PR across these sessions that are included in formal analysis. The pie graph for each comparison represents the corresponding result of Bayesian factor inference, with the brown piece for supporting H1 evidence and the white piece for supporting H0 evidence. Each bar indicates the mean value, and each line placed onto the bar reflects the standard deviation (SD). Each group included 23 participants. Individual pre- and post-neuromodulation observations are shown in panels B and F, with lines connecting observations from the same participant. Group-by-session effects were examined using linear mixed-effects models, and pre–post simple effects were evaluated using Tukey-adjusted comparisons of estimated marginal means. Trends across sessions were assessed using the Mann–Kendall test.

Figure 3—figure supplement 1
Download asset Open asset
Jittered dot points for primary outcomes in the present study.
Figure 4 with 1 supplement
Download asset Open asset
Results of neuromodulation effects to task aversiveness and outcome value.

(A) shows the effects of neuromodulation to increase AUC of task aversiveness for both active group and sham control across sessions that included in formal analysis (sessions 0 (baseline), 2, 3, 5, 6, and 7). Higher AUC indicates lower task aversiveness as a given task is increasingly executed. (B) plots the changes of AUC of task aversiveness for both groups after neuromodulation. (C) shows the effects of neuromodulation to increase outcome value for both active group and sham control across sessions that included in formal analysis. (D) plots the changes of outcome value for both groups after neuromodulation. (E) provides a line chart to show the changes of AUC of task aversiveness across these sessions that included in formal analysis. (F) provides a line chart to show the changes of outcome value in the same manner. Pie graph for each comparison represents the corresponding result of Bayesian factor inference, with brown piece for supporting H1 evidence and white piece for supporting H0 evidence. Each bar indicates mean value, and each line placed onto the bar reflects standard deviation (SD). Each group included 23 participants. Group-by-session effects were examined using linear mixed-effects models, and pre–post simple effects were evaluated using Tukey-adjusted comparisons of estimated marginal means. Trends across sessions were assessed using the Mann–Kendall test.

Figure 4—figure supplement 1
Download asset Open asset
Jittered dot points for secondary outcomes in the present study.
Figure 5
Download asset Open asset
Results of Quasi-Bayesian model for the mediated role of increased task-outcome value in the association between neuromodulation and task-execution willingness (A) and actual procrastination rates (B).

ADE = average direct effect, ACME = average causal mediation effect, CI = confidence interval, ***p < 0.001.

Figure 6
Download asset Open asset
Changes of actual procrastination rates among pre-test, post-test, and 6-month follow-up for active neuromodulation group and sham-control group.

Pre-test means to test the actual procrastination rates before first HD-tDCS neuromodulation. Post-test means to test the actual procrastination rates after last neuromodulation. The 6-month follow-up means to re-investigate the actual procrastination rates after last neuromodulation for 6 months. Each bar indicates the mean value, and the error bars reflect the standard deviation (SD). At the 6-month follow-up, 22 participants in the active neuromodulation group and 23 participants in the sham-control group were assessed. Baseline-to-follow-up comparisons were evaluated using a linear mixed-effects model with Tukey-adjusted simple-effect comparisons. Brackets indicate comparisons between pre-test and the 6-month follow-up within each group.

Tables

Table 1
Demographic information for included participants.

NM represents active neuromodulation group and sham indicates sham-control group. Anxiety symptoms were measured by State-Trait Anxiety Inventory (STAI). Depression symptoms were tested by Self-Rating Depression Scale (SDS). BF10 describes the Bayesian evidence strength to support the alternative hypothesis, with >3 for strong evidence.

Active NMSCp value (BF10)
MaleFemaleMaleFemale
Gender3203200.99 (-)
Age19.61 ± 0.7822.22 ± 1.440.08 (1.03)
SES2.17 ± 0.652.34 ± 0.640.38 (0.41)
Anxiety48.48 ± 6.5247.30 ± 6.870.56 (0.33)
Depression47.13 ± 8.2747.50 ± 9.280.88 (0.29)
Procrastination71.00 ± 5.4772.07 ± 4.770.26 (0.48)
Table 2
Summary for general linear model in predicting changes of task aversiveness and outcome value to actual procrastination.

S.E. means standard error. *p < 0.05.

