The experiment implemented a randomized crossover design consisting of a familiarization trial and 2 experimental trials (Figure 1, see details below). Experimental trials were separated by at least one week to reduce the potential for heat acclimation and were performed at the same time of day for each participant to control for circadian fluctuations in core temperature. Participants were instructed to avoid vigorous exercise and alcohol consumption for 24 h, caffeine for 12 h prior to each experimental session, and to follow their typical meal and hydration practices 24 h prior to each session. Randomization was performed online (random.org) and participants were blinded to the trial type being performed.

Upon arrival to the laboratory, anthropometric measurements (height, mass) and body fat percentage (calculated using the 7-site skinfold technique) were recorded. Participants then completed the Cognitive Failure Questionnaire, which is a 25-item questionnaire that is a self-evaluative measure of general fluid intelligence and related to four factors of absentmindness (memory, distractibility, blunders, and names; Broadbent et al., 1982) where total scores can range between 0 to 100 and the average score is between 19 and 35. Participants were excluded from the study if their score was ≥45, as this score indicates considerable difficulties in completing tasks that require vigilance and may be easily distracted, which could potentially influence the cognitive function data. Of 12 participants initially recruited, one was excluded based on Cognitive Failure Questionnaire score and did not perform the experimental assessments. Next, cerebrovascular reactivity to CO₂ was assessed through using a custom-built end-tidal forcing system. The cerebrovascular reactivity assessment protocol comprised a 5-min baseline (eucapnia), a subsequent 5-min at hypercapnia (+10 mmHg from eucapnia) followed by 5-min at hypocapnia (−10 mmHg from eucapnia), where reactivity was determined by the change in CBF per change in P_etCO₂. Next, participants performed a cognitive test battery (CTB; see details below) in a thermoneutral environment (~22°C, 30% RH) for a total of 3 times to reduce the possibility of a learning effect from repeated exposures (Wallace et al., 2017, 2021). Lastly, an incremental test to exhaustion was performed on a cycle ergometer (Velotron, RacerMate Inc., United States) to determine peak oxygen consumption. The test began with a standardized 5-min warm-up at 100 W, followed by workload increase of 25 W each minute until exhaustion. Peak oxygen consumption was defined as the highest 30-s value measured breath by breath from expired gases collected through a soft silicone facemask connected to an online gas collection system.

Upon arrival, participants voided their bladder and nude body mass (kg) was recorded. A sample of the urine was tested for urine specific gravity (PAL-10S, Atago, Japan) to determine hydration status. Participants were considered euhydrated if USG was ≤1.020, or else the test was rescheduled (ultimately, no trials were rescheduled from hypohydration). Participants were then instrumented and fitted with a two-piece liquid conditioning garment (BCS 4 Cooling System, Med Eng, Canada) consisting of 1/8″ diameter Tygon tubing sewn into a stretchable jacket and pant; the head, hands and feet uncovered. Participants then rested in a semi-recumbent seated position (~5-min) and a baseline measure of P_etCO₂ was recorded. Next participants performed the BASE measure of the CTB (see details below) in a thermoneutral room (~22°C, 30% RH) with no temperature manipulation. Following BASE, participants were fitted with a polyvinyl rain suit and thermal blanket over the liquid conditioning garment with the hands and head uncovered to minimize evaporative heat loss for the rest of the heating protocol.

Similar to a previous study (Wallace et al., 2021), four time-points were tested manipulating both physiological and perceptual thermal strain (Figure 2). The first time-point was BASE where no temperature manipulation occurred. Next, to delineate the sensory displeasure of hot skin separate from increases in core temperature, ~49.0°C water was circulated at 2.5 l·min⁻¹ through the liquid cooling garment and the next CTB was performed once a mean skin temperature ( T ¯ skin ) of ~37.0°C was achieved, creating the NC-HS time-point. To test the effects of hyperthermia on cognitive performance, passive heat stress was continued until there was a rise in rectal temperature by ~∆ + 1.5°C creating a HC-HS time-point; a ∆ + 1.5°C change in T_re was used to ensure a sufficient thermal strain to induce a hyperventilatory hypocapnia response. Additionally, based on pilot work and unpublished reports, our passive heating protocol has interindividual variability in thermal tolerance >38.5°C and led to alterations in ventilatory breathing patterns causing difficulties to clamp P_etCO₂. Therefore, a ∆ + 1.5°C ensured thermal tolerance between all participants and allowed for better control of P_etCO₂ using the end-tidal forcing system. Lastly, upon completion of the CTB, to test the effects of hyperthermia without sensory displeasure of hot skin, ~15–20°C water was circulated through the liquid conditioning garment until T ¯ skin of 35.5°C while minimizing changes in core temperature, creating the HC-CS time-point.

