High-intensity physiological activation disrupts the neural signatures of conflict processing

Abstract

Fluctuations in physical arousal occur constantly along the day and become particularly pronounced at extreme states such as deep sleep or intense physical exertion. While changes in arousal are thought to affect cognitive control, it has been suggested that cognitive control is resilient during the drowsy state as a result of neural compensatory mechanisms. Here, we investigate the higher end of the arousal spectrum by looking at the modulatory effect of high arousal on behavioural and neural markers of cognitive control. We predicted that preserved behavioural measures of cognitive control under high arousal would be accompanied by changes in its typical neural correlates. We conducted an electroencephalography study in which 39 expert cyclists (37 males, 2 females) were presented with an auditory stimulus-response conflict task while cycling on a stationary bike. Participants performed two experimental sessions on different days: one at low intensity and one at high intensity pedalling. Consistent with our predictions, we found no behavioural difference in cognitive conflict measures between the two exercise conditions. However, the typical midfrontal-theta power signature of cognitive control was no longer reliable at high-intensity exercise. Similarly, time-frequency multivariate decoding failed to decode stimulus conflict. On the other hand, we found no difference between intensity levels in whole-brain connectivity measures. Therefore, we suggest that the human cognitive control system is resilient even at high arousal states and propose that the dissociation between behavioural and neural measures could indicate the activation of neural compensatory mechanisms as a response to physiological pressure.SIGNIFICANCE STATEMENTEffects of spontaneous physiological changes on brain and cognition have traditionally been studied in states of decreased arousal. However, virtually no research has been conducted on the higher end of the arousal spectrum. In this study, participants performed an auditory conflict task while cycling at low and high intensity. Behavioural performance was robust in both conditions. However, changes in the typical univariate and multivariate signatures of cognitive control induced by arousal suggest a reconfiguration of the neural processes supervising cognitive control during heightened states of strain.

Staying in control: characterising the mechanisms underlying cognitive control in high and low arousal states

“…Consistent with previous studies examining cognitive control during moderate-to-high physical exercise (Davranche et al, 2009, 2015; Joyce et al, 2014; Schmit et al, 2015), conflict detection and adaptation were both fully preserved during high arousal. As in low arousal, this was not accompanied by the preservation of the classical neural correlate of conflict resolution (Avancini et al, 2023), suggesting some parallelisms between these two strained states of arousal. Yet, the absence of impaired cognitive performance when participants exercised close to the anaerobic threshold deviates from the hypotheses initially proposed.…”

“…Notably, the DDM in their research did not distinguish between automatic and controlled processing, and therefore it could not detect the effect in the amplitude of automatic processes reported here, even if it was present. Interestingly, EEG multivariate spectral decoding analyses of the high arousal data showed that the task-irrelevant information (i.e., spatial location) was only present (decodable) in the baseline condition in the 2-10 Hz band, but not during high-intensity exercise (Avancini et al, 2023). Also consistent with a narrower focus during strained states of high arousal, reduced overt attentional capture (i.e., automatic disruption from abrupt distracting stimuli) has been previously observed right after a bout of intense physical exercise (Llorens et al, 2015).…”

“…The resistance of the bike was controlled by the software MagDays, stimuli were presented using a PsychToolbox software in Matlab, and data were acquired with Brain Vision and AcqKnowledge softwares, and a BIOPAC amplifier, on Dell (USA) computers. Note that EEG data are reported in Avancini et al (2023).…”