Social buffering of the cortisol stress response during the Minnesota Imaging Stress Test in Children

Abstract

Introduction:

To understand neural underpinnings of individual differences in physiological stress responding, most notably of the hypothalamic-pituitaryadrenocortical system, but also of the autonomic system, it is essential to rely on an imaging task that reliably elevates cortisol and measures of the autonomic nervous system activity, such as salivary alpha amylase. When the question also involves neural activity related to social stress buffering, it requires a task that shows differential stress responses as a function of varying social buffering partners. The purpose of this study was to examine whether the Minnesota Imaging Stress Test in Children (MISTiC) with social buffering conditions fulfilled these requirements.

Method:

180 children ages 11 through 15 years (92 female) had salivary cortisol and salivary alpha amylase (sAA) samples taken during the MISTiC, a socially evaluative stressor modeled after the Trier Social Stress Test. Participants were randomly assigned to one of three social buffering conditions: Alone-No Buffer, Parent-as-Buffer, and Researcher-as-Buffer. Buffers interacted briefly with participants audiovisually at multiple points. Saliva samples for cortisol determination were taken 3 times during the hour preceding the MISTiC with the last serving as the pretest (T0) sample. Saliva was then collected post MISTiC at 25, 35, 45, 55, and 65 min after T0. The T0, 25, and 35 samples were assayed for sAA.

Results:

61 % of participants showed a significant increase in cortisol in response to the stressor (i.e., 115 % or greater) with roughly the same showing an increase in sAA. Change from T0 was analyzed for cortisol yielding a significant trials by condition interaction (p < .05). Post-hoc tests showed a significant difference between the Parent-as-Buffer and both the Alone-No Buffer and the Researcher-as-Buffer conditions, thus indicating that parents were still effective buffers for the cortisol response in this age range. The only significant effect for sAA was a trials effect, p < .001 with the same being true for self-ratings of stress, p < .001. Puberty (pre/early vs mid/late) did not moderate the response of social buffering condition on cortisol or sAA.

Conclusion:

The MISTiC is effective in elevating cortisol, sAA and perceived stress. For cortisol, the method used for buffering yielded significant differences by buffer type, suggesting that this paradigm is appropriate for assessing the neural systems underlying the social buffering of stress. Contrary to our prior work, pubertal stage did not moderate the effectiveness of the parent in buffering the child’s cortisol response.

1. Introduction

The study of stress reactivity and regulation is central to research on the developmental origins of physical and mental health (Koss and Gunnar, 2018). While experiences of traumatic stressors during development can have significant and often long-term health consequences (Martins et al., 2022), there are marked individual differences in physiological responses to stressors (McEwen and Stellar, 1993). McEwen and Gianaros (2010) noted that the brain is the central organ of stress reaction and regulation. It is involved in both interpreting the extent to which an event threatens well-being and it activates physiological systems to support adjustments needed to deal with the stressor. Understanding individual differences in stress responding, thus, needs to involve assessment of brain systems that are involved in producing these differences in response. However, it is critical that the stressors studied are ones that actually produce key stress-mediating physiological responses, and most notably can activate the hypothalamic-pituitary-adrenocortical (HPA) system and sympathetic nervous system.

The HPA system, whose stress-mediating hormone is cortisol, is one of the key stress-mediating systems in mammals (Joëls and Baram, 2009). Because of its centrality in responding to stress and because dynamic activity of this neuroendocrine system can be studied non-invasively via saliva, its activation and regulation have been a prime focus of research in human stress and adaptation for decades (Kirschbaum and Hellhammer, 1989; 1994). It has also been a focus of developmental research in both animals and humans for years (Gunnar and Vazquez, 2006). The pathways that activate and regulate the HPA axis are well mapped out in animal models (Herman, 2022). Importantly, these pathways are dependent on the nature of the stressors, particularly the distinction between stressors involving pain which recruit brainstem pathways and those that involve interpretation of threat, which engage limbic and frontal systems (Joëls and Baram, 2009).

