Task description.

(A) An illustrative figure depicting the foraging task. (B) Subjects’ view (left eye) in the task. (C) A perspective view of the task environment. (D) An aggregated top-view heat map of head trajectories of all subjects’ data. Red dots denote baskets’ preset locations.

Experiment 1 behaviour and modelling results.

(A) Each coloured line represents one single subject data with the colour reflecting the value in pain-free condition (VAS=0). The dashed black line’s slope and intercepts are fixed effect estimates from a linear mixed model. The slope’s estimate shows an inverse relationship between pain choice probability and VAS rating, β = −0.0263, 95%CI[−0.0383, 0.0144], t(21.27) = −4.40, p < .001. (B) As described in (A), the dashed line’s slope is the fixed effect’s estimate of choice distance bias, and it shows a positive relationship between the choice distance bias and VAS ratings, β = 0.0113, 95%CI[0.00294, 0.0197], t(29.36) = 2.68, p = .012. (C) The negative phasic pain coefficients (M = −0.592, SD = 0.261) showed the model captured the aversiveness of phasic pain stimuli in this free-operant decision-making task, t(23) = −10.89, p < .001. (D) The moving effort coefficient Cm was separated into a horizontal component (M = −0.115, SD = 0.0967) and a vertical component (M = −0.370, SD = 0.217). The fitted coefficients showed lower effort cost to move horizontally than vertically, t(23) = 6.72, p < .001. (E) VAS ratings at different electric shock intensities. (F) Phasic pain utility values as a function of electric shock intensity. Each coloured line represents one subject’s fitted curve. Black line is the average over all subjects. (G) Linear regression showed significant correlation between VAS pain ratings and model estimated phasic pain values, R2 = 0.146, F (1, 118) = 20.13, p < .001.

Skin conductance changes and their correlation with decision values derived from the model.

(A) Evoked SCR for fixation and fruit pick-up events. Compared to seeing the visual cues, shock stimuli induced a greater SCR when the shock intensities were high. Shaded area is the 95% confidence interval. (B) Both decision values and subjective ratings showed a similar level of correlation with the coefficients of fixation-evoked SCR. For decision values: R2 = 0.040, F (1, 155) = 6.38, p = .013; and for subjective pain ratings: R2 = 0.040, F (1, 155) = 6.44, p = .012. (C) In contrast, subjective pain ratings showed a significant correlation with the coefficients of shock-evoked SCR, R2 = 0.080, F (1, 170) = 14.86, p < .001, whereas the correlation for decision values was weak, R2 = 0.006, F (1, 170) = 0.97, p = .325.

Effects of tonic pain on phasic pain ratings and aversive choice probabilities.

(A, B), Two-way repeated measures ANOVA showed no statistically significant difference in subject’s ratings of phasic pain intensity as a result of tonic pain in low (A) or high (B) phasic pain conditions, F(1,30)=2.01, p=.167, nor was there a significant interaction effect between tonic pain and phasic pain, F(1, 30)=0.87, p=.357. (C, D), Similar linear mixed model fitting results for aversive choice probability. No tonic pain condition: β = −0.0286, 95%CI = [−0.0372, −0.0198], p < .001, intercept = 0.473. Tonic pain condition: β = −0.0280, 95%CI = [−0.0349, − 0.0209], p < .001, intercept = 0.472. (E), Aversive choice probability averaged over phasic pain conditions. Two-way repeated measures ANOVA showed significant effect for phasic pain (including no pain condition), F (2, 60) = 36.872, p < .001, but no effect for tonic pain, F (1, 30) = 0.00, p = .998. The interaction effect between tonic and phasic pain was also not significant F (2, 60) = 0.07, p = .930.

Phasic pain ERPs in different pain conditions.

(A, B) Phasic pain ERP comparison in the same tonic pain conditions. (C, D) Phasic pain ERP comparison in the same phasic pain conditions. (E, F) High phasic pain induced a significantly higher N1-P2 amplitude with or without tonic pain. (G, H) Tonic pain stimulation does not show a significant effect in phasic pain ERP’s N1-P2 amplitude. Two-way repeated measures ANOVA showed the effect for phasic pain was significant, F (1, 30) = 35.42, p < .001. The effect for tonic pain was not significant, F (1, 30) = 0.30, p = .589, and the interaction effect was also not significant, F (1, 30) = 0.58, p = .454.

Tonic pain reduced action velocity.

(A) Average hand speed. A one-tailed paired t-test was conducted to evaluate whether the hand speed was faster without tonic pain (M = 1.01, SD = 0.19) than with tonic pain (M = 0.99, SD = 0.18), t(30) = 2.09, p = .023, with a small to moderate effect size d = 0.37. (B) The average fruit collection rate over the one-minute block. A one-tailed paired t-test was conducted to evaluate whether the collection rate was higher without tonic pain (M = 18.22, SD = 3.45) than with tonic pain (M = 17.90, SD = 3.35), t(30) = 1.93, p = .031, with a small to moderate effect size d = 0.35. (C) A visualisation demonstrating the instantaneous hand speed at different percentages of total distance travelled in reaching to the fruit. Shaded area is the 95% confidence interval of the pairwise difference mean.

Experiment 2 model-fitting results.

(A) Fitted vigour constants were greater in tonic pain conditions (M = 53.61, SD = 27.17) than no tonic pain conditions (M = 38.16, SD = 18.88), t(30) = 2.37, p = .024. (B) Separate vigour constants fitting for each phasic pain conditions. Repeated measures ANOVA showed the effect for tonic pain was significant, F (1, 30) = 6.33, p = .017, but the effect for phasic pain was not significant, F (2, 60) = 0.40, p = .673. The interaction effect was not significant, F (2, 60) = 0.67, p = .515. (C) Phasic pain utility function curves. Solid lines are the average values in each condition. (D) Phasic pain utility function values at 50% electric shock intensity. A paired t-test was conducted and found no significant differences between no tonic pain (M = 4.645, SD = 8.982) and with tonic pain (M = −4.807, SD = 8.146) conditions, t(30) = −0.08, p = .933. (E) Phasic pain utility function values at 50% electric shock intensity plotted against aversive choice probability. Both conditions show a significant correlation between the fitted model values and behavioural choice probability. In blocks without tonic pain, R2 = 0.337, F (1, 29) = 14.75, p < .001; in tonic pain conditions, R2 = 0.311, F (1, 29) = 13.07, p = .001.

(A) Topography of t-values from the LMM, with binary tonic pain condition as the dependent variable and EEG band power 0–0.5 s after the decision point as the independent variable. Empty circles represent channel power showing a significant correlation with tonic pain that does not survive Bonferroni correction. Solid circles indicate significant results after correction. Central-parietal and temporal scalp regions (CP6 and T8) showed strong negative correlation with the alpha power, t = −3.52 for CP6 and t = −3.42 for T8. Significant negative correlation with the beta power was found in parietal scalp regions (P8), t = −3.58. (B) Topography of t-values from the LMM, with continuous vigour constants as the dependent variable and EEG band power 0–0.5 s after the decision point as the independent variable. Negative correlation was found with alpha band power at midline parietal electrode Pz, t = −2.59, p = .010.