Salience or value: what drives priority in pain-cognition interactions?

Supplemental Digital Content is Available in the Text.

When performing a cognitive task while experiencing pain, prioritizing either pain processing or cognitive task pursuit depends on comparative value in males.

Abstract

The evolutionary function of pain is to protect against injury by capturing attention and motivating nocifensive behavior. This makes pain inherently salient and capable of disrupting the pursuit of competing goals. However, this interference can sometimes be overridden by concurrent demands. We have yet to discover how priority is assigned when pursuing competing goals while experiencing pain: through attentional resource competition (salience-based) or motivational conflict (value-based), or both. We investigated these different mechanisms in a study in which 40 healthy adults completed the cognitive branching paradigm, which comprises 2 nested backward letter-matching tasks, 1 with a low-value reward ($0.05) and the other with a high-value reward ($1). Participants performed the tasks first without stimulation and then under 2 conditions: tonic painful heat and tonic nonpainful electric stimuli that were matched for salience. We analyzed performance using linear mixed models and conducted post-hoc Bayesian tests for significant interactions. We found that pain selectively disrupted the low-value task, but not the high-value task, indicating that priority assignment was value-based, whereas the electric stimulus had no impact on task performance. In addition, sex-disaggregated analyses showed that this effect was exclusively seen in males, whereas females exhibited no significant impact of pain on task performance. Our findings suggest that motivational conflict, rather than resource competition, determines priority assignment when pain is concurrent to competing demands. In addition, we show for the first time, sexually dimorphic mechanisms in pain–cognition interactions, which contributes to an emerging literature of sex differences in pain mechanisms and behaviours.

1. Introduction

The purpose of pain is to mitigate harm by drawing attention to potential injury while maintaining evolutionary fitness.^9,35,37,55 This tradeoff is exemplified in goal pursuit while experiencing pain: for some individuals, task performance is adversely affected by pain, whereas for others, task performance improves.⁴³ The underlying mechanism of these differences in pain–cognition interactions remains unclear.

A pervasive model to explain how pain interacts with cognitive processing posits that the inherent salience of pain diverts attention from concurrent goals to sensory and affective processing systems, thereby interrupting ongoing goal pursuit (herein referred to as the salience model).^11,25,49,50 Based on this model, if a competing goal is more salient than the pain being experienced, attentional resources are proportionally shifted to allow goal pursuit, while pain is suppressed.^4,25 This model, however, does not capture how an individual's value assignment determines priority in resource allocation to pain vs competing goals—ie, motivational conflict.¹³

An alternative model that may better explain pain–cognition interactions is the Motivation-Decision Model.¹³ This model posits that pain avoidance and goal pursuit are competing inhibitory motivations: prioritizing the former facilitates pain, and prioritizing the latter reduces pain. It remains unclear how priority is assigned. We introduce a value-based framework where the contextual value of pain is compared to the value of competing goals, and the higher value option is prioritized. The cognitive branching paradigm requires relative valuation through working memory and resource competition,²² where participants track the value of 2 tasks performed concurrently: 1 high-value task, and 1 low-value task. The addition of a third value-based task impairs performance on the lowest value task.⁸ Studies using attention span tasks have shown that pain negatively affects task performance, but that engaging in the task, especially when rewarding,^51,52 can have analgesic effects.^10,47,48 However, these studies do not explicitly assess a value-based model. Previous studies that have explored the tradeoff between pain and value (through monetary gain) report that pain valuation is context-dependent, and that monetary reward is analgesic^4,5 and enhances pain discriminability.⁶ However, they do not use cognitive tasks.

Here, using the cognitive branching paradigm, we investigated whether pain–cognition interactions are driven by salience, value, or both, in a single experimental paradigm. If driven by salience, both painful and nonpainful salient stimuli would equally affect performance regardless of task value (Fig. 1A). If driven by value, pain would disrupt low-value but not high-value tasks, whereas nonpainful stimuli would show no value-based effects (Fig. 1B). If driven by both salience and value, pain and nonpainful salient stimuli would affect low-value tasks more than high-value tasks (Fig. 1C). Participants completed the paradigm under 2 conditions: tonic heat pain and tonic nonpainful electric stimulation matched for salience. Given increasing evidence of sex differences in pain physiology and pain behaviours,^21,36,46 we also explored whether there were sex differences in the pain–cognition paradigm.

Figure 1.

Hypothesized models for pain–cognition interactions. (A) If pain–cognition interactions are driven by salience, a painful and nonpainful salient stimulus should affect performance across both low-value and high-value tasks to the same degree. (B) If pain-interference is driven by value, stimuli will interfere with performance on a competing low-value, but not high-value task. (C) If pain-interference is driven by both salience and value, stimuli should affect performance on low-value tasks, and to a lesser extent, on high-value tasks. (D) If pain does not interfere with performance, there will be no difference in reaction times on either task compared to baseline.

2. Materials and methods

2.1. Experimental design

We aimed to determine whether pain interferes with concurrent goals based on its salience, its value, or a combination of these. To do so, participants attended 3 sessions (Fig. 2). The first session (Fig. 2A) comprised of screening, questionnaires, determining sensory thresholds, calibration, and training on a complex value-based cognitive paradigm: the cognitive branching paradigm.²² In the next 2 sessions (Fig. 2B), participants performed the task without any stimulation (baseline) and while experiencing a stimulus: tonic heat pain or innocuous tonic electric stimulation. In session 2, the stimulus was calibrated to elicit a salience rating of ∼6/10, and the stimulus delivered in session 3 was matched to the salience of the stimulus received in session 2. This allowed us to ensure salience-matching between the 2 types of stimulation.

Figure 2.

