Acute psychosocial stress weakens the sense of agency in healthy adults

Abstract

The sense of agency (SoA) refers to the feeling of being in control of one’s actions and the subsequent consequence of these actions. Emotional context seems to alter the strength of sense of agency. The present study explored the influence of acute psychosocial stress on the SoA by means of the Trier Social Stress Test (TSST). Self-assessment manikin (SAM) and objective physiological indicators (e.g. heart rate, electrodermal activity, and salivary cortisol levels) were utilized to evaluate the effect of the TSST. We also employed the temporal binding effect as an implicit assessment of the participant’s SoA. The results indicated that the stress level of the experimental group after TSST was significantly higher than the control group, whilst the temporal binding scores of the experimental group decreased after TSST manipulation. In short, acute psychosocial stress with intense emotional arousal weakened the sense of agency.

Keywords:

• Sense of agency
• acute psychosocial stress
• the Trier Social Stress Test
• temporal binding effect
• emotional context
• self-awareness

1. Introduction

The sense of agency (SoA) is the feeling of control over our voluntary actions and through them, of events in the outside world (Haggard, 2017). As a crucial component of self-awareness, the sense of agency is a cornerstone of psychological health. SoA primarily relies on the comparison between the predicted consequences of our actions and the actual sensory feedback (Moore & Obhi, 2012). The match of predicted and actual outcomes lead to SoA arises, whereas if they are incongruent, the outcome will likely be attributed to an external cause. For instance, when you push the button to go up by the lift, if the door opens in time, you will feel an enhanced sense of agency over the elevator door.

It has been shown that the emotional modulation of SoA may regulate social behavior (Christensen et al., 2019; Cui et al., 2018; Yoshie & Haggard, 2013). Yoshie and Haggard (2013) provided direct evidence for the modulation of SoA by affective significance of action outcomes. In two experiments, they showed that temporal binding effect, an implicit measure of SoA, was attenuated for negative compared to positive or neutral outcomes. The temporal binding effect refers to the phenomenon in which the perceived time points of a voluntary action and its consequence are attracted to each other (Haggard et al., 2002; Moore & Haggard, 2010). The SoA was also decreased for fear-relevant outcomes relative to neutral ones (Cui et al., 2018).

Though negative outcomes weakened sense of agency, few studies investigated the influence of emotion state on sense of agency. A study found that arousal states with negative emotional valence, such as fear and anger, decreased sense of agency (Christensen et al., 2019). A recent study has shown inconsistent results using virtual reality (VR) technology to explore stress state on embodied SoA (Stern et al., 2020). Fourteen participants performed virtual hand tasks into a “neutral” block or a “stress” block. They found that mild psychosocial stress did not affect SoA in a virtual reality context. Therefore, the influence of acute psychosocial stress on SoA deserved further study.

The Trier Social Stress Test (TSST) is an ecologically valid stressor that involves social evaluation and unpredictability, forcing participants to speak in front of an expressionless audience and to complete an unexpected mental arithmetic test (Kirschbaum et al., 1993). Stress-induced in TSST could be more persuasively verified by physical indicators. Heart rate is thought to be a standard physiological indicator of the effect of TSST (Allen et al., 2014). Furthermore, numerous studies have shown that TSST leads to hypothalamus–pituitary–adrenal (HPA) axis activation, and cortisol is the final product of HPA axis activation (Gaab et al., 2003; Kirschbaum et al., 1999). Research suggested that electrical skin levels are also sensitive to a wide range of stressors (Posada-Quintero et al., 2017).

As mentioned above, there is still a lack of systematic research on the influence of stress state on the sense of agency. We employed temporal binding effect as an implicit measure of the sense of agency. This effect was first studied using the Libet Clock paradigm (Haggard et al., 2002), and the interval estimation paradigm was derived afterward (Humphreys & Buehner, 2009; Zhao et al., 2013). In the present study, participants were randomly divided into an experimental and a control group. The participant’s stress state in experimental group were induced through the TSST paradigm, while the control group did not. We hypothesized that TSST would cause a significant increase in participants’ stress levels, as well as a substantial decrease in the temporal binding effect.