To determine the role of P_etCO₂ on cognitive function during passive heat stress, participants performed identical experimental trials that differed only based on inspired gas manipulation: i) a poikilocapnic (POIKI) condition where P_etCO₂ was recorded but not manipulated; and ii) an isocapnic (ISO) condition where P_etCO₂ was clamped to eucapnic levels determined at BASE. Participants were fitted with a soft silicone facemask (Hans Rudolph, United States) attached to a T-shaped two-way non-rebreathing valve connected to an online gas collection system (ML206 Gas Analyzer, AD Instruments; United States). Measures of inspired ventilation ( V ˙ I , L·min⁻¹) and breathing frequency ( f ˙ b r , breaths·min⁻¹), were measured using a pneumotach. Inspired and expired fractions of CO₂ were sampled at 1 kHz to determine breath-by-breath P_etCO₂ and were sampled through nafion tubing in a desiccant-packed drying chamber at a flow rate of 200 ml·min⁻¹. A custom built end-tidal forcing system (for details see: Hartley et al., 2016) was used to manipulate P_etCO₂ throughout the experimental trials by manipulating inspired CO₂ concentration to ‘force’ P_etCO₂ towards the desired value. Gas concentrations were controlled through solenoid valves which independently controlled gas flow from cylinders of compressed medical grade breathing air (20.93% oxygen, 0.03% CO₂, balance nitrogen), and 100% CO₂. Inspired air volumes were delivered to an air reservoir (~5 l) via a humidification chamber (~500 ml). Both temperature and humidity of the gas was regulated to replicate ambient room conditions through the end-tidal forcing system.

As a separate quality control for determining the accuracy and signal drift of our gas analyzer over the course of the experimental duration, CO₂% and ventilation volume were tested every 60-min over a 180-min period on one participant under thermoneutral conditions using ventilation rates of 10–20 l·min⁻¹. The average CO₂ signal drift was ∆ + 0.01 ± 0.01% CO₂ compared to room air, ∆–0.01 ± 0.01% CO₂ compared to the calibration gas, and volume was ∆ + 0.01 ± 0.025 l over the 180-min period. Overall, minimal CO₂ signal drift over the course of the experimental trials (< 3 h in duration) ensuring accuracy of the end-tidal forcing system in clamping P_etCO₂ throughout the experimental protocol.

Before the commencement of BASE, participants were instrumented with a flexible thermistor (Mon-A-Therm Core, Mallinkrodt Medical, United States), self-inserted 15 cm beyond the anal sphincter to measure rectal temperature (T_re) sampled at 4 Hz. Four thermocouples (VC-T-24-190 Omega Environmental Inc., Canada) were used to determine weighted T ¯ skin at four sites defined as T ¯ skin = 0.3 a r m + 0.3 c h e s t + 0.2 t h i g h + 0.2 c a l f on the right side of the body sampled at 4 Hz. Heart rate was calculated using R-R intervals using a standard three-lead electrocardiogram (MLA2340, AD Instruments; United States). Bi-lateral middle cerebral artery velocity (MCA_v) was assessed using a 2 MHz pulsed transcranial Doppler ultrasound system (Doppler-Box, Compumedics GmbH, Germany). The probes were positioned over the temporal window and were held in place using a secure and comfortable head frame (M600 Headframe, Spencer Technologies, United States). To identify and optimize signals, positioning of the probes were determined by using techniques described by Willie et al. (2011). Data was visually inspected and removed, post-hoc, if significant artifacts were detected (e.g., movement during CTB) or poor signal waveform quality. Systolic blood pressure (SBP) and diastolic blood pressure (DBP) were taken from a manual sphygmomanometer (Aneroid Sphygmomanometer, Welch Allyn Hillrom, United States) on their left arm by the same experienced researcher before the CTB at each time-point. Mean arterial pressure (MAP) was calculated as: M A P = ( ( 2 * D B P ) + S B P ) / 3 . All physiological data (except MAP) were continuously sampled throughout the experiment and are the average over the course of the entire CTB. Subjective assessments of the environmental conditions were assessed using a 1–4 scale to measure thermal comfort and a 1–7 scale for thermal sensation (Gagge et al., 1967) and were collected upon the completion of the CTB at BASE, NC-HS, HC-HS, and HC-CS.