Summary of the methods across sessions. Each box represents a procedure within the session and text underneath refers to inclusion criteria for each. (A) In session 1, participants completed consent and questionnaires, followed by sensory testing, during which their thresholds for warmth, heat pain, and electric stimuli were tested. Next, they were familiarized to a range of heat and electric stimulus intensities, which were readministered in a subsequent ratings step to obtain painful, intensity, salience, and unpleasantness ratings and predict the temperature and current needed to elicit a ∼6/10 salience. Participants were then trained on the cognitive branching paradigm and had to achieve task proficiency to qualify for sessions 2 and 3. (B) In sessions 2 and 3, participants performed the cognitive branching paradigm at baseline with no stimulus followed by either tonic heat pain or tonic electric stimuli, the order of which was counterbalanced (tonic heat is shown in session 2 here). Capsaicin was placed on the lateral right calf and covered with a thermal probe to incubate for 15 minutes. Afterwards, the temperature was increased to the predicted temperature that would elicit ∼6/10 from session 1 minus 5°C. The cognitive branching paradigm was performed with capsaicin-heat pain and participants qualified for session 3 if >60% of the paradigm was perceived as painful. In session 3, participants performed the branching paradigm at baseline with no stimulus followed by with a tonic electric stimulus that was salience-matched to the average salience of tonic heat pain in session 2. Participant data were included in analyses if the electric stimuli were nonpainful for >60% of the session and if the average electric salience was within 0.5/10 of the average heat pain salience.

2.2. Sample size determination

Based on sample sizes of similar studies examining pain–cognition interactions contemporaneous to the start of this study (2018),^30,48 we sought to recruit 20 participants. We also performed a power calculation running a linear mixed model on pilot data and found that a sample of 20 led to 93.7% power. Given evidence of sex differences in pain mechanisms,^21,36,46 and the dearth of data on pain–cognition interactions, we sought to investigate the effects of the task within sex (ie, sex disaggregated analyses). As such, we doubled the sample size to ensure that these analyses would be adequately powered.

2.3. Participants

Figure 3 provides a CONSORT diagram of participants. We recruited and screened 171 adults from the University of Toronto environment. Inclusion screening criteria for participants were having good general health and being between 21 and 45 years of age to mitigate any age-related confounds. Exclusion criteria were self-reported neurological disorders, cognitive impairment, allergy to chili peppers, acute pain on the day of testing, troublesome persistent or recurrent pain or pain-related discomfort in the past 2 months, and present or past chronic pain. One hundred eighteen individuals meeting screening criteria provided written informed consent to procedures approved by the University of Toronto's Human Research Ethics Board (#33465). Additional screening criteria were then assessed based on depressive symptoms and cognitive function (see Section 2.4). Participants were compensated commensurate to their participation, and those. who completed the full study also received a task bonus, calculated using their total score on 1 of the 4 stimulation conditions scaled to a maximum of $75. Briefly, 91 participants attended all 3 sessions. During the experiment, further exclusion criteria excluded 51 participants, as follows: 25 had unmatched salience ratings between pain and electric stimuli (see Statistical Analysis—Ratings), 1 had error rates at chance level, 4 were capsaicin nonresponders (did not ever find capsaicin painful), 5 perceived electric stimuli as painful, 2 were hypersensitive to capsaicin, 3 had a high tolerance to capsaicin (temperatures beyond safe range needed to elicit pain with capsaicin), and 11 had an insufficient number of trials that were painful (>3/8 runs). This left a final sample of 40 participants (20 males and 20 females). The exclusion criteria were essential to control for as many factors as possible to properly test the value-based model of pain. Although this limits the ecological validity of the study, this approach was required to parse the salience and value contributions to pain–cognition interactions.

Figure 3.

Consort diagram of participants consented, completed, and included in analysis.

2.4. Session 1: questionnaires, sensory thresholds, and training

Participants completed 2 screening questionnaires: Beck's Depression Inventory (BDI)³ and the Mini Mental State Examination (MMSE).¹⁴ If participants scored BDI ≥ 21 or reported suicidal ideation, they were excluded from the study. Furthermore, if participants scored <27 on the MMSE, they were also excluded, as they were unlikely to gain adequate proficiency in the task (based on pilot data). Further exclusion criteria during this session were (1) failure to gain proficiency in the cognitive branching paradigm (see Cognitive branching paradigm below), (2) heat pain thresholds > 48°C, (3) hypersensitivity to heat stimuli, (4) electric threshold beyond safe levels (>15 mA), and (5) pain reported during the electric stimulation.

2.4.1. Sensory thresholds

We tested sensory thresholds to ensure participants perceive heat stimuli within a normal range without capsaicin. Warm thresholds and heat pain thresholds were measured using the method of limits⁶⁰ to determine the minimum temperature detected as warm and painful, respectively. A 30 mm × 30 mm thermal probe (Medoc Advanced Medical Systems Ltd., Ramat Yishai, Israel) was placed on the right lateral calf and the temperature was increased with a ramp rate of 1°C/s from a baseline of 32°C (to a maximum of 50°C). Participants pressed a button at the first warm or painful sensation, depending on the test, and the temperature was recorded. They repeated each test 4 times. We calculated the warmth detection threshold as the average across the 4 trials, whereas the heat pain threshold was the average of the last 3 trials, as the first trial was consistently lower (see Fig. S1, http://links.lww.com/PAIN/C345).

To test whether pain exerts its interference in a salience-based manner, participants would also perform the tasks with a nonpainful, salience-matched tonic electric stimulus in subsequent sessions. Thus, we collected electric thresholds during the training session as a baseline to later match the salience of electric stimuli to heat pain. To measure electric thresholds, we placed 2 hydrogel surface electrodes (Ag/AgCl) to the right sural nerve and a DS7A current stimulator (Digitimer Ltd., Welwyn Garden City, the United Kingdom) produced square wave pulses lasting 50 μs. We used the staircase method: we increased the current in steps of 1 mA starting from 0 until participants detected a nonpainful “light tap sensation,” then lowered it by 0.5 mA until they could no longer identify a stimulus.⁴⁵ The threshold was determined by increasing current by 0.25 mA until participants once again detected a stimulus, then up and down by 0.05 mA until 5 positive responses were reported.