βSEp valueOdd ratio (OR)adj R2
ΔTask aversiveness0.620.420.131.850.29
ΔOutcome value0.85*0.420.042.34
Key resources table
Reagent type (species) or resourceDesignationSource or referenceIdentifiersAdditional information
Biological sample (Homo sapiens)Adults with chronic procrastinationThis paperAdults scoring >66 on the General Procrastination Scale recruited from Southwest University; n = 1682 screened, 53 enrolled and randomized, and 46 included in the final analysis; see Materials and methods, Participants
Software, algorithmR (v4.4.1)R Development Core Team, 2024; https://www.r-project.orgRRID:SCR_001905Statistical analyses
Software, algorithmlme4CRAN; https://cran.r-project.org/package=lme4RRID:SCR_015654Linear mixed-effects models
Software, algorithmlmerTestCRAN; https://cran.r-project.org/package=lmerTestRRID:SCR_015656Satterthwaite’s method for LMM
Software, algorithmemmeansCRAN; https://cran.r-project.org/package=emmeansRRID:SCR_018734Estimated marginal means and simple effects
Software, algorithmmediationImai et al., 2010; https://cran.r-project.org/package=mediationRRID:SCR_026984Quasi-Bayesian causal mediation analysis with MCMC sampling
Software, algorithmglmmTMBBrooks et al., 2017; CRANRRID:SCR_025512Beta regression for sensitivity analysis
Software, algorithmggplot2 (v3.5.2)CRAN; https://cran.r-project.org/package=ggplot2RRID:SCR_014601Data visualization
Software, algorithmMATLAB (2021)MathWorks, Inc; https://www.mathworks.comRRID:SCR_001622Data processing
Software, algorithmGraphPad PrismGraphPad Software; https://www.graphpad.comRRID:SCR_002798Data visualization
Software, algorithmG*PowerFaul et al., 2007RRID:SCR_013726A priori power estimation
Software, algorithmExperience sampling mobile appThis paperCustom-developed app for ecological momentary assessment of task aversiveness, outcome value, and task-execution willingness; see Materials and methods, Experimental design and procedure
Other4 × 1 multichannel HD-tDCS stimulation systemSoterix Medical Inc; https://www.soterixmedical.comStimulator with 4 × 1 ring multichannel stimulation adapter (MSA); anodal HD-tDCS, 2.0 mA, 20 min per session, 7 sessions; see Materials and methods, HD-tDCS protocol
OtherHigh-definition neuronavigation systemANT Neuro Inc, GermanyTarget localization of the left DLPFC (F3); see Materials and methods, HD-tDCS protocol
Appendix 1—table 1
Reconstructed preregistration statement table.
QuestionHypothesisSampling plan (e.g., power analysis)Analysis planInterpretation given to different outcomes
Whether the multi-session HD-tDCS over the left DLPFC could reduce real-world procrastination?This HD-tDCS neuromodulation could increasingly mitigate real-world procrastinationWe capitalize on the G*Power software to estimate the minimum sample size obtaining the medium effect size (α err prob = 0.05, 1 − β err prob = 0.80, f = 0.25) by building a repeated measures, within-between, and interaction effect ANOVA model (Faul et al., 2007). In this vein, the medium effect size can be reached once total sample size n = 18, noncentrality λ = 15.75, critical F = 2.1945, numerator Df = 6, and denominator Df = 96. In addition, owing to the repeated measures for each participant, the two-dimensional plot, so-called Power Contours estimation, would be used to visualize and calculate the acceptable sample size by corresponding experimental design (Baker et al., 2021)(a) The generalized mixed-effect linear model (GLMM) would be constructed using 2 (active vs. sham) × 2 (before first neuromodulation vs. after last neuromodulation) full factorial design for procrastination willingness (i.e., self-reported scores) and actual procrastination rate (1 − self-reported task completion rate).
(b) Markov Chain Monte Carlo Generalized linear mixed effects model (MCMCglmm) would be built to re-estimate results that probed above for validation.
(c) Temporal difference-to-difference model would be used to obtain the difference coefficient (Δ) individually, and independent t-tests with Jeffreys–Zellner–Siow Bayesian examinations for between-group differences upon procrastination willingness (i.e., self-reported scores) and actual procrastination rate (1 − self-reported task completion rate) would be conducted.
(d) Between-group comparison for the counts of reporting task procrastination across participants are conducted by x2 test with φ correction or Fisher exact test would be carried out.
(e) The joint model of longitudinal and survival data (JM-LAD), in conjunction with machine learning algorithm, was adopted to capture multi-session effects individually.		-
Whether the multi-session HD-tDCS over the left DLPFC could reduce task aversiveness and increase outcome value?Multi-session HD-tDCS could decrease task aversiveness and increase outcome values meanwhileSee aboveThe generalized mixed-effect linear model (GLMM) would be constructed using 2 (active vs. sham) × 2 (before first neuromodulation vs. after last neuromodulation) full factorial design for task aversiveness (i.e., AUC of five task time points) and outcome value (i.e., AUC of five task time points).-
Why multi-session HD-tDCS could ameliorate procrastination willingness and real-world procrastination behavior?(a) Multi-session HD-tDCS could increase task-execution willingness and decrease real-world procrastination rate by undermining task aversiveness.
(b) Multi-session HD-tDCS could increase task-execution willingness and decrease real-world procrastination rate by increasing outcome value.		No suitable method to make this sample size estimation(a) Generalized linear model to correlate Δ Task aversiveness and Δ Outcome value to Δ Procrastination willingness.
(b) Generalized linear model to correlate Δ Task aversiveness and Δ Outcome value to Δ Real-worldProcrastination.
(c) Quasi-Bayesian causal mediation analysis to model the task aversiveness as the mediator for interpreting the association of HD-tDCS intervention to changes of procrastination willingness and real-world procrastination.
(d) Quasi-Bayesian causal mediation analysis to model the outcome value as the mediator for interpreting the association of HD-tDCS intervention to changes of procrastination willingness and real-world procrastination.		-
Does this multi-session HD-tDCS neuromodulation have a lasting-effect?This effect can be tested in the 6-month follow-up(a) The repeated-measure ANOVA model would be built to test real-world procrastination rates across three time points (pre-, post-, and 6-month follow-up test).
(b) Jeffreys–Zellner–Siow Bayesian factor model would be used to validate the above results derived from repeated-measure ANOVA.		-
Appendix 1—table 2
The number of reporting unexpected event effects in both groups across all sessions.