2. Methods

2.1. Participants

A total of 59 participants (41 females, 22.05 ± 2.43 years) were recruited. Twenty-five were excluded (see 2.7 Data analysis for details), leaving 34 eligible participants, 17 subjects per group (10 females in the experimental group, 21.20 ± 2.66 years; 13 females in the control group, 23.15 ± 2.67 years). All participants were right-handed and healthy university students aged 18–28 years old who had normal vision or corrected-to-normal vision. The incorporated criteria also included: no surgery in the last six months, no smoking, no alcohol abuse, regular life and rest, and none of the women used oral contraceptives (Allen et al., 2017).

Sample sizes were determined using G-Power 3.1.9.4 (Faul et al., 2009). A mixed-design analysis of variance with a within-subjects factor of the stage (pretest and posttest) and a between-subjects factor of the group (experimental group and control group) was conducted for temporal binding scores. We required 34 participants to detect a medium effect size (Cohen’s f = 0.25) at 80% power, setting α = 0.05, Number of groups = 2, Number of measurements = 2, Correlation among repeated measures = 0.50, and non-spherical correction ε = 1. The stopping rule targeted 34 eligible participants whose data could be further analyzed (Gu et al., 2021).

2.2. General procedure

Before the experiment, participants needed to complete the time estimation test online at home. Firstly, they learned different length of the sound (100 ms, 500 ms, and 900 ms) by listening and then carrying on the test. Those who did not pass the test cannot participate in the experiment. All participants gave written informed consent and were paid for their participation. The experiment was conducted in accordance with the Declaration of Helsinki and approved by the Institutional Review Board of the Institute of Psychology, Chinese Academy of Sciences.

The experimental procedure is shown in Figure 1. Participants arrived at the laboratory about 13:00 or 15:00. They began a 20-min break during which they completed the Trait Anxiety Scale, SoA scale, and SAM scale (pretest). The heart rate and electrodermal activity pads of the physiological synthesizers were connected. Heart rate, electrodermal activity (HR _base, EDA _base), and the first tube of saliva cortisol (SC _base) were then collected.

Figure 1. Schematic illustration of the procedure. Saliva cortisol (SC) was collected five times at baseline, before the TSST, and after the TSST 0 min/15 min/30 min. Note: Heart rate (HR) and electrodermal activity (EDA) were recorded eight times for 5 min at baseline, before the TSST, throughout the preparation/speech/arithmetic, and at 0min/15 min/30 min after the TSST manipulation. Two interval estimation tasks were performed before and after the TSST, respectively. H&E: Heart rate (HR) and electrodermal activity (EDA), HR_base(baseline), HR_bef (before TSST), HR_pre(preparation), HR_spe(speech), HR_Air(Arithmetic), HR_0min(TSST_0min), HR_15min(TSST_15min), HR_30min(TSST_30min), SC: Salivary cortisol; TSST: Trier Social Stress Test.

An illustration of timeline that salivary cortisol, heart rate, and electrodermal activity were measured.

Next, the interval estimation task was conducted by E-Prime 2.0 for approximately 30 min. At the end of the task, heart rate, electrodermal activity (HR _bef, EDA _bef), and the second tube of saliva cortisol (SC _pre) were collected.

Subsequently, the TSST was conducted which consisted of preparation, interview, and mental arithmetic for a total of 15 min. Heart rate and electrodermal activity (HR _pre, HR _spe, HR _ari, and EDA _pre, EDA _spe, EDA _ari) were recorded. After the TSST, heart rate, skin electrodes (HR_0min, EDA_0min), and the third tube of saliva cortisol (SC_0min) were collected. Then, participants filled out the SAM scale (posttest).

A second interval estimation task was conducted lasting approximately 30 min. During a 5-min break within the temporal estimation task, heart rate, electrodermal activity (HR_15min, EDA_15min), and the fourth tube of saliva cortisol (SC_15min) were collected. At the end of the interval estimation, heart rate, electrodermal activity (HR_30min, EDA_30min), and the fifth tube of saliva cortisol (SC_30min) were collected.

2.3. Scales

2.3.1. Trait Anxiety Inventory and SoA scales

The State-Trait Anxiety Inventory (STAI) has been widely used for the assessment of anxiety in various populations. A Chinese version of the STAI's Trait Anxiety Inventory (T-AI) was used in this study. The trait anxiety subscales are all scored on a scale of 1 to 4, the higher the score, the more pronounced the individual’s anxiety trait.