2.4.2. Stimulus familiarization

Once thresholds were collected, participants underwent a procedure to familiarize them to heat and electric sensations to aid their ratings of the stimuli in subsequent sessions. Participants received 5 to 8 stimuli (duration of 8 seconds) with each stimulus increasing in intensity. The starting heat intensity was the midpoint between their warmth detection threshold and their heat pain threshold, and the starting electric intensity was their threshold +1 mA. Participants informed the experimenter each time they detected a sensation, and whether it was painful (yes/no). We defined pain as “an unpleasant sensory and emotional experience associated with, or resembling that associated with, actual or potential tissue damage.”⁴¹ They were told that if the stimulus hurt, felt like it was or could potentially be damaging their skin, that would be considered painful. Stimulus intensities increased by 1 step (1°C or 1 mA) until participants reported maximal tolerance, pain evoked by the electric stimuli, or stimulus intensities that reached the maximal stimulation intensity of 49°C (for safety purposes). This process enabled participants to feel a range of innocuous to noxious sensations to better discriminate and compare stimuli before using rating scales during the experiment.

2.4.3. Rating Scales

We administered the same range of stimuli as in the familiarization procedure, but this time they were presented in a pseudorandom order, and participants quantified each stimulus using 3 rating scales: intensity, salience, and unpleasantness. We defined intensity as “the strength or magnitude of the stimulus,” salience as “the extent to which the stimulus is likely to grab and direct your attention,”²⁵ and unpleasantness as “the extent to which something bothers you or causes a negative emotional response.” Participants were provided with a more detailed description of these dimensions with an analogy, and they could consult it, if they wished (see Supplementary Materials, http://links.lww.com/PAIN/C345). Each scale ranged from 0 to 10 (intensity: 0 = not intense, 10 = most intense imaginable; salience: 0 = not salient, 10 = extremely salient, unpleasantness: 0 = not unpleasant, 10 = extremely unpleasant). Using the ratings collected, the FORECAST function in Excel predicted the temperature and current needed to elicit a moderate salience rating of ∼6/10 in each participant, and these values were used as starting points for stimulation in the experiment sessions. Once sensory testing was completed, participants learned the cognitive branching paradigm.

2.4.4. Cognitive branching paradigm

To test the interference of pain on cognition, we modified the cognitive branching paradigm developed by Koechlin et al.^8,22 Participants viewed letters of the word TABLET displayed in a pseudorandom order as a backward letter-matching paradigm using e-Prime 3.0 software (Psychology Software Tools, Pittsburgh, PA, the United States of America). The paradigm comprises 3 phases—primary, secondary, and return. The phase is determined by shapes that frame the letter: primary is denoted by a square frame, secondary is denoted by a circle, and the return phase is also denoted by a square. Participants first practiced the primary phase to become proficient. The task requires participants to determine whether the first letter is the start of the sequence (a “T” in TABLET) and continue the backward letter-matching paradigm, answering whether the current letter follows the preceding letter consecutively in the spelling of the word. Participants were asked to respond with a key press as quickly and accurately as possible to trials to indicate affirmative and negative answers, respectively.

Once participants were comfortable with the primary phase, they learned the branching paradigm. Each block had 4 trial types. The “primary” trial was the first letter of the primary phase, the “secondary” trial was the first letter of the secondary phase, the “return” trial was the first letter of the return to primary phase, and “average” trials were the remaining trials of each phase (eg, average trials in the primary phase represent all the trials after the primary/first trial, ie, trials 2–end).

Each paradigm block began with 3 to 5 trials in the primary phase, followed by 3 to 5 trials for the interrupting second phase, and 2 to 3 trials of the return phase where they resume the primary phase (Fig. 4A). Each secondary phase required the participant to start a new sequence of the matching task. The participant was required to remember the last letter of the primary phase while performing the secondary phase to successfully resume the task in the return phase: this process is called cognitive branching.

Figure 4.

Experimental Task Design. (A) A schematic of 1 block of the cognitive branching paradigm. The paradigm begins with a letter in the word “TABLET” presented in a square-shaped frame (orange square). Participants answer whether this first letter is a “T” (yes/no). In all remaining trials of the primary task (second-end), participants perform a backward letter-matching task, answering whether the letter currently presented immediately follows the letter previously presented in the spelling of the word “TABLET” (yes/no). Once the phase shifts from primary to secondary, participants will restart the backward letter-matching task, answering whether the first letter presented in the new shape is a “T” (green circle). Participants then continue spelling “TABLET” within the secondary phase while also holding their place in the primary phase in mind while they perform the secondary phase, so that when the phase returns to primary (purple-filled square), they will resume the primary phase where they left off, and continue spelling “TABLET.” (B) Experimental design of 1 task run of a stimulation condition. Participants performed 4 blocks of the cognitive branching paradigm while experiencing heat or electric stimulation throughout. Blue squares represent letters in the primary and return phases, and purple circles represent letters in the secondary phase. After the 4 blocks were completed, stimulation was turned off, and participants rated the intensity, salience, and unpleasantness of the stimulus they experienced during the paradigm, followed by a brief break before the next run began. This was repeated 8 times for each stimulation condition.

Each correct trial in the primary and return phase was worth a high value ($1), whereas each correct trial in the secondary phase was worth a low value ($0.05). During cognitive branching, the primary and secondary phases represented the 2 active tasks to be performed, and the addition of a third task has been shown to disrupt performance on the task with the lowest reward.⁸

Previous work has examined reaction time (RT) in a factorial design where primary and secondary tasks vary between high and low reward and found that higher reward improves RT on each task, except on the secondary task when primary rewards are high (ie, both primary and secondary tasks have equal, high rewards).⁸ This indicates that, in this condition, the anticipation of the primary reward devalues the reward of the secondary task. Previous evidence has shown that pain value is context dependent.⁵¹ As such, we added pain to the branching paradigm to determine whether pain poses a burden on cognition in a value-based manner. We predicted that the value assigned to pain would be compared to the reward of the tasks: if the value assigned to pain is greater than the reward of the task, pain will interfere with performance, whereas if the reward of the task is greater than the value assigned to pain, it will not (Fig. 1).

