Healthy adults aged 18 to 35 were recruited as part of the Biomarkers of Stress States (BOSS) study run at the University of California, Santa Barbara with the goal of investigating how different types of acute stress (noxious stimulation, social stress, mental fatigue and physical fatigue) impact human physiology, brain and behavior. All participants were initially pre-screened via phone interview to determine their eligibility to participate. Participants were considered ineligible if any of the following criteria applied: heart condition or joint issues, recent surgeries that would inhibit movement, BMI>30, currently taking blood pressure medication or any psychostimulants or antidepressants. Participants also agreed to provide a blood sample and multiple saliva samples throughout the study. Participants were instructed to eat and drink in accordance with their typical daily routine and they were not specifically instructed to abstain from caffeinated beverages. They were instructed to abstain from strenuous exercise in the 48 h prior to each testing session. Informed consent was provided at the beginning of each session, and all procedures were approved by Western IRB and the U.S. Army Human Research Protection Office, and conformed to UC Santa Barbara Human Subjects Committee policies.

Participants were then invited to attend the lab to complete two preliminary testing sessions on two separate days. In session one, all participants underwent a trial 90 s CPT exposure (described in Section 3.3) to determine whether they were able to withstand the discomfort associated with the CPT, and participants who were unable to last the full 90 s were not allowed to participate in subsequent sessions (n = 1). In session two, participants completed a graded VO₂max test on a stationary exercise bike to assess their aerobic fitness capacity (Vyntus CPX Metabolic Cart and Viasprint 150P stationary bike, Vyaire Medical, Yorba Linda, CA, United States). VO₂max data are reported here, but were primarily collected as part of the larger BOSS study for another purpose.

A subset of n = 42 participants were then randomly allocated to the noxious stimulation (CPT) condition. Participant age, sex and physiological data are reported in Table 1. Participants were all members of the UC Santa Barbara campus community and were either current students (n = 34), had recently graduated and were in full-time employment (n = 5) or were unemployed (n = 4). The ethnic background of the sample was as follows: white/Caucasian (n = 16), Asian (n = 9), Hispanic/Latino/Spanish Origin (n = 6), black/African-American (n = 3), Middle-Eastern/North-African (n = 2), White/Hispanic (n = 3), White/Asian (n = 1), Hispanic/Asian (n = 1), other (n = 2). The total household annual income reported by participants was as follows: <$20,000 (n = 5), $20,000 to $34,999 (n = 4), $35,000 to $49,999 (n = 6), $50,000 to $74,999 (n = 5), $75,000 to $99,999 (n = 7), > $100000 (n = 10). Six participants declined to report annual household income. Data collected from participants assigned to other stress conditions as part of the larger BOSS protocol are not reported in this manuscript. All participants received compensation of $20 per hour for the time they spent in the lab, with an additional $25 bonus for completing the entire BOSS protocol.

Given that the main goal of the larger BOSS study was to investigate stress effects on cognitive performance and underlying physiology, sample size was estimated based on what is typical in the literature for cognitive tasks and not investigations into CPT effects per se. However, the present sample size exceeds the sample size of comparable CPT investigations. For example, two recent investigations validating the “both feet” version of the CPT used in this study sampled 24 participants (Larra et al., 2015) and 28 participants (Bachmann et al., 2018) and investigations into CPT habituation effects test between 5 and 28 participants (Glaser and Whittow, 1957; Glaser et al., 1959; Zbrożyna and Krebbel, 1985; Zbrożyna and Westwood, 1988, 1990; Stancák et al., 1996; Barati et al., 2013).

In sessions three, four and five, participants attended the lab and completed a sequence of cognitive tasks while physiological and neural measures were recorded. Each session was completed on a separate day. Session three was a baseline recording (i.e., no stress manipulation) whereas session four involved a stress manipulation and session five involved an active control manipulation. Sessions four and five were completed in a counterbalanced order (order was randomly assigned based on participant number). In this paper, data are reported from sessions four and five only.