The Sense of Agency scale (SoAs) contained both the positive sense of agency (SoPA) and the negative sense of agency (SoNA). It had a total of 11 items, 5 for SoPA and 6 for SoNA (Tapal et al., 2017).

2.3.2. SAM scale

The Self-Assessment Scale (SAM) is shown in Figure 2. The SAM is a nonverbal assessment technique that directly measures the valence of pleasure and arousal, associated with a person’s emotional response to various stimuli.

Figure 2. Illustration the paper version of SAM. Note: (a,b) Adapted version of the Self-Assessment Model for potency and arousal ratings. Valence and arousal were rated on a 9-point Likert scale, (a) for emotional pleasantness, where 1 represents the sadness and 9 represents the happiness, and (b) for emotional arousal, where 1 represents the calmness and 9 represents excitement.

Ten pictures were divided into a and b groups. The first five pictures from the saddest to the happiest, representing the mood for a; the last five pictures from the calmest to about to explode, representing the degree of arousal for b.

2.4. Acute stress and control conditions

2.4.1. TSST manipulation

The Trier Social Stress Test (TSST) is a widely used method inducing acute stress responses in a laboratory setting, which consists of a 5-min preparation, a 5-min presentation, and a 5-min mental arithmetic task. In the interview presentation, the participant was guided to apply for a job position and to prepare for the following interview. Taking notes with pen and paper was allowed, but the paper would be taken away during the presentation. In the presentation phase, the participant was led to an individual interview room with three interviewers who might ask questions at any time and would be informed that there was a video camera recording. In the mental arithmetic phase, the participant was asked to subtract 13 from 1022 consecutively and to report each result of the mental arithmetic loudly. If the result was incorrect, the participant had to start again from 1022 until the end of the session. All three interviewers pretended to be very calm and reserved.

2.4.2. Control treatment

The control condition was consistent with the physical and psychological demands of the stress task, but not stressful by omitting the social and self-relevant components. Participant was only required to read a neutral text loudly to the camera, followed by 5 min of written arithmetic at the end.

2.5. Interval estimation task

The present study employed a modified version of interval estimation paradigm (Zhao et al., 2013). The interval estimation task was conducted by E-Prime 2.0 for approximately 30 min. The task was completed in two phases. Before the formal experiment, participants were asked to pass an online test at home in advance. Those who failed to passed the test were not allowed to enter the formal experiment. The formal experimental phase was completed in the laboratory.

Practice phase: Participants were trained to be familiar with different temporal intervals (100 ms, 500 ms, or 900 ms). They were told that 1 second equals 1000 ms. Participants first saw a prompt for the length of time, such as 100 ms. A fixation cross then appeared on a gray blank screen followed by a tone. The prompted time is the temporal interval between the disappearance of the fixation cross and the appearance of the subsequent tone. Each temporal interval was trained 5 times. Participants could practice again if they were not familiar with the task.

Test phase: The test phase was followed by the practice phase. A time interval was randomly selected and presented for the test phase, including 100 ms, 500 ms, and 900 ms. First, a fixation cross appeared on a gray blank screen followed by a tone. Participants needed to estimate the temporal interval between the fixation cross and tone. A total of three trials were presented and the participant passed if all answers were correct.

Experimental phase: The formal experimental phase contained voluntary and involuntary conditions. To avoid sequential effects, the order of the voluntary and involuntary conditions was counterbalanced. Each condition contained a total of 95 trials, of which 5 were practice trials (not analyzed).

In the voluntary condition, a fixation cross appeared first and the participant pressed "Z" on the keyboard for left or "M" for right. The fixation cross disappeared followed by a "left" or "right" feedback tone. Then a fixed time interval (250 ms, 550 ms, or 850 ms) appeared. The interval between the disappearance of the fixation cross and the appearance of the feedback tone were then required to be reported for participants by clicking on a 0–1000 ms timescale.

In the involuntary condition, a fixation cross appeared and the participant heard a key beep. The fixation cross then disappeared simultaneously. A "left" or "right" feedback tone appeared after a fixed interval (250 ms, 550 ms, or 850 ms). The time interval between the disappearance of the fixation cross and the appearance of the feedback tone were required to be reported for participants by clicking on a 0–1000 ms timescale, See Figure 3.