After learning the branching paradigm, participants completed 3 practice runs (it takes 3 exposures to become proficient in the tasks)³⁴ with 8 randomized blocks in each to generate an average accuracy and RT score. Participants required an average accuracy ≥90%, return accuracy ≥80%, and average RT ≤ 900 milliseconds in at least 1 run to qualify for subsequent sessions. It was important that participants were proficient in the tasks to minimize learning effects on performance.

2.5. Sessions 2 and 3: experiment

Participants who passed the training session attended 2 experimental sessions on separate days to perform the cognitive branching paradigm while experiencing tonic heat pain or innocuous electric stimulation. The 2 sessions were always within 7 days of each other. Given that the experiment occurred over separate days, participants first performed the paradigm without any stimulation (baseline) to account for any potential confounds. We counterbalanced the session order of pain and electric stimuli across participants. Stimuli were salience-matched across sessions.

2.5.1. Tonic heat pain condition

We administered a chemical heat-based persistent model of pain, where a 10% topical capsaicin cream (compounded by Pace Pharmacy, Toronto, ON, Canada) combined with thermal stimuli at intensities below the heat pain threshold were used to induce primary heat hyperalgesia.³⁸ As previously done,^15,28 1 g of capsaicin cream was applied to a 30 × 30 mm square on the right lateral calf and held in place with a clear adhesive bandage (Tegaderm, 3M, St-Paul, MN). The same thermal probe from training was placed overtop the bandage for 15 minutes at 32°C to allow the capsaicin to incubate and sensitize the skin. After incubation, we asked participants whether they felt that the baseline temperature was warmer than at the start of incubation, and whether it was now painful and salient. Given the sensitization that occurs in response to capsaicin, we subtracted 5°C from the temperature predicted to elicit ∼6/10 salience (determined in session 1, see Rating Scales) and used this temperature as a marker of a normal response to capsaicin. If the temperature needed to reach the target salience rating was much higher than the predicted temperature minus 5°C, participants had high tolerance, and if the temperature needed to be much lower, they were hypersensitive. We increased the heat from baseline between 1 and 3°C (depending on the predicted temperature) for 8 seconds asking participants to rate the sensation for intensity, salience, and unpleasantness. We adjusted the temperatures until we obtained either a ∼6/10 salience rating (session 2), or a salience rating within 0.5/10 of the average salience rating of the electric condition from session 2 (session 3). Participants who rated the baseline temperature as > 4/10 for salience and as painful were defined as hypersensitive to capsaicin and excluded, as increasing the temperature even 1–3°C led to > 8/10 salience. If, during the experiment, capsaicin was rated as nonpainful more than twice at 46°C, participants were categorized as having high tolerance and were excluded for safety purposes.

2.5.2. Tonic electric control stimulus

The electric stimulus comprised a continuous train of high frequency electric pulses. To do so, an external DG2A Train/Delay Generator (Digitimer Ltd) continuously triggered the Digitimer DS7A at a frequency of 100 Hz. The current predicted to elicit a salience of ∼6/10 from the first session was applied before beginning the experiment to confirm it was not painful. We also adjusted the current to ensure that we obtained either a ∼6/10 salience rating (session 2), or a salience rating within 0.5/10 of the average salience rating of the pain condition from session 2 (session 3).

2.5.3. Stimulation condition duration

During each stimulation condition, participants performed the cognitive branching paradigm in blocks of 4, followed by a break and ratings, and this was repeated 8 times. A set of 4 task blocks and subsequent break were termed a “run.” Figure 4B depicts the structure of a single task run for a stimulation condition. On a given trial, the letter was presented for 500 milliseconds, followed by a jittered intertrial interval with a fixation cross averaging 3000 milliseconds (ranging from 2600 to 3400 ms) where a response could be made. If a response was not provided within the time window, the next trial began. Each run averaged 10 trials per block, totalling 320 trials per condition. In total, there were 32 primary, 32 secondary, and 32 return trials, 128 average high-value task trials, and 96 average low-value task trials per condition, per participant. The duration of a single task run was 2.5 minutes.

2.5.4. Stimulus ratings

After each task run, participants had a 1-minute break (where the stimulation was stopped). On the computer, they rated whether the stimulus was painful (yes/no) and provided ratings of stimulus intensity, salience, and unpleasantness they perceived during the previous task run (see Rating Scales). Specifically, participants responded to the following prompts: “Was the stimulus on your leg painful?”, and “How intense/salient/unpleasant was the stimulus on your leg?” Participants reported ratings using designated keyboard keys (f and j keys on a QWERTY keyboard). There were 8 ratings per stimulation condition (baseline, electric, heat). Before resuming the subsequent run, stimuli were adjusted by 0.5 to 1°C and 0.3 to 1 mA depending on the ratings provided to ensure stable perceptual ratings, and to maintain salience-matching.

2.6. Statistical analysis

2.6.1. Task data curation

All analyses were performed using RStudio,^39,56,57 unless otherwise indicated. Separate analyses were conducted for RT and error rates. Error rate analyses are presented in the Supplementary Materials, http://links.lww.com/PAIN/C345. Only average trials (second–end trials in each phase) were used in analyses, as performance in the primary, secondary, and return trials may reflect the surprise of the task changing and not the working memory process itself needed to answer the trials. Reaction times of correct task trials in each condition were split by task reward (high vs low). Outliers, defined as RTs of correct trials >2 standard deviations outside the mean, were removed. To compare RTs between pain and electric conditions, we baseline-corrected by regressing out the mean baseline RT of each stimulus and reward condition from each trial of the stimulation condition RTs for each participant. We used these corrected, preprocessed RTs in analyses.