Each participant started both sessions at either 8a.m. or 1p.m. On both days, participants were informed at the start of the session which condition they would be exposed to (i.e. treatment or control) and that there would be five cold water exposure (CPT) or five warm water exposure (warm pressor test; WPT) trials in total. Participants were then prepared for the physiological recording by fitting disposable foam electrodes (EL500, BIOPAC Systems Inc., Goleta, CA, United States) to the neck and torso. Prior to the placement of each electrode, an approximate one-inch area of skin was disinfected and gently exfoliated with an abrasive pad, followed by application of Nu-Prep exfoliation gel (EVPREP, BIOPAC). The area was fanned dry and then a small amount of electrode gel (GEL100, BIOPAC) was placed on each electrode before being placed on participants. For collection of impedance cardiography (ICG), 8 electrodes were placed on the torso and neck, two on each side of the neck and two on each side of the torso (Bernstein, 1986). For ECG, two electrodes were placed on the chest, one under the right collarbone and one under the left rib cage. A total of 10 electrodes were applied.

Participants were then fitted with an EEG cap consisting of 63 scalp electrodes arranged in accordance with the 10–20 system (BrainAmp MR, Brain Products, Berlin, Germany) and positioned at a desk in front of a computer monitor (120 cm viewing distance). A CNAP Monitor 500 (BIOPAC) blood pressure cuff was attached to the right upper arm and a finger cuff to the middle and index fingers for continuous blood pressure (CBP) monitoring. An eye-tracker (Eyelink 1000 Plus, SR Research, Ltd., Mississauga, Ontario, Canada) was positioned on the desk at ∼60 cm distance from the participant’s eyes and an infrared camera (FLIR Systems, Goleta, CA, United States) was positioned at ∼65 cm distance. The infrared camera data were recorded for another purpose (Kumar et al., 2021). The participant rested on a chin support (SR Research, Ltd.) in order to stabilize the head. A fully instrumented participant is shown in Figure 1A.

Participants then underwent five CPT immersions in the treatment session and five WPT sessions in the control session. The CPT/WPT protocol is detailed in Figure 1B and described in the next section. Each individual CPT/WPT immersion within a session is referred to as a “Trial”, using T1-T5 as shorthand to denote Trials 1–5. Session order was counterbalanced between participants. Immediately following each trial, participants engaged in a cognitive task as part of the BOSS protocol (data pertaining to the cognitive tasks were collected for another purpose and are not reported here). Each cognitive task had a different duration, and time to produce each of the saliva samples varied between subjects, so time between CPT/WPT exposures was variable (mean = 24 m 55 s, SD = 310 s). The variability in time between CPT/WPT exposures and the potential for the cognitive tasks to impact the stress response are limiting factors of this dataset, and this is explored in the “Limitations and Future Directions” section of the discussion. The complete experimental session structure is outlined in Figure 1C.

Participants were allowed to drink water during the protocol and they were allowed to take bathroom breaks in between tasks/CPT exposures.

A version of the CPT/WPT where participants immersed both feet in water was chosen for two key reasons. First, participants needed to be able to use both hands immediately after the termination of the CPT to make responses to the cognitive tasks, hence immersing the hands was not suitable as numbness in the hands may impair task performance. Second, recent evidence suggests that the stress response may be enhanced when participants are required to immerse both feet in water when compared to single arm (Larra et al., 2015). This version of CPT has been used in several recent studies and is well validated (Bachmann et al., 2018; Zhang et al., 2018).

For each trial, a large metal bucket was positioned just in front of the participant’s feet and a towel was placed under the feet. A digital thermometer was positioned in the bucket. For treatment sessions, the bucket was pre-filled with cold water and ice at least 15 min prior to the first exposure (∼33 °f/0.6°C) whereas for control sessions the bucket was pre-filled with lukewarm water (∼92 °f/33.3°C). The bucket was filled to ∼10 cm deep. The water was stirred between trials and measured between trials to ensure that a consistent temperature was maintained throughout the session. The temperature in the laboratory was thermostatically maintained at ∼ 22°C during the protocols.