The sympathetic adrenomedullary system is the other major stress-mediating system (Goldstein and Kopin, 2008). This system can be non-invasively measured using an index of the pre-ejection period or PEP. Unfortunately, this measure is more challenging to obtain in the strong magnetic field of the MRI environment. Instead, one can use a measure of salivary alpha amylase (sAA) which is an indirect index of activity of the autonomic nervous system (ANS; Ali and Nater, 2020).

The most effective psychological stressor task that elevates cortisol and sAA is the Trier Social Stress Test (TSST), a public speaking task that also involves performing mental arithmetic while being judged and filmed (Kirschbaum et al., 1993). This task is a social-evaluative stressor that threatens the social self with being judged poorly by others, a potent form of threat for humans (Dickerson and Kemeny, 2004). Social evaluation is especially potent during the adolescent period of human development (Somerville, 2013). Indeed, sensitivity to social evaluation appears to increase with pubertal development more so than age, suggesting neurobiological and hormonal changes are involved in this sensitivity (van den Bos et al., 2014).

The heightened sensitivity to social evaluation among adolescents may also help explain why parents become less able to buffer or reduce the HPA stress response for adolescents compared to children (Hostinar et al., 2015). The shift in potency of parental social buffering of the HPA axis also appears to be related to puberty more so than age (Doom et al., 2015). Doom and colleagues (2015) carefully screened and selected youth between 11.7 and 15 years of age to enroll roughly equal numbers of youth in pre/early pubertal (Tanner stages 1 and 2) compared to mid/later pubertal (stages 3, 4 and 5) groups. These groups were then randomly assigned to conditions of parental buffering versus researcher buffering, which meant supporting the participant as they prepared their speech. Among the prepubertal children, the parent was a highly effective buffer relative to the researcher, reducing the cortisol response to the in-person TSST for children. There was no difference between the parent and researcher buffers for the mid/late pubertal children.

The neural changes that result in developmental alterations in the impact of parents as social buffers of the cortisol response are not understood. Yet, we know that puberty has marked effects on many aspects of brain development, perhaps especially aspects of the social brain (Andrews et al., 2021). However, in order to examine neural changes with development related to social buffering, we need an imaging task that reliably elevates cortisol as well as one that shows a differential cortisol response as a function of varying social buffering partners.

Many attempts to study activation and regulation of the HPA axis in imaging contexts have used variants of the TSST, typically relying on the math section because discrete event markers are required for a task-based analysis of brain activation (e.g., Montreal Imaging Stress Test or MIST; Dedovic et al., 2005). Although the TSST versions used with adults are typically more potent than those used with children (see Seddon et al., 2020 compared to Goodman et al., 2017), the MIST and tasks like it tend to produce a relatively small cortisol response among about 50 % of the participants. To our knowledge, math-only tasks like the MIST are not effective in elevating cortisol among children or adolescents, although modified versions of the MIST have been used to study brain activation in adolescents (e.g., Celen et al., 2024).

We previously reported on the cortisol response to an imaging task that adapted both the speech and math portions of the TSST for use in the MRI scanner (Minnesota Imaging Stress Test in Children, MISTiC, Herzberg et al., 2020). In the current paper we describe using the MISTiC to assess cortisol and sAA reactivity under three social buffering conditions: buffering by parent, buffering by friendly female researcher, and no buffering (i.e., alone). The HPA axis has been the focus of much of the research on social stress buffering in humans and in other animals (Hennessy et al., 2009). This study expands on our previously published work which assessed parent and researcher buffer conditions in traditional behavioral versions of the TSST with children (Doom et al., 2015; Hostinar et al., 2015). Notably, in those previous studies, there was no alone condition. Thus, while among pubertal adolescents in previous TSST studies we did not find a difference between parent and researcher as a buffer, we do not know whether one or both types of social partners might have affected the cortisol response if compared to the standard alone condition. That is, while parents and researchers may not differ in their effectiveness as social buffers with increased pubertal stage, they might still be better than being alone during this stressor task.