2.6.2. Ratings

We removed runs of the paradigm that were reported as painful during electric stimulation and nonpainful during heat pain stimulation. We averaged the ratings across each rating type (intensity, salience, unpleasantness) per stimulation condition. If the difference between the average salience rating was > |0.5|/10, up to 3 ratings and their corresponding task runs (up to 84 out of 224 trials) were removed from each condition. If, after removing those runs, the mean difference in salience ratings between the pain and electric conditions was > |0.5|/10, or there were fewer than 60% of the task trials remaining (<5 runs), we excluded the participant.

Given that salience-matching was a crucial part of our study, several statistical tests were performed to compare average salience ratings across stimulation conditions. The mean salience ratings of the participants who met the strict matching criteria were analyzed using IBM SPSS Statistics v28 for Mac (IBM Corp., Armonk, NY the United States of America). We tested the similarity of salience ratings between pain and electric stimuli using a Pearson correlation. Cronbach alpha was derived to confirm the reliability of salience ratings across modalities. To determine whether there were consistent differences between these ratings based on stimulus type (eg, pain is always slightly more salient than electric), we performed a Bayesian paired samples t test. To confirm equivalence of salience ratings, we also performed the two one-sided test (TOST).²⁴ To assess whether there was any significant habituation in salience across task runs that differed between stimulation conditions, a repeated measures ANOVA was performed (Supplementary Materials, http://links.lww.com/PAIN/C345). In a separate cohort, we tested whether the salience of the capsaicin-heat pain model fluctuates over time independent of the cognitive task (Supplementary Materials, http://links.lww.com/PAIN/C345). The association between intensity ratings and unpleasantness ratings was tested with a Pearson correlation, and differences were assessed with a paired samples t test. Significance was set to P < 0.05.

2.6.3. Linear mixed models

We performed an analysis comparing preprocessed RTs between each stimulation condition and its respective baseline condition and a separate analysis comparing baseline-corrected RTs between pain and electric stimulation conditions. Linear mixed models with the lmerTest package in RStudio²³ investigated differences at the level of (1) whole group; (2) within sex via sex disaggregated tests, and, upon viewing a sex disaggregated effect; (3) between sexes via interaction. Each model had a random slope and intercept with RT as the dependent variable, and participant ID as the random effect. In models 1 and 2, fixed effects were the two-way interaction between stimulation condition (baseline vs stimulus OR pain vs electric) and task reward (high vs low). In model 3, fixed effects were the three-way interaction between sex, stimulation condition, and task reward. Significance was set to P < 0.05.

2.6.4. Bayesian informative hypothesis evaluation

We did not perform post-hoc pairwise comparisons as they have been shown to lead to greater false positives and do not capture the contribution of interdependent conditions that may be jointly driving behaviour.¹⁶ Instead, significant interactions were further investigated using Bayesian informative hypothesis evaluation with the bain package in RStudio.^18,20 This method allows you to mathematically represent hypotheses that offer possible explanations of behaviour of how each experimental condition is related to the other.¹⁶ Each hypothesis is compared through model selection and is given a posterior model probability calculated through Bayes theorem (assuming equal prior model probabilities). Based on these probabilities and the data, a Bayes factor is computed to compare each hypothesis to an unconstrained hypothesis (Hu) that assumes no relationship between conditions (the null). We used this method to determine whether pain was interfering using a salience-based hypothesis, value-based hypothesis, value + salience hypothesis, or having no effect. Estimated marginal means from the linear mixed models for each stimulation condition and task reward were used as inputs of the analysis, and hypotheses were compared to Hu. The hypothesis with the largest Bayes factor and posterior model probability was considered the most likely hypothesis. Hypotheses were structured as follows and the mathematical representations of each can be found in Table S1, http://links.lww.com/PAIN/C345:

• (1) If pain interferes based on its salience, pain should affect performance on the low-value and high-value tasks equally compared to its baseline, and the same effect should be seen in the electric condition compared to its baseline;

• (2) If pain interferes based on its assigned value, pain should affect performance on the low-value but not the high-value task relative to its baseline, and this effect should not be seen in the electric condition;

• (3) If pain interferes based on its salience and value, pain should affect performance on both low-value and high-value tasks, but more so on low-value tasks;

• (4) If pain does not interfere, it will have the same performance as the baseline and electric conditions.

Upon observing a sex difference in the linear mixed model, our hypotheses were that for males, pain should affect performance in a value-based manner, whereas for females, there should be no effect. We thus tested 2 hypotheses, the equations of which can be found in Table S2, http://links.lww.com/PAIN/C345:

• (1) Pain affects performance in a value-based manner in both males and females (no sex difference);

• (2) Pain affects performance in a value-based manner in males, but has no effect on females.

3. Results

3.1. Participants

Forty participants (20 females and 20 males, mean ± SD age = 24.88 ± 3.15 years) performed a cognitive branching paradigm comprising of 2 tasks: 1 low-value and 1 high-value task.⁸ Questionnaire data can be found in Table 1. For each participant, 6.17 ± 1.03 (mean ± SD) runs of the pain stimulation condition and 6.32 ± 1.02 runs of the electric stimulation condition were included in analyses (Table 2). The number of runs excluded per participant were 1.05 ± 1.15 for heat runs rated as nonpainful and 0.18 ± 0.45 for electric runs rated as painful (Table 2). The temperature applied across task runs for the pain stimulation was 38.42°C ± 5.8, and the current applied across task runs for the electric stimulation was 6.77 mA ± 2.6. Crucially, we matched the perceived salience of the stimulation conditions for each participant, and salience was defined to participants as “the extent to which a stimulus grabs and directs your attention during the task” as previously done.³³ To salience-match, the number of heat pain runs and electric runs removed was 0.75 ± 1.01 and 1.48 ± 1.04 (Table 2), respectively.