A blood pressure cuff was fitted to the right hand/arm and calibrated prior to starting the recording. A fixation cross was then presented on screen. The participant was instructed to fixate on this cross throughout the duration of the experiment and remain as still and relaxed as possible throughout the entire procedure. The protocol was then initiated (Figure 1B). The protocol started with a baseline period (40 s) where the participant was just required to relax and maintain their gaze on the fixation cross. Next, a tone sounded and a written instruction appeared asking the participant to position their feet on the edge of the bucket in preparation for immersion (25 s). A tone then sounded and the participant was instructed to immerse their feet in the bucket and try to relax and stay as still as possible (90 s). At the end of the immersion period, another tone sounded and the participant was instructed to remove both feet from the bucket and place them on a towel for the recovery period (40 s) and try to relax and minimize any further movement. At the end of the recovery period the participant was asked to report the amount of pain they had experienced on a scale from 0 (no pain) to 100 (extremely painful). Upon completion of the procedure the participant was then given the opportunity to dry and wrap their feet in a dry towel for comfort, until the next trial. Importantly, during the baseline, CPT and recovery periods participants were instructed to keep their feet flat on the ground or the bottom of the ice bucket and to remain as still as possible in order to obtain steady and reliable blood pressure, ECG and impedance cardiograph readings.

During each trial, EEG was recorded with 63 scalp electrodes at 1000 Hz sampling rate (BrainAmp MR, Brain Products). ICG, ECG and CBP were sampled at 4000 Hz with an MP150 Amplifier and displayed and stored on a laptop running AcqKnowledge Software Version 5.0.2 (BIOPAC Systems, Inc.). Binocular pupillometry was sampled at 500 Hz (Eyelink 1000 Plus, SR Research). Saliva samples were collected at three points throughout each session: one at the beginning of the experiment, one after the second CPT/WPT and one after the final CPT/WpT. Samples were collected using the passive drool technique, where the participant was required to salivate through a straw into a vial. Saliva samples were transferred to a freezer (-80°C) at the end of each session.

All data are were analyzed and interpreted under a common framework. First, for each measure, two-way ANOVAs were computed to determine whether there were any main effects of condition, habituation across trials, and any condition x trial interactions. Significant interactions were then explored by computing pairwise comparisons to investigate intra-session habituation across trials. First, T1 and T5 are compared, to determine whether the stress-response habituates over the entire session. Second, to gain a sense of whether these habituation effects are driven by habituation across earlier or later exposures within the session, T1 and T3 are compared (early-session habituation) and T3 and T5 are compared (late-session habituation). All statistical tests were conducted using a non-parametric permutation-based resampling approach (Section 2.7).

For cardiovascular and pupillometry measures specifically, the appropriate tests are computed for each second of trial data, to gain insight into the time course of intra-trial stress and habituation dynamics. For the time-course analyses, ANOVA and t-test outcomes where p _null<0.05 are represented by the presence of colored horizontal bars superimposed onto the plots, where the presence of a bar at a given time point indicates a significant difference. ANOVA outcomes are presented in-between the treatment and control plots, and pairwise comparison outcomes are presented at the base of the treatment and control plots for timepoints where a significant condition x trial interaction is observed, where each color pair represents a different comparison i.e., red/blue = T1 vs. T5 (habituation across entire session), red/yellow = T1 vs. T3 (early -session habituation), yellow/blue = T3 vs. T5 (late-session habituation). The goal of analyzing the raw data was to study stress effects that occurred prior to foot immersion (i.e., 0–65 s), hence the “immersion” and “recovery” periods are greyed out in the raw plots and statistical comparisions are not reported for these time periods. The goal of analyzing the baseline corrected data was to study stress effects evoked by foot immersion (i.e., 65 s–195 s), hence the “baseline” and “prep” periods are greyed out in the baseline corrected plots and statistical comparisions are not reported for these time periods.

To address whether a change occurred within either condition for all the cardiovascular variables we computed additional three-way ANOVA analyses on the raw data that incorporated change in baseline as a factor, comparing data recorded during the pre-CPT baseline sessions with data recorded during and after the CPT exposures. In our main baseline-corrected analyses we computed baseline by averaging over the final 15 s of the baseline period, so for consistency we also applied a 15 s moving window average to the data recorded during and after the CPT (i.e. 65-155 s). We then computed a 2 [baseline: pre-CPT, during/post CPT] x 2 [condition: treatment, control] x 5 [trial: T1>T5] ANOVA for all timepoints ranging from 65–140 s. Consistent with our main analyses, the complete results from the ANOVAs are plotted with the horizontal bars between the figures representing significant main and interaction effects (Supplementary Figure S1).