In the current paper, we examined 3 hypotheses. (1) The MISTiC would be effective in producing an increase in cortisol and in sAA in the majority of participants. Thus, we predicted that we would replicate findings for cortisol reported by Herzberg and colleagues (2020). (2) Compared to the Alone-No Buffer condition, the youth in the Parent-as-Buffer would show a smaller cortisol response or more rapid return to baseline. (3) Youth in mid-late pubertal stages would show a reduced Parent-as-Buffer effect compared to youth at pre-early stages. We made no prediction as to whether the Researcher-As-Buffer would or would not differ from the alone condition. Hypothesis 3 is consistent with our earlier work (Doom et al., 2015) and with the gradual move away from family and increased reliance on friends or others with adolescent development (Helsen, Vollebergh and Meeus, 2000). We were underpowered to include sex as a factor in addition to condition and puberty, however, in sensitivity analyses presented in supplemental materials we examined the possibility of sex effects as required by PNEC.

2. Method

2.1. Participants

The participants were 186 (97 female) youth aged 11–15 years old (M=12.97, SD=1.05) who were recruited from a departmental registry of families interested in having their child participate in research. Exclusion criteria were premature birth (<37 weeks), serious medical or neurological condition, use of systemic glucocorticoids, or implanted metal including braces which would negatively impact MRI scanning. Once screened and after agreement to participate, participants were randomly assigned to one of three buffering conditions as described below. Of the 186 who came in for testing, four withdrew, one had cortisol values that were biologically implausible, and one could not be tested because of equipment failure. Thus, there were 180 participants who contributed data to this report. See Table 1 for demographic information for each buffering condition.

Table 1.

Counts, means, standard deviations for factors and potential covariates.

Variable	Alone-No Buffer	Parent-as-Buffer	Researcher-as-Buffer
	N = 64	N = 61	N = 55
Female/Male (assigned at birth)	32/32	30/31	30/25
Pubertal Stage (% 1–2 vs 3–5)	47/53 %	49/51 %	39/61 %
	M (SD)	M (SD)	M (SD)
Age (years)	13.05 (1.07)	12.98 (1.02)	12.88 (1.10)
% White	86	84	87
Total Household Income	8.36 (2.2)	8.26 (2.3)	8.14 (2.1)
Time of Arrival (hours)	15.46 (1.14)	15.31 (1.32)	15.88 (1.22)
Positive relationship quality - Mean of both Parents	3.09 (0.66)	3.00 (0.75)	3.10 (0.70)
Negative relationship quality - Mean of both Parents	1.77 (0.53)	1.74 (0.51)	1.73 (0.54)

Note. N = 180. Only data on participants whose cortisol data are reported in this publication are listed. Data in parentheses are standard deviations. Income of 8 = $125,000–150,000.

2.2. Procedures

Participants were recruited to roughly balance pubertal stage (pre/early stages 1–2 vs mid/late stages 3–5) within age (11 through 12 years, 13 through 14 years) and across conditions. [Note that pubertal stage data were then collected again from both parent and child report at the first session.] Once sorted by age, sex and parent-reported pubertal status, participants were randomly assigned to condition. Nearly all of the participants lived in two-parent households. We asked the families to designate which parent was the primary caregiving parent and this parent took part in all of the remaining activities. For the Parent-as-Buffer condition, the mother was the buffering parent in 85 % of the cases. The participant and their primary caregiving parent then completed two research visits: an initial 2-hour video conference (session 1) and a second 2-hour in-person testing session (session 2) at the University of Minnesota Center for Magnetic Resonance Research. During the video-conference parents and youth completed consenting procedures, reviewed the upcoming in-person session, and the child completed behavioral questionnaires using electronic survey entry into REDCap (Harris et al., 2009). Parents completed REDCap questionnaires on their own time. The in-person imaging session included an acclimation phase followed by the MISTiC. Participants arrived for the in-person imaging session between roughly 1300 and 1830 h with the median being 1545 h. This wide range in testing time reflected the need to co-ordinate MRI availability with family and researcher schedules.

