1. Neuroscience
Download icon

Forced choices reveal a trade-off between cognitive effort and physical pain

  1. Todd A Vogel  Is a corresponding author
  2. Zachary M Savelson
  3. A Ross Otto
  4. Mathieu Roy  Is a corresponding author
  1. Department of Psychology, McGill University, Canada
  2. Institute of Cognitive Science, Carleton University, Canada
Research Article
Cite this article as: eLife 2020;9:e59410 doi: 10.7554/eLife.59410
5 figures, 3 tables, 1 data set and 2 additional files

Figures

Example of a trial for the decision-making task.

Participants made a choice between a level of the N-back task and a level of pain. If participants chose the effortful option, they performed the N-back task at the given level. If the painful option was chosen, participants received the corresponding painful stimulus and avoided performing the N-back task.

Averaged heatmap of choice behaviour across participants as a function of pain stimulus level (horizontal axis) and N-back level (vertical axis) on offer (see Supplementary file 1C for model estimates).

Lighter colours indicate a higher percentage of accepting the painful stimulus.

Parabolic scaling of effort levels and averaged heatmap of choices predicted from the computational model.

(a) Higher offered effort levels led to a parabolic increase in the aversive SV of effort. This value was compared against the pain level on offer before being fit to participants’ choices. (b) Averaged heatmap of choices predicted from the parabolic model fit to effort levels. Lighter colours indicate a higher probability of choosing the pain option.

Estimated beta weights (presented in log-units) for the choice RT model.

Positive values indicate an increase in predicted RTs (i.e. slowing down) and negative values indicate a decrease. RTs significantly increased at higher levels of the chosen good (i.e. at higher levels of SVeffort when choosing the effort option and higher levels of pain when choosing the pain option). These effects remained even after controlling for the difficulty of the choices (Supplementary file 1J), controlling for a person’s overall proportion of choices (Supplementary file 1K), and in those participants exhibiting an no overall bias towards one option over the other (Supplementary file 1I). Error bars represent the 95% CI based on the SEs from the model; ***p<0.001 (see Supplementary file 1F for full information on coefficient estimates).

Predicted probabilities of accepting the pain option at each N-back and pain level moderated by need for cognition and pain catastrophizing.

(a) Those higher in need for cognition were more likely to choose the effortful option, especially when pain and effort offers were both in the higher, more aversive range (e.g. 4-back vs. 80/100 pain). (b) Similarly, those higher in pain catastrophizing preferred to avoid the pain option (and thus choose the cognitive effort option) at increasing N-back levels. NFC = need for cognition, PCS = pain catastrophizing. The grey dashed line indicates the indifference point between the two choices (i.e. the point where the probability of choosing either the pain or effort option is 50%).

Tables

Table 1
Goodness-of-fit Estimates from Computational Models on Cognitive Effort Levels.
LinearExponentialParabolicHyperbolic
AIC1223.321245.431170.661293.65
  1. Note. Separate models were fit for each mathematical function on participants’ data; AIC = Akaike Information Criterion (Akaike, 1974).

Author response table 1
Variableβ95% CIdftp
(Choose Pain Intercept)8.13[8.01, 8.24]40.15135.03<.001 ***
(Choose Effort Intercept)7.92[7.82, 8.02]31.99164.95<.001 ***
Trial Number−0.01[−0.01, −0.01]31.237.19<.001 ***
Choose Pain × Pain predicted−0.06[−0.18, 0.06]35.831.09.28
Choose Effort × Pain predicted0.05[−0.06, 0.17]22.890.89.38
Choose Pain × SVeffort0.02[−0.02, 0.07]28.170.99.33
Choose Effort × SVeffort0.14[0.10, 0.18]25.866.58<.001***
Choose Pain × Pain level0.13[0.08, 0.17]25.785.74<.001 ***
Choose Effort × Pain level−0.06[−0.08, −0.03]30.634.91<.001 ***
Author response table 2
Variableβ95% CIdftp
(Choose Pain Intercept)8.06[7.94, 8.18]33.13134.18<.001 ***
(Choose Effort Intercept)7.93[7.84, 8.03]32.75161.52<.001 ***
Trial Number−0.01[−0.01, −0.01]31.137.18<.001 ***
Choose Pain × Previous choice0.10[0.03, 0.18]1363.832.91.004 **
Choose Effort × Previous choice−0.02[−0.07, 0.03]1343.700.71.48
Choose Pain × SVeffort0.01[−0.02, 0.05]21.620.69.50
Choose Effort × SVeffort0.14[0.10, 0.19]24.906.77<.001 ***
Choose Pain × Pain level0.14[0.09, 0.18]20.746.24<.001 ***
Choose Effort × Pain level−0.06[−0.08, −0.04]28.995.45<.001 ***

Data availability

All data analyzed for this study can be found on OSF (https://osf.io/n4cht/).

The following data sets were generated
    1. Vogel TA
    2. Otto R
    (2020) Open Science Framework
    ID n4cht. Forced Choices Reveal a Trade-Off between Cognitive Effort and Physical Pain.

Additional files

Supplementary file 1

This file contains the supplementary tables and figures referenced in the main text.

(A) contains the average performance on the levels of the N-back task. (B) contains the average pain ratings for the thermal stimuli. (C) contains model estimates from the multilevel logistic regression on choices. (D) contains model estimates from the multilevel logistic regression on choices after controlling for trial number. (E) contains model estimates from the multilevel logistic regression on choices after controlling for the choice made on the previous trial. (F) contains model estimates from the multilevel linear regression on choice response times (RTs). (G) contains the heatmaps of the average predicted RTs from the multilevel linear model. (H) contains scatter plots of predicted vs observed RTs for each combination of effort and pain levels. (I) contains the model estimates from multilevel linear regression on choice RTs for the subset of participants who displayed no significant bias towards either choice option. (J) contains the model estimates from the multilevel linear regression on choice RTs after controlling for the difficulty of the choices. (K) contains model estimates from the multilevel linear regression on choice RTs while including a subject-level predictor of a person’s overall probability of choosing the pain option. (L) contains model estimates from the multilevel logistic regression examining the influence of need for cognition. (M) contains model estimates from the multilevel logistic regression examining the influence of pain catastrophizing. (N) contains model estimates from multilevel logistic regressions examining the influence of the other measures of individual differences. (O) contains correlations between the measures of individual differences.

https://cdn.elifesciences.org/articles/59410/elife-59410-supp1-v2.docx
Transparent reporting form
https://cdn.elifesciences.org/articles/59410/elife-59410-transrepform-v2.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)

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

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