For the EEG data, baseline uncorrected ERSP data are presented averaged across the baseline period (1–40 s) and baseline corrected ERSP data averaged across early (65–94 s), mid (95–124 s), late (125–154 s) and recovery (155–190 s) periods. As before, inspection of the baseline uncorrected data provides insight into any potential changes in brain oscillations that occur prior to exposure, and inspection of baseline corrected data allows us to look at changes evoked by exposure. The data were also parsed into different frequency bands associated with cognitive function i.e. delta [1–3 Hz], theta [4–7 Hz], alpha [8–14 Hz] and beta [15–30 Hz]. A 2 [stress condition] x 5 [trial] ANOVA was then computed for each scalp electrode for each time period and frequency band, and any robust condition x trial interactions were computed by averaging across all electrodes that contribute to the interaction and computing pairwise comparisons as described above.

Statistical significance in all tests (except the rank order analysis) was assessed by using a non-parametric permutation-based resampling approach to empirically approximate null distributions for F and t statistics (Foster et al., 2016; Bullock et al., 2017; Bullock et al., 2021). This testing approach was adopted because it has the advantage of being robust to violations of normality. Specifically, condition labels were shuffled within participants and 1000 iterations of the appropriate statistical tests were ran (repeated-measures ANOVAs and/or paired-samples t tests) to generate null distributions of F and t statistics. Once null distributions had been obtained, reliable differences were tested for by computing the probability of obtaining F and t statistics from each of the null distributions that were greater than the observed F and t statistics. The F and t statistics are reported in the text along with the critical p-value (labelled p _null) which represents the probability of observing a value greater than this in the null distribution. For the time-course data analyses (Cardiovascular Measures, Pupillometry) ANOVA and t-test outcomes where p _null<0.05 are represented by the presence of colored horizontal bars superimposed onto the plots, where the presence of a bar at a given time point indicates a significant difference. ANOVA outcomes are presented in-between the treatment and control plots, and pairwise comparisons are presented at the base of the treatment and control plots for timepoints where a significant condition x trial interaction is observed (see Section 3.2 for detailed description). For self-report, EEG and cortisol analyses, test results are reported in the text. Here, to gain a more precise estimate of the observed statistic’s position in the null distribution, tests are reported as p _null<0.05, p _null<0.01 and p _null<0.001. A test result of p _null>0.05 indicates that the result was not statistically reliable. The time-course analyses presented here rely on repeated comparisons at multiple timepoints, which raises the possibility of increased inferential error, however, the effects that are described are present across multiple timepoints and no inference relies on a single comparison but rather a consistent pattern across time.

All custom analysis scripts are available via GitHub (https://github.com/attlab/CPT_Adaptation). Data are available upon request.

Participants reported significantly greater pain immediately after treatment trials when compared to control trials [F (1,37) = 392.54, p _null<0.001, η _p ² = 0.91] (Figure 3). There was also an interaction between condition and trial [F (4,148) = 8.06, p _null<0.001, η _p ² = 0.18]. Pairwise comparisons revealed that this interaction was driven by reduced pain ratings as a function to repeated exposures over the course of the treatment session [T1>T5; t (37) = 3.41, p _null<0.001] and this was driven by late-session habituation [T3>T5; t (37) = 3.33, p _null<0.001] not early-session habituation [T1>T3; t (37) = 1.09, p _null>0.05]. There was very little variability in pain ratings in control trials, with most participants rating all five trials as inducing no pain (rating = 0), so pairwise comparisons were not computed for control trials.

All cardiovascular and pupillometry measures are presented in both raw and baseline corrected format and analyzed using a common framework (see Section 2.6). Resting baseline levels for all cardiovascular measures are reported in Table 1. Note that these resting recordings were taken from the final 15 s of the baseline period in the first control exposure (i.e., lukewarm water) which. We reason that participants were in a relatively relaxed state at this time given that they were explicitly aware the forthcoming exposure was to lukewarm water; however, it is possible that anticipation of the exposure and elevated stress levels from being in the lab and having sensors applied to the face and body may be associated with elevated ABP and increased physiological arousal. Hence, these values may deviate from a true resting baseline.