During acclimation, study procedures and safety screening for the MRI were reviewed, and brief questionnaires about stressors, medications, and food intake were completed in REDCap. Participants completed an MRI scan simulation that presented the MRI sounds while inside a scanner bore to ensure comfort in the scanning environment. Participants then changed into scrubs to eliminate metal in the scan room and were given a bathroom break. Once in the scanning environment, a friendly technician prepared the youth for the fMRI including ear protection that allowed for communication and padding to reduce head motion. Once prepared, the participant watched ~5 min of a youth-friendly animated movie during structural imaging and completed ~5 min of resting state scanning. The acclimation phase lasted approximately 1 h (M=1.27 hr, SD=.16). The change from the acclimation to the stress phase was marked by a shift from interaction with the friendly technician to interaction with the judges (i.e., a pair of research staff members dressed in white lab coats).

The stress phase used the MISTiC paradigm (Herzberg et al., 2020). Judges (one male, one female) maintained neutral facial and vocal expressions and guided participants through public speaking and judged math tasks in the MRI scanner using a live video feed (picture-in-picture) projected on a screen in the MRI bore. Participants were instructed to prepare and deliver a speech as if they were a new student in a classroom and were introducing themselves to their classmates (Yim et al., 2010). For 5 min, they prepared the speech. They then verbally delivered the speech to the two judges (5 min). If the participant stopped talking, the judges informed the participant that their time was not yet up and to please continue. Judges ended the speech after 5 min even if the participant was still talking. Judges then explained the math portion of the task and remained on screen for 5 min to “judge” performance while the participant used a button box to respond to challenging but age-appropriate arithmetic problems. Throughout all judged segments, each judge periodically glanced down and wrote a note on a clipboard, then looked back to the participant, as if writing evaluations of the performance. When the 5 min of math ended, the judges departed, signaling the end of the stress phase. The participants in all 3 conditions were told that the judged portion of the study was over. The friendly tech immediately took over and the participant watched another 10 min of the animated movie. The tech administered an unjudged math task fMRI scan, during which still images of the judges, eyes-downcast, were presented to the participant. The images were included to control for face processing effects in the fMRI data. See Fig. 1 for a full timeline of the session.

Fig. 1.

MISTiC session timeline illustrating saliva sample times relative to the onset of the stressor task, buffering interaction points, MRI scanning and task order. Social buffer interactions with participants coincided with saliva sampling points. All samples were assayed for cortisol, while starred samples were also assayed for salivary alpha amylase. The acclimation phase appears in gray, while the stress and recovery phases are in orange.

2.3. Conditions

Participants were recruited and agreed to participate prior to being randomly assigned to one of three conditions: Alone-No Buffer, where participants performed the MISTiC without a social buffer; Parent-as-Buffer, where participants received social buffering from the family-defined primary parent (85 % female); Researcher-as-Buffer, where participants received social buffering from a friendly female researcher. Immediately prior to scanning and coincident with each saliva sampling event, buffers were introduced audiovisually to the participant for a brief (~30 sec) interaction opportunity. Buffers were instructed to chat and check-in in their natural manner, but to avoid reference to any task performance.

Note: The original design was to have the buffer physically present next to the scanner bed and able to interact with the participant in-person when each saliva sample was taken, which required sliding the participant partway out of the bore. However, just as data collection began, the COVID-19 pandemic shut down all in-person testing. Even when research scanning was allowed again, pandemic restrictions limited the number of people who could be present in the scanner control room and allowed only the participant and the technician to be present in the scanner bay itself. Thus, all buffering was conducted remotely by video.

2.4. Measures

2.4.1. Potential covariates

2.4.1.1. Network of relationship inventory.

Behavioral Systems Version (NRI-BSV, Furman and Buhrmester, 2009) was administered during the initial video-conference session. These data were collected to ensure that youth with better parental relationships, which might influence cortisol or brain responses, were not by chance assigned to one buffering condition more than others. Scores were computed for positive (alpha=.91) and negative (alpha=.92) subscales for relations with each parent. In 4 cases the child was being reared by a single mother and so the parent data reflected only one set of scores.