Table 1.

Group mean of self-report questionnaires.

Questionnaire	Mean (±SD)
Beck's depression inventory II	4.4 ± 4.7
State-trait anxiety inventory
State	29 ± 7.7
Trait	35.3 ± 11.6
Pain catastrophizing scale	10.8 ± 9.8
Rumination	4.3 ± 3.8
Magnification	2.6 ± 2.4
Helplessness	4.0 ± 4.3

Table 2.

Data Quality Assurance/Quality Control for task runs based on stimulation condition.

Stimulation condition	Mean (±SD)
Heat pain
No. of runs removed because of nonpainful report*	1.1 ± 1.1
No. of runs removed to salience-match	0.8 ± 1.0
Remaining number of painful runs out of 8	6.2 ± 1.0
Remaining number of trials	172.9 ± 29.0
Electric
No. of runs removed because of pain report†	0.2 ± 0.4
No. of runs removed to salience-match	1.5 ± 1.0
Remaining number of painful runs out of 8	6.3 ± 1.0
Remaining number of trials	177.1 ± 28.6

^*Participants completed 8 runs, which must all be painful in the pain condition. Nonpainful runs were removed.

^†Participants completed 8 runs, which must all be nonpainful in the electric condition. Painful runs were removed.

3.2. Salience-matching pain and electric stimuli

To determine whether pain–cognition interactions are driven by salience or by value, it was essential that each participant perceived heat and electric stimuli as equally salient. We confirmed that the salience was matched between stimuli for each participant (Fig. 5) using a Pearson correlation, the TOST, Cronbach alpha for reliability, and a Bayesian paired samples t test: salience ratings were significantly correlated (r = 0.99, P < 0.001), equivalent (TOST: P < 1e-17), reliably similar (α = 0.996), and their means did not significantly differ (BF₀₁ = 3.33, t₃₉ = −1.36, P = 0.18, Table 3). Therefore, salience ratings were matched across stimulation types, ruling out salience as a factor explaining differences in performance between stimulation conditions. In addition, we demonstrate there was no difference in salience ratings throughout the experiment between stimulation conditions (Table S3, http://links.lww.com/PAIN/C345), and in a separate cohort, the salience of capsaicin-heat pain remained persistent over time (Fig. S2, http://links.lww.com/PAIN/C345). Intensity ratings and unpleasantness ratings were also correlated (intensity: r = 0.75, unpleasantness: r = 0.82), but the means of the unpleasantness ratings were different (t = 7.53, P < 0.001), with the pain unpleasantness rated higher than the electric unpleasantness (Table 3). This is to be expected given that the electric stimulation was not painful, but despite the lower unpleasantness, it was still perceived as salient and intense as the heat pain.

Figure 5.

Average heat pain and nonpainful electric stimuli were salience-matched. (A) Average salience ratings in response to tonic heat pain (red) and tonic electric stimulation (blue), where each participant is represented by a horizontal line, indicating how closely matched salience ratings were between stimulation conditions. (B) Average pain and electric salience ratings were highly correlated, and the pink dashed lines reflect 95% confidence intervals.

Table 3.

Group mean ± SD ratings across task runs per stimulation condition.

Stimulation condition	Pain rating (numeric rating scale, /10)	Electric rating (numeric rating scale, /10)	r (P)	T (P)
Intensity	5.5 ± 1.3	5.2 ± 1.4	0.75 (<0.001)	1.82 (0.076)
Salience	5.2 ± 1.3	5.2 ± 1.4	0.99 (<0.001)	1.36 (0.18)
Unpleasantness	5.0 ± 1.6	3.8 ± 1.9	0.82 (<0.001)	7.53 (<0.001)

3.3. Reaction times: baseline comparisons

Mean RTs and error rates for each stimulation condition can be found in Table S4, http://links.lww.com/PAIN/C345. Comparing the pain condition to baseline, there was a significant reward-by-stimulus interaction (t = 2.64, P = 0.012) indicating that pain slowed RTs in the low-value but not the high-value task (Fig. 6A). Bayesian informative hypothesis evaluation demonstrated that pain interfered with competing tasks in line with a value-based hypothesis (BF₁₀ = 9.86, 69.98% posterior model probability, Fig. S3, http://links.lww.com/PAIN/C345). Comparing the electric condition to baseline, there was a main effect of reward (t = 7.66, P = 2.52e-09) but no main effect of stimulus (t = 0.66, P = 0.51), and no reward-by-stimulus interaction (t = 1.48, P = 0.14), indicating that the electric stimulus did not affect RTs (Fig. 6B).

Figure 6.

Reaction times of each stimulation condition compared to their respective baseline conditions. Average reaction times with 95% confidence intervals during (A) tonic heat pain in red and (B) tonic electric stimulation in blue compared to baseline in black.

3.4. Reaction times: pain vs electric

To assess the impact of stimulus type (pain vs innocuous electric) on task performance, we compared baseline-corrected RTs between conditions. We observed a significant reward-by-stimulus interaction (t = 3.38, P = 0.0016), where pain and electric stimuli impacted RTs differently based on task value (Fig. 7). Specifically, the RT difference between pain and electric stimuli was significantly greater in the low-value task compared to the high-value task. To further unpack this finding, we performed Bayesian informative hypothesis evaluation¹⁶ using the estimated marginal means of each condition and to determine whether priority was assigned between pain and the task based on value, salience/no effect, or the combination of value and salience. In this comparison, salience and no effect are the same hypothesis, as the salience-based hypothesis assumes that pain and electric stimuli affect performance to the same degree on both tasks compared to baseline, and thus no difference is expected in their RTs, which also aligns with the no effect hypothesis. We demonstrated support for the hypothesis that pain interfered in a value-based manner (BF₁₀ = 2.43, 70.03% posterior model probability, Fig. S4, http://links.lww.com/PAIN/C345).