Analyses of anticipatory and evoked stress effects on neural oscillations are presented in Figures 9–11. A common analytical framework was applied to the EEG data that was similar to the framework applied to cardiovascular and pupillometry data (see Section 2.6 for a detailed account), but for data reduction purposes the EEG data are parsed into key frequency bands (delta, theta, alpha, beta) and time periods (baseline, early-immersion, mid-immersion, late-immersion and recovery).

Cortisol was elevated during treatment relative to the control session [F (1,41) = 4.74, p _null<0.05] and declined over the course of the experiment from Sample 1 (Pre T1) to Sample 3 (Post T5) in both conditions [F (2,82) = 14.07, p _null<0.001]. Condition and sample did not interact [F (2,82) = 2.13, p _null>0.05] (Figure 12A). Circulating cortisol levels fluctuate throughout the day, typically exhibiting a sharp increase in morning hours and then a gradual decline throughout the day until bottoming out during nighttime sleep. In the present study, a participant’s treatment and control sessions either both started at 8 a.m. or 1 p.m. and this likely affected baseline cortisol levels, which may also impact cortisol reactivity to the CPT/WPT trials. To assess this possibility, a mixed-measures ANOVA was computed, with start time [AM, PM] as the between-participants factor and stress condition [treatment, control] and sample [Pre T1, Pre T3, Post T5] as within-participants factors. There were no significant effects (all p _null > 0.05; Figure 12B).

Exposure to both cold and warm water evoked an initial increase in HR, MAP and CO, reduction in SV and TPR, and shortened PEP and LVET. This response was increased as a function of cold water relative to warm water during the immersion period for several variables (HR, CO, SV, MAP, PEP and LVET). The observed increased sympathetic activity and decreased parasympathetic activity (decreasing HF throughout immersion) are consistent with an alerting/defense response that is evoked by many different forms of environmental and mental stressors (Hilton, 1982; Zbrozyna and Westwood, 1993) as well as previous investigations that use the CPT to induce stress in human participants (e.g., Lovallo 1975; Larra et al., 2015; Bachmann et al., 2018). The increase of HR, CO and MAP in the setting of reduced TPR early in the immersion period is notable. Under resting conditions, a rise in MAP, regulated through baroreceptors would evoke a fall rather than a rise in HR and a reduction in TPR. However, the baroreceptor reflex is centrally inhibited during the alerting/defense response (Mifflin et al., 1988) and is known to decrease in sensitivity with mental stress (Conway et al., 1983; Steptoe and Sawada, 1989), so this may explain why MAP and HR rise together. The initial reduction in TPR in response to cold water could be driven by vasodilation in some muscle beds and vasoconstriction elsewhere as mediated by the alerting/defense response. However, this cannot be disambiguated in the current experiment from a second thermoregulatory response to cold. Cold exposure to the skin evokes splanchnic and renal vasodilation, which could reduce global TPR (Morrison and Nakamura, 2011). The reduction of SV is also notable as it would be expected to rise with an alerting/defense response. We speculate that it is a consequence of a mixed stressor/thermal stimulus induced by CPT leading to a potential redirection of blood volume away from skin and towards the splanchnic circulation. This could reduce venous return to the heart, end diastolic volume and SV.

Cold-water exposure also evoked rapid pupil dilation within 10 s of exposure, followed by sustained dilation in early but not later trials. Pupil diameter changes are controlled by the dilator and sphincter muscles, which are influenced by sympathetic and parasympathetic branches of the ANS, respectively (Steinhauer et al., 2004; Bradley et al., 2008), hence these data help to further characterize the nature of the autonomic response to pain. The initial dilation is consistent with previous single-exposure data that demonstrate pupil dilation within 30 s of cold-water exposure (Tassorelli et al., 1995; Tassorelli et al., 1998). Pupil dilation was then sustained through to the end of the cold-exposure and part of the recovery period. In previous pupillometry studies the pupil returned to baseline 1–2 min into the cold-water exposure (Tassorelli et al., 1995; Tassorelli et al., 1998), whereas the present data show a sustained response in early trials. This may reflect a stronger response to immersing both feet in the present study when compared to immersing a single hand, as per the previous work (Tassorelli et al., 1995; Tassorelli et al., 1998).