Here, we conducted a post-neuromodulation investigation to require participants to report whether their performance for completing tasks was influenced by additional impacts outside normal conditions, such as ‘get the flu’, ‘get a fever’, ‘mandatory assignment for other tasks’, and ‘unexpected emergency events’. If in this case, they should report what unexpected events occur to uncontrollably disrupt task execution. In the S1, all the 44 participants reported expected effects from the neuromodulation. In the S1, one participant reported having the flu. In the S3, both participants in the NM group reported a mandatory meeting assignment, while one participant in the SC reported a bicycle accident and an additional two reported mandatory meeting assignments. In the S4, all the 39 participants reported that their task performances were fully disrupted by the weekend. S5–S7 reported the similar additional AE mentioned previously.

S0S1S2S3S4S5S6S7
NM0/2321/231/232/2320/230/233/235/23
SC0/2323/230/233/2319/231/230/233/23
x2-.023-.200.026--.500
Appendix 1—table 3
LMM to robustness check with outcome as subjective procrastination willingness.
Modelβ (group*treatment_day)Std. error (SE)T valuep value
Full model–7.781.79–4.35<0.001
Removing session #7–9.782.25–4.34<0.001
Removing session #6–7.271.83–3.98<0.001
Removing sessions #6 and #7 both–10.023.38–2.960.005
Appendix 1—table 4
LMM to robustness check with outcome as real-world procrastination rates.
Modelβ (group*treatment_day)Std. error (SE)T valuep value
Full model–7.382.45–3.020.004
Removing session #7–10.343.16–3.270.002
Removing session #6–6.552.53–2.590.011
Removing sessions #6 and #7 both–11.074.36–2.540.013
Appendix 1—table 5
Sensitivity analysis results using beta regression.
DVEffectbSEzpConsistency with Gaussian model
Actual procrastination rateGroup–0.7600.202–3.77<0.001Yes (p < 0.001)
Treatment Day0.2250.0643.50<0.001Yes (p < 0.001)
Group × Day interaction–0.1820.091–2.010.045Yes (p = 0.004)
Task-execution willingnessGroup–1.0810.282–3.83<0.001Yes (p < 0.001)
Treatment Day0.4460.0647.00<0.001Yes (p < 0.001)
Group × Day interaction–0.3750.086–4.35<0.001Yes (p < 0.001)
Appendix 1—table 6
Summary for mediation model in predicting task-execution willingness by using treatment (active ms-tDCS vs. sham) from mediated effect of increased task outcome.
Estimate95% lower95% upperp value
ACME21.95***10.6734.900.0001
ADE11.14–2.2025.100.10
Total effect33.10***19.0047.800.0001