2.4.1.2. Pubertal.

Stage was measured by the Morris and Udry (1980) questionnaire during the video conferencing session 1 with the youth endorsing breast/testicle and pubic hair developmental stage using visual representations on a 5-point Tanner-like scale. Both children and parents reported on the child’s pubertal status. Child-report was used (n = 168) when available. In 9 instances, children did not complete the report so the parent-report data were used. In 3 cases, neither child nor parent completed the report, in these cases the parent’s initial report during recruitment was used. Breast and testicle reports were used as an index of central puberty and divided into pre-early (stages 1 and 2) versus mid-late (stages 3 through 5), to match procedures used in our prior study (Doom et al., 2015).

2.4.2. Stress measures

2.4.2.1. Cortisol.

Nine saliva samples of interest were collected throughout the session using Salimetrics SalivaBio swabs for the added safety provided by the 6” length, since participants were often providing samples while supine in the scanner bore. Samples 1–4 were taken during the acclimation phase with sample 4 being considered Time 0 (T0) for the stressor task, while sample 5 was taken 25 min after T0 which is typically the timing to peak cortisol response (Dickerson and Kemeny, 2004), and samples 6–9 were taken during what was expected to be recovery and return to prestressor levels. Sample 1 was taken within 1–2 min of arrival (T-80); Sample 2 (T-40) was taken just prior to entering the scanning environment and after the experience of the mock scanner. Sample 3 (T-20) was taken after the participant was prepared to scan. Sample 4 (T0) was taken following the 10 min of structural and resting state scans. Sample 5 (T + 25) was taken after the end of the MISTiC stressor. The remaining samples, taken at 10-minute intervals and were designed to capture duration of elevation and return to baseline. Fig. 1 shows a complete timeline of procedures, saliva sampling, and social buffering.

Samples were promptly stored at −20 ° C until being shipped to the lab for time-resolved fluorescent immunoassay (DELFIA) for cortisol determination in μg/dL. Samples were assayed in duplicate with coefficients of variation under 10 %. Data for one participant was dropped due to biologically implausible values, 1 % of all samples were winsorized to within 3.3 Z. Individual samples were log10 transformed to resolve positive skew.

2.4.2.2. Salivary alpha amylase (sAA).

Saliva samples 4–6 (T0, T + 25, and T + 35) were also used to assess sAA response and recovery, and were assayed using an enzymatic assay with coefficients of variation under 10 %. Like cortisol, all samples from a given participant were assayed in the same batch, < 1 % of values were winsorized to within 3.3Z, and individual samples were log10 transformed to resolve positive skew.

2.4.2.3. Self-report stress.

After scanning was complete, participants rated how stressed they felt at various points in the session using 5-pt Likert scales (during speech preparation, speech delivery, judged math, and once the session was over).

2.4.3. Data analysis plan

Data were analyzed using IBM SPSS Statistics (version 28) software. Planned preliminary analyses included a power analysis and analyses of potential covariates (analyses of variance and chi-square as appropriate) to determine whether the variable differed by condition. Only those differing by condition were included as covariates. We did not attempt to control for stage of the menstrual cycle as the majority of girls in this age group either do not yet menstruate or do not yet have a regular menstrual cycle (Dorn et al., 2006). An analysis of the acclimation period was conducted to verify the baseline from which response would be judged. That is, log-transformed cortisol values of samples 1–4 (T-80, T-40, T-20, and T0) were analyzed to examine differences by condition and time during the roughly one-hour acclimation phase.

Primary analyses examined stress and buffering. First, cortisol response rates were examined to evaluate the strength of the MISTiC manipulation. Next, GLM repeated measures were conducted on stress measures as a function of buffering condition: Alone-No Buffer, Parent-as-Buffer, Researcher-as-Buffer (3) and pubertal status: pre/early vs mid/late puberty (2). For cortisol, change scores (sample - T0 baseline) were analyzed to control for individual differences in baseline (T0) cortisol level while allowing the pattern of change over time to remain. Thus, a repeated measures analysis of cortisol trials (5) by conditions (3) by pubertal status (2) was performed. sAA was examined in a repeated measures analysis across 3 trials with condition (3) and pubertal status (2) as between subject factors. For subjective stress ratings, a trials (4) by condition (3) by pubertal status (2) repeated measures analysis was performed. Because there were only 3 sAA samples, to preserve power we did not analyze change from T0 as we did in the cortisol analysis.