Midway through the cold-exposure timeline, changes became apparent in oscillatory brain activity. Cold-water exposure evoked an increase in frontal delta and theta activation. These patterns of activation are broadly consistent with previous work showing increased bilateral frontal delta and theta during a single CPT exposure (Ferracuti et al., 1994; Chang et al., 2002), although delta effects are noticeably more left-lateralized in the present data. These results support the notion that low-frequency EEG activity is a characteristic of the response in the human brain to the pain associated with the CPT (Knyazev, 2012), and the topography of the activation overlaps with hemodynamic imaging studies that show increased activation of pre-frontal brain regions as a function of pain (Casey, 1999; Frankenstein et al., 2001). The changes in low-frequency oscillatory activity were sustained throughout the cold-water exposure. Together, these EEG results replicate some previous findings and suggest that multiple frequency bands are involved in pain processing.

Several cardiovascular variables habituated across the repeated cold-water exposures. HR was the first cardiovascular measure to habituate, in that the increase in HR was already reduced in Trial 3 relative to Trial 1, whereas the increase in MAP and CO largely habituated in later trials (i.e., Trial 5 relative to Trial 3). The habituation in both the shortening of the LVET response and the drop in SV in response to repeated cold water immersions also occurred in later trials (i.e., Trial 5 relative to Trial 3) The habituation of HR and MAP observed here is consistent with previous work in humans showing reduced HR and SBP/DBP responses to repeated exposure of the hand or foot to cold water (Glaser and Whittow, 1957; Glaser et al., 1959; Zbrożyna and Krebbel, 1985). The increase in pupil diameter also habituated early in the session (i.e., Trial 1 vs. Trial 3; note that while the habituation effect was numerically larger during treatment when compared to control, there was still significant habituation in both). Frontal-midline theta activity declined over the course of both cold- and warm-water exposures, perhaps reflecting reduced cognitive control requirements over the course of each session as participants become increasingly familiar with task demands (Cavanagh and Frank, 2014). The EEG alpha band was also modulated during repeated cold-water exposure towards the end of the exposure and recovery periods, demonstrating a complicated pattern of synchronization and desynchronization over the course of the session. Finally, subjective ratings of pain in response to cold exposure declined over the course of the study, with most of the habituation occurring between Trials 3–5. This habituation in self-reported pain is broadly consistent with previous work showing decreased pain thresholds over repeated cold exposures within a single-day session (Smith et al., 2009) and also across sessions ran on multiple days (Carman and Knight, 1992).

Changes in pupil diameter and in oscillatory brain activity were observed in the period immediately prior to cold-exposure. There is evidence to suggest that pupillary activity is a physiological marker of emotional regulation (Siegle et al., 2003; Urry et al., 2006; van Reekum et al., 2007) and pupil size is also a reliable indicator of cognitive effort during the anticipation of events, with increased effort being associated with larger pupil diameter (Moresi et al., 2008) for both aversive and neutral events (Bitsios et al., 2004). The increased pupil diameter observed here prior to cold water exposure may therefore reflect a combination of anticipatory arousal and increased cognitive effort as the participant mentally prepares themselves for the forthcoming cold-induced discomfort. There was also suppression of alpha and beta activity across central EEG electrodes. This may indicate disinhibition of somatosensory cortex, which typically characterizes activated cortical regions that are preparing to process information or initiate a movement (Pfurtscheller et al., 1997), thus suggesting a link between somatosensory disinhibition and anticipation of pain. This is broadly consistent with previous work showing alpha suppression in response to observing or imagining pain in others (Whitmarsh et al., 2011; Hoenen et al., 2015), so it is possible that in the present study alpha is consistently suppressed prior to cold-water exposure trials due to participants imagining their own forthcoming pain. This effect does not habituate with repeated pain exposure.

When considering the anticipatory effects reported here, it is important to keep in mind that participants were informed at the beginning of each session which condition they would be exposed to (i.e., treatment or control) and that there would be five exposures total. Participants also had prior experience with the CPT from the initial calibration session, so in the treatment condition the knowledge that they were about to undergo five CPTs likely increased their anticipatory stress levels throughout the entire session. Furthermore, the repeated exposures to the CPT may have compounded anticipatory stress effects throughout the session.