  1. *p < 0.05; **p < 0.01; ***p < 0.001.

Appendix 1—table 7
Summary for mediation model in predicting actual procrastination by using treatment (active ms-tDCS vs. sham) from mediated effect of increased task outcome.
Estimate95% lower95% upperp value
ACME30.91**11.9349.630.002
ADE2.77–7.1213.380.61
Total effect33.68**18.1049.740.002

  1. *p < 0.05; **p < 0.01; ***p < 0.001.

Appendix 1—table 8
Summary for mediation model in predicting task aversiveness by using treatment (active ms-tDCS vs. sham) from mediated effect of increased task outcome.
Estimate95% lower95% upperp value
ACME00.000.001.000
ADE25.09**8.2942.000.0036
Total effect25.09**8.2942.000.0036

  1. *p < 0.05; **p < 0.01; ***p < 0.001.

Author response table 1
Comparison to statistics derived from model with covariates (i.e., age, gender, SES, Emotions) and without covariates.
DVEffectCovariatebetaSEtp95% CI
Taskexecution WillingnessGroupIncluded-22.185.66-3.92<. 001[-33.50, -10.86]
GroupExcluded-21.575.50-3.92<. 001[-32.57, -10.57]
Treatment dayIncluded9.531.287.47<. 001[6.97, 12.09]
Treatment dayExcluded9.181.316.99<. 001[6.56, 11.80]
- 6 -
Actual procrastinat ion rateGroup×Day interactionIncluded-7.841.80-4.36<. 001[-11.44, -4.24]
Group×Day interactionExcluded-7.481.86-4.03<. 001[-11.20, -3.76]
GroupIncluded-25.966.49-4.00<. 001[-38.94, -12.98]
GroupExcluded-24.776.19-4.00<. 001[-37.15, -12.39]
Treatment dayIncluded8.921.735.15<. 001[5.46, 12.38]
Treatment dayExcluded8.991.775.09<. 001[5.45, 12.53]
Group×Day interactionIncluded-7.372.44-3.02. 004[-12.25, -2.49]
Group×Day interactionExcluded-7.442.50-2.98. 005[-12.44, -2.44]
Author response table 2
DVEffectbSEzpConsistency with Gaussian model
Actual procrastination rateGroup-0.7600.202-3.77< . 001Yes (p < .001)
Treatment Day0.2250.0643.50< . 001Yes (p < .001)
Group×Day interaction-0.1820.091-2.01. 045Yes (p=.004)
Group-1.0810.282-3.83< . 001Yes (p < .001)
Task-execution WillingnessTreatment Day0.4460.0647.00< . 001Yes (p < .001)
Group×Day interaction-0.3750.086-4.35< . 001Yes (p < .001)

Additional files

MDAR checklist
https://cdn.elifesciences.org/articles/108241/elife-108241-mdarchecklist1-v1.docx

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Zhiyi Chen
  2. Zhilin Ren
  3. Wei Li
  4. Zhenzhen Huo
  5. Zhuanzheng Wang
  6. Ye Liu
  7. Bowen Hu
  8. Wanting Chen
  9. Ting Xu
  10. Leonov Artemiy
  11. Chenyan Zhang
  12. Bernhard Hommel
  13. Tingyong Feng
(2026)
Modulating task-outcome value to mitigate real-world procrastination via noninvasive brain stimulation
eLife 14:RP108241.
https://doi.org/10.7554/eLife.108241.5