Does neural variability during acute psychosocial stress capture stress-related brain–body associations?

This is the authors' abstract. We don't add key points for this paper.

Acute psychosocial stress and brain-body coupling Acute psychosocial stress elicits coordinated neural, endocrine, cardiovascular, affective, and behavioral responses that enable adaptation to changing environmental demands (De Kloet et al., 2005; Lupien et al., 2009). These responses unfold across different timescales and show substantial inter-individual variability. At the peripheral level, the autonomic nervous system responds rapidly through sympathetic activation, whereas activation of the hypothalamic-pituitary-adrenocortical (HPA) axis results in the delayed release of cortisol (De Kloet et al., 2005; Jankord & Herman, 2008). At the neural level, psychosocial stress engages distributed limbic, salience-processing, default-mode, and executive-control regions involved in threat detection, self-referential processing, cognitive control, and regulation of physiological arousal (Hermans et al., 2014; Menon, 2011; Van Oort et al., 2017). The ScanSTRESS paradigm provides an established functional magnetic resonance imaging (fMRI)- compatible framework for investigating these coordinated stress responses (see Figure 1). During alternating stress and control blocks, participants perform demanding mental arithmetic and mental rotation tasks under time pressure, social evaluation, and negative feedback (Streit et al., 2014). ScanSTRESS reliably induces neural, endocrine, cardiovascular, affective, and behavioral stress responses (Berretz et al., 2021; Henze et al., 2025; Noack et al., 2019). Our recent harmonized mega-analysis of multi-site ScanSTRESS data demonstrated that conventional task-based blood oxygenation level dependent (BOLD) responses are systematically associated with different components of the acute stress response (Henze et al., 2025). Cortisol responses were associated with activation in salience- and default-mode-related regions, heart rate with activity in default-mode regions, negative affect with salience- and default-mode activity, and task performance with default-mode and central-executive regions. Several of these brain-body associations additionally differed between male and female participants. Thus, mean task-related BOLD responses capture meaningful inter-individual variation in endocrine, cardiovascular, affective, and behavioral stress reactivity. Importantly, neural stress responses also change over the course of stress exposure. Previous ScanSTRESS studies identified an exposure-time effect characterized by decreasing neural activation from the first to the second ScanSTRESS run (Henze et al., 2020, 2025). Altered exposure-time effects have also been observed in clinical populations, including individuals with burnout symptoms during ScanSTRESS (Bärtl et al., 2024) and patients with major depressive disorder during the Montreal Imaging Stress Task (MIST, (Dong et al., 2022). Notably, such temporal alterations can emerge even in the absence of pronounced differences in overall mean neural stress responses. These findings suggest that temporal characteristics of neural stress processing may contain meaningful information beyond average activation levels and further motivate the investigation of measures that capture dynamic aspects of the BOLD signal. Beyond mean BOLD: Neural variability Task-based fMRI studies conventionally quantify neural responses as average BOLD signal differences between experimental conditions. Although this meanBOLD approach provides information about the overall magnitude and direction of neural responses, it necessarily collapses fluctuations occurring within a condition into a single average value. This may be particularly limiting for acute psychosocial stress, during which participants continuously adapt to changing cognitive demands, social-evaluative threat, negative feedback, and physiological arousal (Dickerson & Kemeny, 2004). Beyond this conceptual limitation, meanBOLD measures also show limited test-retest reliability in many task-based fMRI contexts, with average intraclass correlations often below .40 (Elliott et al., 2020; Tsikonofilos et al., 2025). Together, these conceptual and psychometric limitations motivate the investigation of complementary neural measures that may better capture dynamic aspects of brain function and individual differences in stress processing. Neural variability (NV) provides a complementary way of characterizing such temporal fluctuations. It can be quantified as the standard deviation of the BOLD signal over time (BOLD-SD) and thereby captures moment-to-moment variation in neural activity rather than its average level. BOLD variability, historically often regarded primarily as noise, has increasingly been associated with meaningful aspects of neural functioning, including cognitive performance, flexible transitions between brain states, and adaptation to changing task demands (Garrett et al., 2011; Uddin, 2020; Waschke et al., 2021). Importantly, BOLD-SD appears to capture information that is not redundant with conventional meanBOLD. Direct comparisons have demonstrated largely non-overlapping spatial patterns of BOLD-SD and meanBOLD, and BOLD-SD has shown substantial predictive utility for individual differences in cognitive and clinical characteristics (Garrett et al., 2010, 2011; Månsson et al., 2022; Schneider et al., 2024; Wehrheim et al., 2024; Wei et al., 2023; Xie et al., 2018). These findings raise the possibility that BOLD-SD may capture dynamic aspects of neural stress processing that remain hidden when neural responses are summarized by their average magnitude alone. Research gap Despite growing evidence that NV reflects meaningful individual differences in brain function, BOLD-SD has not yet been systematically examined during acute psychosocial stress induction. It therefore remains unknown whether stress-related neural variability is associated with the endocrine, cardiovascular, affective, and behavioral components of an acute psychosocial stress response and whether it provides information beyond conventional meanBOLD measures. The present study builds directly on the ScanSTRESS mega-analysis of Henze et al. (2025). Using the same harmonized multi-site framework, we will examine whether stress-related BOLD-SD is associated with cortisol responses, heart rate, negative affect, and task performance. Given previous evidence that neural stress responses and their associations with peripheral and affective stress-response measures differ between male and female participants (Henze et al., 2025), we will additionally investigate sex differences in stress-related BOLD-SD and in its associations with these stress-response measures.