To measure progressive changes in cognitive performance, a ~ 12-15 min customized CTB (CogState, United States) was performed at BASE, NC-HS, HC-HS, and HC-CS. This customized CTB focused on the assessment of executive function and working memory. Executive function is an umbrella term for a collection of higher-order cognitive processes and is the ability to plan and execute behavior while being able to dynamically update goals, store and acquire information in working memory, switch among tasks, inhibit behaviors, process errors and being able to perform these behaviors in changing environments (McCabe et al., 2010). Multiple tested were chosen to encapsulate different components of executive function in order to determine potential task-dependent changes in cognitive function. The CTB consisted of a Groton Maze Learning Task (GMLT), a detection task, a two-back task, a set-shifting task, and a GMLT recall task. The order of the tasks was identical CTB was performed. However, the stimuli in each task was randomized to ensure similar task difficulty each time the CTB was performed. For each task, participants were instructed to respond as quickly and accurately as they could.

The GMLT is a touch screen-based test that measures executive function through error detection and spatial memory. The test consists of a hidden 28-step pathway on a 10 × 10 grid of squares. A blue tile on the top left corner of the screen indicates the starting location while a red circle on the bottom right indicates the finish location. The GMLT is performed five times per test and is randomized and matched for difficulty on every trial assessment to minimize learning effects. Performance changes were measured for the total duration (s) and total number of errors (#) measured during the five-block period and for the last maze (GMLT 5). A GMLT recall test (GLMT–Recall) was performed at the end of each CTB, which required the participant to recall the same hidden pathway from the initial five-block period. The total testing period took ~5–7-min.

A detection task was used to test psychomotor reaction time. A face down card was presented on the screen and the participant was required to press a key as soon as the card was turned over. This process continued until the task is completed with 35 cards being presented with a 2 s interval in between. Performance was measured for speed (mean of the log10 transformed reaction times for correct responses) and for total number of errors (#). The task required ~2-min to complete.

A 2-back test was used to measure of attention and visual working memory. This test required the participant to determine if the card being presented is identical to the card presented two cards ago. There was a total of 48 cards presented and the participant could select either “yes” or “no” for each card. Performance for this task was measured for speed of processing (mean of the log₁₀ transformed reaction times for correct responses) and total number of errors (#). The task required ~2-min to complete.

A set-shifting task was used measure of cognitive flexibility. In this test, the participant was asked to answer the question “is this the target card?.” The participant was presented with a playing card in the center of the screen with either the word “number” or “colour” above it and had to select “yes” or “no.” The only feedback presented to the participant was that the next card would not be displayed until the correct response is made. The target card changes throughout the test which could be either from one colour to the other (i.e., from a red target card to a black target card or intra-dimensional shift) or from “colour” to “number” (i.e., from a red target card to a number two target card or extra-dimensional shift). The participant was not told when these changes occurred and needed to re-learn the new target card to continue with the test. Performance was measured based on speed of processing (mean of the log₁₀ transformed reaction times for correct responses) and total number of errors (#). The task required ~3-min to complete.