3. RESULTS

3.1. Preliminary analyses

3.1.1. Power analyses for cortisol

A power analysis using G*Power was examined for the primary cortisol model. Based on the average correlation among the cortisol measures of 0.81, with an alpha of 0.05, our power was 0.80. While we had no sex by condition hypotheses, analysis for sex is a requirement of the PNEC journal. A power analysis determined there was insufficient power to analyze sex by puberty by condition by trials design. Results for this underpowered analysis are provided for description as a supplement to meet journal requirements.

3.1.2. Potential covariates

Age in years, percent of white participants, household income, positive and negative relations with parents, and time of arrival were tested as potential covariates. Time of arrival differed by condition, F(2, 177)= 3.29, p = .04. Because of the circadian rhythm in cortisol, time-of-arrival was entered as a covariate in the cortisol analysis.

3.1.3. Acclimation cortisol

Cortisol during the acclimation period was analyzed to determine whether cortisol decreased across the hour before the MISTiC was conducted and whether there were puberty or condition differences in acclimation cortisol that needed to be taken into account in the primary cortisol response analysis. A 2 (pre/early vs mid/post) by 3 (condition) by 4 (trials) analysis was conducted with trials as a repeated measure. The analysis failed the Mauchley test for sphericity and thus Greenhouse-Geiser correction was employed. Only a significant trials effect was obtained, F(1.8, 303.88)= 109.21, p < .001. The log transformed means were: −0.93, −1.09, −1.13, and −1.16, from arrival to just before introduction of the MISTiC with SD’s ranging from.24 to.26. Cortisol at T0 just prior to the TSST was further examined to determine that it could be used as the baseline by conducting a 2 (pubertal status) by 3 (conditions) Analysis of Variance. The results showed no significant effects of pubertal status, F(1174)= 0.12, p = .74, condition (2174)= .51, p = .60, or interaction of puberal status and condition, F (2, 174)= 1.68, p = .19. Thus the 4th acclimation sample, T0, taken right before the introduction of the MISTiC was used as the pre-test or baseline measure.

3.2. Stress measures

3.2.1. Responder analysis

As a check on whether the MISTiC produced a significant cortisol response in the majority of participants, we calculated the percent increase from T0 cortisol to the maximum cortisol level each participant achieved. A cut point of 115 % was used to define a meaningful response based on Miller and colleagues (2013). Using this definition, 61 % of the participants showed a cortisol response. By condition the response rates were Alone-No Buffer, 58 %, Parent-as-Buffer, 69 % and Researcher-as Buffer, 55 %. Response rates were not statistically different by condition, χ²(2)= 2.79, p = .25. A roughly comparable 62 % of participants showed an increase in sAA. sAA response rates defined as an increase greater than 1 SE were 59 %, 62 %, and 64 % and were not significantly different by condition, χ²(2)= .24, p = .89.