Figure 7.

Pain slowed reaction times on low-value, but not high-value tasks, compared to an electric stimulus. Mean baseline-corrected reaction times and 95% confidence intervals are plotted during the pain (red) and electric (blue) stimulation conditions.

3.5. Reaction times: sex disaggregated baseline comparisons

To the best of our knowledge, there have been no findings of sex differences in the pain–cognition interactions literature. As such, we had no explicit hypotheses but doubled the sample size given by our power calculation to perform adequately powered sex disaggregated analyses as a secondary, exploratory aim, to determine whether there are sexually dimorphic behavioural strategies used. Comparing pain stimulation to baseline in males, we found a main effect of stimulus (t = 2.26, P = 0.036) and reward (t = 6.34, P = 3.88e-06), but no reward-by-stimulus interaction (t = 1.89, P = 0.073) (see Fig. S5a, http://links.lww.com/PAIN/C345). In females, we only found a main effect of reward (t = 3.55, P = 0.0021), but no main effect of stimulus (t = 0.22, P = 0.83), or reward-by-stimulus interaction (t = 1.85, P = 0.077) (see Fig. S5b, http://links.lww.com/PAIN/C345). Comparing electric stimulation to baseline, we found a significant reward-by-stimulus interaction in males (t = 2.16, P = 0.039; Fig. S6a, http://links.lww.com/PAIN/C345) and a main effect of reward in females (t = 5.77, P = 1.62e-05; Fig. S6b, http://links.lww.com/PAIN/C345) but no main effect of stimulus (t = 0.41, P = 0.69), or reward-by-stimulus interaction (t = 0.72, P = 0.48).

3.6. Reaction times: sex disaggregated: pain vs electric

We observed a significant reward-by-stimulus interaction (t = 3.72, P = 0.0014) in males, whereas there was no such interaction in females (t = 0.81, P = 0.43), indicating the interaction at the group level was likely because of males (Fig. 8A). Similar to the group level, we showed that the interaction in males follows a value-based hypothesis (BF₁₀ = 2.36, 70% posterior model probability). Females did not show any differences in performance during pain and electric stimulation for either task reward (t = 1.36, P = 0.19) (Fig. 8B).

Figure 8.

Pain interfered with performance in a value-based manner in (A) males but did not interfere at all in (B) females. Mean baseline-corrected RTs and 95% confidence intervals for heat pain (red) and electric (blue) conditions are plotted for (A) males and (B) females. RT, reaction time.

3.7. Reaction times: sex differences

Finally, to confirm sex differences, we ran another linear mixed model with sex as an additional fixed effect, and a Bayesian informative hypothesis evaluation.¹⁶ We found a significant reward-by-stimulus-by-sex interaction (t = 2.58, P = 0.014), indicating that the significant difference observed at the group level between pain and electric stimuli in the low-value task compared to the high-value task depends on sex (Fig. 8). Specifically, the interaction is stronger in males than in females. We ran Bayesian informative hypothesis evaluation to determine whether pain interfered with competing tasks in value-based manner in males and females, or in males only. We found strong evidence supporting the combination of a value-based hypothesis in males and a null effect in females (BF₁₀ = 15.6, 66.13% posterior model probability, Fig. S7, http://links.lww.com/PAIN/C345). We also observed that females were faster than males across all conditions (t = 2.21, P = 0.033) and did not have a main effect of reward (t = 0.31, P = 0.76).

4. Discussion

This study directly investigated whether pain–cognition interactions are driven by salience, value, or both. We found that pain interfered with performance in a value-based manner, affecting low-value but not high-value tasks, whereas salience-matched electric stimuli had no effect on RT. This effect was sex-specific: males showed value-based interference, whereas females showed no interference from either stimulus compared to baseline. These findings reveal previously unreported sex differences in priority assignment in pain–cognition interactions. Therefore, future studies of pain–cognition interactions should use a value-based framework, and to explicitly investigate sex differences.

The most pervasive model to explain pain–cognition interactions posits that pain competes with concurrent goals for limited attentional resources based on salience.²⁵ Our study addresses critical gaps in understanding priority assignment. Pain is an inherently attention-grabbing signal that warns of potential harm and initiates appropriate behavioural responses.¹¹ It is thought to interrupt ongoing cognitive processes through bottom-up attentional capture,²⁵ and top-down processes selectively shift attention to stimuli that are relevant to ongoing goals, thus regulating bottom-up processes.²⁷ There is a reciprocity of pain and cognition such that the allocation of attentional resources determines whether pain will be interruptive, or whether cognitive engagement will distract from and modulate pain perception.³² Accordingly, healthy adults who performed cognitive tasks while experiencing pain fell into 1 of 2 phenotypes: those who performed worse on the task, and those who performed better.^12,43 Nonetheless, it remains unclear what factors influence priority assignment of concurrent goals or pain, and whether such priority assignment can be extended to any salient stimulus when competing with concurrent goals. Unlike previous research using control stimuli that were not salience-matched,^{17,31,44,52,54} we used electric stimuli individually matched to heat pain for salience, ensuring performance differences were not salience-dependent. Although salience was self-reported, we have shown that self-report is a reliable measure for salience-matching.¹⁹ This novel approach—examining both salience and value in a single paradigm with properly matched stimuli—provides stronger evidence that value, rather than salience alone, drives pain–cognition interactions, with important sex-specific effects.