All continuous variable data are presented as the mean ± SD averaged over each time-point while completing the CTB. All physiological and cognitive variables were analyzed using separate condition (ISO vs. POIKI) x time-point (BASE, NC-HS, HC-HS, HC-CS) repeated measures ANOVAs. If data violated sphericity (Mauchley’s test, p < 0.05) the Greenhouse–Geisser was used. A Bonferroni post-hoc analysis was used to test significant timepoint main effects. Pre and Post body mass was assessed using a condition (ISO x POIKI) by time (PRE x POST) repeated measures ANOVA. If a significant condition x timepoint interaction was calculation, paired sample t tests were used to test the main effects at specific time-points (e.g., BASE vs. BASE) to determine if there was a condition effect. To reduce the likelihood of Type I error due to multiple comparison, the α value was revised based on number comparisons (2, i.e., HC-HS POIKI verus HC-HS ISO), therefore p = 0.025 for paired samples t test comparisons.

All ordinal data (Thermal Comfort, Thermal Sensation) is presented as the median (quartiles 1 and 3) and were analyzed using separate condition (ISO vs. POIKI) x time-point (BASE, NC-HS, HC-HS, HC-CS) repeated measures ANOVAs, with a Wilcoxon signed-rank test used to compare at specific time points if a condition effect was present. All analyses were performed using IBM SPSS Statistics for Windows (version 26.0; IBM Corp., Armonk, N.Y., United States).

The experimental design was successful in creating four distinct temperature and perceptual time-points (Table 2). There was a time-point effect for T_re (Table 2; p < 0.001, η_p² = 1.00), where pairwise comparisons determined T_re was significantly higher in the HC-HS and HC-CS compared to BASE and NC-HS (all p ≤ 0.001). Pairwise comparisons demonstrated no differences in T_re between BASE and NC-HS (p = 1.00), but a significant difference compared to either HC-HS and HC-CS (both p < 0.05). There was a significant conditon effect (p = 0.014, η_p² = 0.766) and time-point effect for T̅_skin (Table 2; p < 0.001, η_p² = 1.00), where T ¯ skin was significantly different at all time-points (all p ≤ 0.001). There was a significant timepoint effect for Thermal Comfort (p ≤ 0.001, η_p² = 1.00); where NC-HS (p = 0.001) HC-HS (p ≤ 0.001), but not HC-CS (p = 0.057) were different from BASE; there was no Thermal Comfort difference between NC-HS and HC-CS (p = 0.296). Additionally, there was a significant time-point effect for Thermal Sensation (p ≤ 0.001, η_p² = 1.00), where pairwise comparisons demonstrate differences between each time-point (all p < 0.05).

Urine specific gravity prior to the experimental trials was <1.020 and was not different between conditions (p = 0.138, ISO: 1.008 ± 0.006, POIKI: 1.011 ± 0.006). There was a significant time-point effect (p ≤ 0.001) for body mass from PRE (ISO: 76.2 ± 10.1 kg, POIKI: 76.2 ± 9.9 kg) to POST (ISO: 74.5 ± 10.2 kg ~ −2.0% loss, POIKI: 74.7 ± 10.1 kg, ~ −2.0% loss) with no differences between conditions (p = 0.640).

There was a significant time-point effect for V ˙ I (p ≤ 0.001, η_p² = 0.999), condition (p = 0.005, η_p² = 0.894), and condition x time-point interaction (p = 0.001, η_p² = 0.964; Figure 2A). Pairwise comparisons revealed a hyperthermia-induced hyperventilatory response where V ˙ I significantly increased from BASE in both HC-HS (p = 0.002) and HC-CS (p = 0.034). HC-HS was significantly higher than all other time-points (all p < 0.05), while there were no differences between NC-HS and BASE (p = 0.961) or NC-HS and HC-CS (p = 0.105). Paired samples t-tests indicated a significant difference in V ˙ I between ISO and POIKI at HC-HS (ISO: 22.6 ± 7.5 l·min⁻¹, POIKI: 15.3 ± 2.7 l·min⁻¹, p = 0.005) and HC-CS (ISO: 17.3 ± 4.3 l·min⁻¹, POIKI: 11.2 ± 2.3 l·min⁻¹, p = 0.002). Similarly, f ˙ b r demonstrated a significant time-point (p = 0.007, η_p² = 0.853) and condition (p = 0.03, η_p² = 0.628) effect, with a non-significant condition × time-point effect (p = 0.619, η_p² = 0.089; Figure 2B). Pairwise comparisons revealed that compared to BASE, f ˙ b r was significantly higher at HC-HS (p = 0.021) and HC-CS (p ≤ 0.001). Furthermore, HC-CS was significantly higher than NC-HS (p = 0.006), with no other significant differences between conditions (all p > 0.05).