3.2.2. Cortisol change analysis

The Mauchly test was significant, thus Greenhouse-Geiser correction was used to test significance. With arrival time in the analysis, there was a significant effect of trials by condition, F(4.6, 398.15)= 2.55, p = .03. There was neither a puberty by trials, F(2.3, 398.15)= 1.13, p = .33, nor a puberty by conditions by trials effect, F(4.6, 398.15)= 0.58, p = .70). Post-hoc tests showed that condition by trials effects for the Parent-as-Buffer condition differed from the Alone-No Buffer condition, F(2.5, 301.6)= 4.08, p = .01, with a trend to differ from the Researcher-as-Buffer condition, F(2.5, 272.6)= 2.7, p = .06; while the Alone-No Buffer and Researcher-as-Buffer conditions did not differ, F(1.95, 222.36)= 1.18, p = .31. Based on typical TSST patterns, we anticipated a rise from baseline peaking near T + 25 and then a recovery to baseline. With a change score of each sample referenced to baseline, you would expect diminishing values after peak. Thus, in our analysis of change scores, we predicted a linear decrease over time. As shown in Fig. 2 this is what we observed for the Parent-as-Buffer condition, F(1,60)= 65.23, p < .001. The pattern of change scores for the Researcher-as-Buffer was also linear, decreasing after peaking at T + 25, F(1,54)= 7.6, p = .008. Finally, for the Alone-No Buffer condition, the pattern of change was actually quadratic, F(1,63)= 11.8, p = .001; this is graphically represented as a later mean peak for this condition. To determine whether the Alone-No Buffer condition was significant later than that of the other conditions, we computed the difference between the response at T + 25 and T + 35. We then conducted a one-way ANOVA. The groups differed significantly, F(2177) = 5.71, p = .004. Bonferroni tests showed that the Alone-No Buffer condition differed from the other two conditions (p’s < .05), while the two social buffer conditions did not differ from one another, p = .10.

Fig. 2.

Change in cortisol from sample 4 to sample 9 (T0 to T + 65) following the MISTiC by condition. Sample 5 (T + 25) immediately follows speech preparation, judged speech and judged math. Y-axis reference line at 0 represents T0 sample cortisol level.

3.2.3. sAA analysis

The Mauchly test was not significant, so no correction was used. There was a significant effect of trials, F(2, 346)= 19.17, p < .001. Within-subject contrasts showed the effect was quadratic, F(1, 173)= 41.74, p < .001: sAA rose then fell from T0 to T + 35. There were no significant main effects of condition or pubertal status, nor interaction effects, p’s > .10. See Fig. 3.

Fig. 3.

The response of sAA to the MISTiC. As only the trials effect was significant, this was the only effect graphed.

3.2.4. Subjective stress analysis

The Mauchley test for sphericity was violated, thus Greenhouse-Geiser corrections were imposed. There was a significant trials effect, F(4.05, 705.14)= 310.61, p < .001, but no other significant effects. As shown in Fig. 4, youth rated the speech delivery to the judges as the most stressful period of the session.

Fig. 4.

Self-report of stress on the 5 point Likert scale, ranging from 1 =Not at all to 5 = A Whole Lot. As only the trials effect was significant, only that effect is graphed.

4. Discussion

The results clearly indicated a significant cortisol response to the MISTiC. This is the second study that has found that this stressor task elevated cortisol in young adolescents (Herzberg et al., 2020). Sixty-one percent of the participants exhibited an elevation of 115 % or more, which is a common criterion for a significant increase (Miller et al., 2013) and is typically what is seen for TSST responses in this age group (e.g., Doom et al., 2015). The task also elicited a significant sAA response, and self-reported stress also increased, averaging 4 on the 5-point Likert scale for delivering the judged speech. Thus, this task appears well-suited for assessing brain activity during a social-evaluative stressor that also elicits a clear physiological stress response.

The effectiveness of social buffers to affect the cortisol response to the MISTiC was statistically significant for the Parent-as-Buffer group. Cortisol response mean was the highest at 25 min after the onset of the TSST, as is typically found for the HPA axis, and then decreased until, by the last sampling time, the mean had returned to within the pretest or baseline region. The peak mean for Alone-No Buffer condition was later than that for the two social buffer conditions. Since all participants were told at the end of the judged math section that the judges were leaving and the judged portion of the session was over, it seems unlikely that the participants in the Alone-No Buffer condition were less sure than those in the other two conditions that they were no longer being evaluated. Thus it is not clear why the peak mean in the Alone-No Buffer condition was later than in the two social buffering conditions.