The motivation-decision model posits that pain elicits a motivational state to avoid and escape it, but amidst conflicting motivators (thirst, hunger), individuals must decide whether to prioritize escaping pain or pursuing goals, as these behaviors are mutually exclusive and inhibit each other.¹³ It is thought that if goal pursuit “wins out,” pain is inhibited, whereas when pain avoidance “wins out,” pain is facilitated. Although unclear how each option is prioritized, a task paradigm that is well suited to test this model is the cognitive branching paradigm, which has been used to delineate the neural underpinnings of complex problem-solving and concurrent goal pursuit.^8,22,34 Cognitive branching involves simultaneous tracking of 2 concurrent goals. Adding a third value-based goal leads to the deprioritization of the lowest-value goal. The valuation of pain has been shown to occur in a relative fashion and is context dependent: a painful stimulus of moderate intensity is assigned a different value depending on whether it is paired with a lower intensity or higher intensity painful stimulus.^51,58

The cognitive branching paradigm was ideal for testing our value-based framework, as it relies on relative task valuation. We hypothesized that if pain's perceived value is compared against competing tasks, it would affect low-value but not high-value task performance, whereas the electric stimulus would have no effect. We found that pain only slowed RTs in the low-value task, compared to a salience-matched innocuous electric stimulus, and a baseline condition. This indicates that pain does not interfere with concurrent goals in a salience-based manner: otherwise, pain would have affected RT on all tasks and the electric stimulus would have had the same effect. Our results are consistent with other studies demonstrating that greater task reward and difficulty reduce pain interference and perceived intensity, indicating that the cognitive task is prioritized.^4,5,26,52,53 In addition, the high-value task was performed quicker than the low-value task, consistent with previous cognitive branching findings.⁸ These findings follow the motivation-decision model, where there is a motivation to prioritize task performance and a motivation to prioritize pain, and the inhibitory effect they exert on one another depends on relative valuation. Another model, the motivational intensity theory, posits that task difficulty and the importance of success on a task determine the resources allocated.⁷ In cases where task difficulty is not clear, as in the current study, priority is assigned solely on success importance.⁴² Based on this framework, the monetary value of the low-value task would be deprioritized. Our findings are consistent with the motivation-decision and motivational intensity theories, and in line with studies that show that increased motivation shields against pain interference on task performance.^10,47,48,52

Surprisingly, we also found a sex difference in strategies used to assign priority. We found that males followed a value-based framework, whereas females did not. We discovered that female RTs were no different during stimulation and baseline—for either stimulus modality. Although there has been robust discussion of sex differences in pain,^21,29,36 this is the first evidence of sexually dimorphic strategies in pain–cognition interactions: our findings indicate that males use a value-based framework, whereas the result was inconclusive in females. It is possible that females had a floor effect in task performance (ie, RT) in the branching paradigm, therefore, not capturing effects of value or salience. Investigating error rates (see Supplementary Materials, http://links.lww.com/PAIN/C345), we found that in the electric condition, more errors were made in females, regardless of task reward, but this was not observed in the pain condition, or when comparing baseline-corrected pain vs baseline-corrected electric error rates. Reward did not affect error rates in females in any condition (in contrast to the effect of reward on RTs in females). In males, we found that the low-value task had more errors than the high-value task, regardless of stimulus type. It is thus possible that males and females prioritize different components in speed-accuracy trade offs: in our sample, males had slower RTs than females, overall; whereas females had an effect of stimulus on error rates, this was not seen in males. These findings are consistent with evidence that women and girls report using different pain management strategies than men and boys: the former report using emotional and cognitive reframing as well as social support, whereas the latter report using avoidance and distraction.^21,40 Although not assessed in the present study, it is possible that females used emotional and cognitive reframing, which was more effective at reducing the impact of pain on value-based cognitive tasks. Given that testing for sex differences in RTs was exploratory, although adequately powered based on our power calculation, future studies investigating pain–cognition interactions should assess sex differences in task performance—both RT and error rates, and use tasks that may overcome the potential floor effect observed in females.

Gender was not measured in this sample, and given evidence that gender roles and norms influence sex differences in pain thresholds, tolerance, and ratings,¹ future studies should collect self-report gender-related metrics, such as gender role expectations of pain⁵⁹ to determine whether this better explains sexual dimorphic findings. Future studies examining pain–cognition interactions should build on these findings by including value-based tasks and investigating sex and gender differences.

A limitation of the study is the complex, highly controlled experimental design. Strict salience-matching, the cognitive branching paradigm, and the capsaicin-heat pain model do not reflect a real-world experience of performing a task in the presence of pain and thus the results have limited ecological validity and generalizability. However, the setup was essential to control for factors that may drive priority assignment of pain–cognition interactions, and it is not possible to do so in real life. Second, although the cognitive branching paradigm promotes relative valuation, it does not allow for forced choice to better understand the motivational conflict of the differing monetary rewards. Given that the goal of the present study was to delineate between the role of salience and value in priority assignment, future studies should continue to investigate the interaction between pain and value using forced choice paradigms. Finally, although the capsaicin-heat pain model elicits sensitization, which commonly occurs in the chronification of pain and is thus considered a more ecologically valid pain model than electrocutaneous stimulation or acute heat, models that better approximate pain outside of the laboratory are needed. Future studies should assess more realistic models of acute pain, such as the orthodontic separators model,² and how they interfere with performance on cognitive tasks that involve decision-making and multitasking more relevant to daily life to better understand these dynamics in a more generalizable manner.

Our findings show that pain uniquely competes with concurrent goals, with priority assigned in a value-based manner, whereas a salience-matched nonpainful control stimulus had no effect on task performance. We further show that this effect was entirely seen in males. These findings highlight the importance of using a value-based framework and testing for sex differences when investigating pain–cognition interactions.

Conflict of interest statement

The authors have no conflicts to report.

Supplemental digital content

Supplemental digital content associated with this article can be found online at http://links.lww.com/PAIN/C345.

Supplementary Material

jop-166-2851-s001.pdf ^{(556.6KB, pdf)}

jop-166-2851-s002.pdf ^{(311.1KB, pdf)}