There was a significant condition (p = 0.002, η_p² = 0.952), time-point (p = 0.002, η_p² = 0.985), and condition x time-point interaction (p ≤ 0.001, η_p² = 0.999) for P_etCO₂ (Figure 2C). Pairwise comparisons demonstrated that P_etCO₂ was significantly different from BASE at HC-HS (p = 0.01) and HC-CS (p = 0.022), but not NC-CS (p = 0.09). Paired samples t-tests demonstrated there was a significant difference between ISO and POIKI at HC-HS (ISO: 42.3 ± 2.2 mmHg, POIKI: 31.7 ± 6.2 mmHg, p = 0.001) and HC-CS (ISO: 40.7 ± 2.1 mmHg, POIKI: 34.2 ± 5.4 mmHg, p = 0.004), but not BASE (ISO: 40.3 ± 2.4 mmHg, POIKI: 39.9 ± 2.1 mmHg, p = 0.515), or NC-CS (ISO: 39.6 ± 2.3 mmHg, POIKI: 37.9 ± 3.3 mmHg, p = 0.203). At HC-HS, the relative change from BASE was ~ + 4.0% in ISO and ~ −26.0% in POIKI.

Due to technical issues, data for MCA_v was limited to n = 8. Within this reduced dataset, there was a significant time-point effect (p = 0.006, η_p² = 0.907), condition (p = 0.013, η_p² = 0.790), but not condition x time-point interaction (p = 0.493, η_p² = 0.198) for MCA_v (Figure 2D). Pairwise comparisons demonstrated that, compared to BASE, MCA_v was significantly different at NC-HS (p = 0.014) and HC-HS (p = 0.003), but not HC-CS (p = 0.216). HC-HS was significantly lower than all other time-points (all p < 0.05). NC-HS was not significantly different than HC-CS (p = 1.000). Values for MCA_v were BASE (ISO: 67.7 ± 9.7 cm·s⁻¹, POIKI: 66.3 ± 8.1 cm·s⁻¹), NC-HS (ISO: 63.1 ± 10.0 cm·s⁻¹, POIKI: 61.5 ± 9.5 cm·s⁻¹), HC-HS (ISO: 55.4 ± 9.3 cm·s⁻¹, POIKI: 48.6 ± 9.1 cm·s⁻¹), and HC-CS (ISO: 61.0 ± 10.4 cm·s⁻¹, POIKI: 57.0 ± 10.7 cm·s⁻¹).

Due to poor signal quality of electrocardiogram, one participant’s data was removed from the heart rate analyses (n = 10). A significant time-point effect was seen for heart rate (Figure 3A; p < 0.001, η_p² = 1.00) with no condition (p = 0.512, η_p² = 0.094) or interaction (p = 0.062, η_p² = 0.600). Pairwise comparisons demonstrated that heart rate was significantly different from each other at all time-points (all p < 0.05). There was no condition (p = 0.759, η_p² = 0.059), time-point (p = 0.473, η_p² = 0.214), or condition x time-point interaction (p = 0.264, η_p² = 0.331) for MAP (Figure 3B). There was a significant time-point effect for SBP (Figure 3C, p ≤ 0.001, η_p² = 1.00), with no condition (p = 0.322, η_p² = 0.157) or condition x time-point interaction (p = 0.06, η_p² = 0.589). Pairwise comparisons revealed SBP was significantly higher at HC-HS compared to all other time-point (p ≤ 0.001), while there were no differences between the other time-point (all p = 1.000). For DBP (Figure 3D), the time-point effect approached significance (p = 0.052, η_p² = 0.769), however pairwise comparisons demonstrated no differences between any time-point (all p > 0.05). There was no condition (p = 0.738, η_p² = 0.061) or condition × time-point interaction (p = 0.268, η_p² = 0.318) for DBP.