Changes across several cardiovascular measures were also observed in the time-period prior to both cold and warm water exposures. While it is possible that changes in these pre-exposure measures may in part be driven by anticipation of the forthcoming exposures, any anticipatory effects from the second trial onwards are likely conflated with carry-over responses from previous trials, so we cannot make any strong statements about anticipatory stress effects per se.

Elevated cortisol at the outset of both treatment and control conditions and prior to any CPT/WPT exposures indicated that anticipation of both sessions induced stress in participants. Prior experience with the CPT in an earlier calibration session, as well as the ∼ 1 h of instrumentation at the beginning of the session may have contributed to the stress. The lack of interaction between treatment and control in later samples indicates that cortisol was not sensitive to repeated CPT exposures, which is broadly consistent with previous work demonstrating no cortisol response to a single unilateral arm CPT exposure (McRae et al., 2006; Duncko et al., 2007; Schwabe et al., 2008) and suggests that repeated CPT exposures do not activate the HPA. However, it has been previously demonstrated that a single 3-minute CPT (both feet) can evoke a robust cortisol response (Bachmann et al., 2018), so it is possible that the 90 s CPT exposures used in the present study were not sufficient to evoke a strong HPA response. Furthermore, it is possible that elevated stress at the beginning of both sessions may have masked any CPT evoked responses. Experimental designs intending to use cortisol as an index of HPA activation by the CPT should be mindful of lab-visit induced stress effects as well as CPT duration.

It is important to acknowledge that the present dataset was collected as part of a larger study into the effects of stress on the human brain and behavior, so the CPT protocol and suite of measures were optimized for the main goals of that study i.e., investigating how CPT exposure impacts subsequent performance across a battery of cognitive tasks and the physiological correlates of any adaptive behavioral changes. There are several limitations associated with this dataset and our findings that must therefore be discussed.

First, the cognitive tasks completed by participants between CPT/WPT exposures as part of the larger protocol may have confounding effects on the stress responses and this may have impacted results presented here. Furthermore, the duration of the cognitive tasks varied, meaning that the recovery time between CPT/WPT exposures was not entirely consistent throughout each session, which may have also impacted the results. Second, participants were also not asked to modify their consumption of caffeinated beverages prior to testing sessions, as the main testing sessions were long and it was a concern that requiring them to abstain from caffeine may lead to excessive sleepiness with detrimental effects on cognitive task performance. However, it is important to acknowledge that differences in caffeine levels between individuals may influence the cardiovascular responses to the CPT. Third, no attempt was made to test female participants at a specific stage of the menstrual cycle and that variability in estrogen and other hormone levels may impact our results. Fourth, given the scale of the larger BOSS study and the length of each testing session, it was necessary to test each participant in either morning sessions starting at 8 a.m. or afternoon sessions starting at 1p.m. Many cardiovascular variables such as HR, HF, BP and CO are known to have circadian rhythms (Millar-Craig et al., 1978; Furlan et al., 1990; Yamasaki et al., 1996; Guo and Stein, 2003) so it is possible that the variability in baseline values may have affected both anticipatory and evoked responses to the CPT manipulation. Fifth, the repeated measures design led to systematic baseline trends in several cardiovascular variables and the precise source of these trends is difficult to determine. Our main analyses were baseline corrected, but it is still possible that these trends may have impacted the outcome of our results. Despite these limitations, the unique combination of measures and multiple CPT exposures within a session provide a novel and valuable opportunity to study stress habituation across multiple physiological and neural responses.

We also acknowledge that there is growing interest in establishing which cortical regions have the most influence over ANS, and anatomically there is evidence from human fMRI and intercranial EEG studies that regions such as insular and anterior cingulate cortex are major contributors to both sympathetic and parasympathetic responses (e.g. Macey et al., 2012; Cechetto 2014; Pal et al., 2021; Roquet and Conti 2021). Here, we do not attempt to explore any direct relationship between our scalp recorded EEG measures and ANS modifications as this is a complex question that is beyond the scope of this paper, but we are interested in exploring this relationship in future analyses of our data.

The goal of the present study was to interrogate the multiplex of physiological and neural responses to repeated cold- or warm-pressor tests using an extensive suite of measures. The findings elucidate the time-course of stress effects and habituation over repeated exposures in multiple cardiovascular, pupillometry and EEG measures, providing valuable insight into anticipatory and evoked stress effects on the ANS and CNS.