In previous work, puberty moderated the effectiveness of parental social buffering, with effective parent buffering in youth who had not started puberty (stage 1) or were in its early phase (stage 2), but less effective buffering for those at mid to late pubertal stages (3–5). This was not the case in the present study. The behavior and context of the social buffer in the MISTiC, however, was quite different from that of the social buffer in previous work (e.g., Doom et al., 2015, Hostinar et al., 2015). In previous studies, the social buffer helped the participants prepare for the speech portion of the TSST and were physically present. In the current study, because the MISTiC was designed to collect imaging data during speech preparation, the participant could not be talking. Instead, for 30 s before speech preparation, the buffer interacted with the participant via headphones and monitor. Then the participant did not see the buffer again until after the end of the stress portion of the MISTiC (judged math) when the buffer again interacted for 30 s. Furthermore, in the present study the buffer was instructed NOT to talk about task performance. As the buffering conditions were quite different from previous studies, the modifications raise the possibility that the loss of parental effectiveness with increased pubertal stage may have to do with having the parent help the adolescent prepare for social performance, as opposed to simply being present. Thus, in previous work the reduction in parental buffering with more advanced puberty might be related to the type of support provided.

Notably, when we examined sAA, we did not find any condition effects, even though we observed a significant sAA response to the MISTiC. We also did not find condition or condition by trials effect for the self-report measure, although we obtained significant increases in reports of stress during the MISTiC. Thus, the effect of the parent as a social buffer was observed only for the HPA axis. This may explain why much of the research on stress buffering has focused on this neuroendocrine system (Hennessy et al., 2009).

It is also notable that the Research-as-Buffer did not appear to as effectively influence the participants’ responses, either for cortisol, sAA or self-report of stress. Unlike the parent, the researcher was not someone the participant knew. They met the researcher first while in the scan environment. Perhaps a familiar person, even if not the parent, could be effective in either reducing the magnitude of the response or producing a more rapid return to baseline. Recently, however, we examined whether a good friend could buffer the cortisol response to the TSST for adolescents (Filetti et al., 2025). Having the friend help the youth prepare for the TSST did not reduce the cortisol response, in fact, it increased it. Thus, across all of the studies conducted with children and adolescents that have examined whether a social buffer could affect the cortisol responses to a social evaluative stressor, the only type of buffer that has been effective when they are just there as support is the parent.

While there were strengths to this study, including the random assignment to condition and the balancing of many potential confounders across conditions, there were also limitations that need to be considered. The COVID-19 pandemic required a design shift. We were not able to have the buffers in the scanning room with the participant, a manipulation which appeared to effectively reduce neural stress markers in anxious youth in a prior imaging study (Conner et al., 2012). Thus, although the adaptations that were made allowed this study to be conducted during the pandemic years, they clearly altered the delivery of social buffering, and thus may have affected the results. Further, the participants reflected a restricted range of racial/ethnic variation with most being white and living in relatively highly resourced families. There is certainly evidence that social evaluation of the type used in the present study may have different impacts on minoritized populations and/or those from lower socioeconomic groups when being judged within a university context (Keenan et al., 2021). Thus, it will be important to conduct similar studies of this sort with more representative groups of youth. Although not a limitation, once females begin to have regular menstrual cycles, the cortisol response differs between the follicular and luteal phases. We did not attempt to control menstrual phase because only about half of the girls were menstruating and, in this age range, cycles are often irregular making it difficult to schedule testing for a particular menstrual phase (Dorn et al., 2006).

In sum, the MISTiC is a social evaluative imaging task that effectively elevates cortisol, sAA and self-reports of stress in the majority of individuals in the early adolescent (11–14 year) age range. The percentage of responders is similar to that observed among youth of similar ages during in-person versions of the TSST for children. The effectiveness of the parent as a social buffer was modest in that it did not reduce the magnitude of the response but did facilitate return to pretest or baseline levels. It would be informative to repeat this study using other buffering methods, such as hand holding, that have been shown to reduce cortisol responses to threat in imaging studies with adults (Coan et al., 2006).

Appendix A. Supporting information

Supplementary data associated with this article can be found in the online version at doi:10.1016/j.psyneuen.2026.107760.