Figure 3. Diagram of the interval estimation process. Note: A fixation cross appears at the center of screen. The participant presses "Z" for left or "M" for right on the keyboard (voluntary condition) or hears a sound of key beep (involuntary condition). The fixation cross disappears followed by a fixed time interval with a feedback tone "left" or "right." The participant was asked to report the time interval by clicking on a 0–1000 ms timeline.

The experiment contained voluntary condition and involuntary condition. For the voluntary condition, participant was asked to estimate the temporal interval between disappearance of fixation cross which caused by voluntary action and the appearance of feedback tone. As for the involuntary condition, the only difference from voluntary condition was that the fixation cross was passively disappeared.

2.6. Physiological index collection

2.6.1. Salivary cortisol collection

A total of 5 saliva samples were collected from each participant using a biosharp (Beijing, China) BS-20-M 2 ml centrifuge collection tube. Samples were frozen in a −22 °C refrigerator within 2 h of collection until further analysis. Samples were thawed and centrifuged at 4000 rpm for 20 min before analysis by enzyme-linked immunoassay (ELISA) with biotin-labeled SC in a competitive assay (ELISA-Kit, WuHan Chundu Biotechnology Co., Ltd, China). The sensitivity of cortisol was 0.1 nmol/L. The linear regression of samples with the expected concentration correlation coefficient R-value was above 0.99. The intra-batch coefficient of variation and inter-batch coefficient of variation was less than 15%.

2.6.2. Heart rate and electrodermal activity acquisition

Heart rate and electrodermal activity were collected from participants using a BIOPAC MP160 polysomnographic recorder (USA) with a PPGED-R amplifier. The collected electrodermal activity and heart rate were recorded and analyzed by AcqKnowledge 4.0. The participant’s index finger, middle finger, and ring finger were disinfected with alcohol cotton balls and left to dry. Two electrode pads were then installed on the abdominal part of the first knuckle of the participant’s index and middle fingers. The positive and negative electrodes of electrodermal activity were connected to the two electrode pads. The heart rate recording electrode was connected to the ring finger. Participants were instructed to turn off electronic devices such as cell phones to prevent potential interference with the device signal.

2.7. Data analysis

Questionnaire analysis: Pearson correlation was conducted among trait anxiety scores, SoPA, and SoNA. SAM was analyzed with emotional valence expressed as a2a1 and emotional arousal as b2b1, where a2a1 = (a _post − a _pre), b2b1 = (b _post − b _pre), and independent samples t-test was conducted for a2a1 and b2b1.

Analysis of physiological indicators: To examine the validity of stress inducement, two-way mixed analyses of variances (ANOVAs) were conducted for heart rate (HR), skin conductance level (SCL), nonspecific skin conductance responses (NS.SCRs), and salivary cortisol (SC) with groups (experimental group and control group) as a between-subjects measure and time as a repeated measure. HR was averaged across each collecting period using the AcqKnowledge 4.0 software and was defined as the number of beats per minute (BPM). In order to get more accurate data, we set the noise rejection to 5% of peak, and the peak interval window to 40–180 BPM. There were two principal measures of relatively long-term tonic EDA states: skin conductance level (SCL) and nonspecific skin conductance responses (NS.SCRs). SCL and NS.SCRs have been widely used as indices of sympathetic nervous system arousal (Boucsein et al., 2012). Firstly, the acquired raw signal is down-sampled, and the low-pass filter is fixed at 1 Hz. Secondly, we set EDA preference below, baseline estimation window width 5 seconds, SCR threshold level 0.01 µs, reject SCRs under 10% of max. For SCL, in order to avoid the possibility of negative values and individual differences in the values of the raw skin conductance data, we use a discrete normalization method to transform the raw skin conductance values, x* = (x − x_min)/(x_max − x_min) (HAN Ying et al., 2018). The SCL was averaged across each measuring period. The NS.SCRs was the sum of the number of SCRs recorded in a 5-min period (Posada-Quintero et al., 2017). To examine whether the stress-induced physiological indicators would be associated with the subjective SAM reports, we conducted Pearson’s correlations between the change rate of the physiological indicators (△HR, △SCL, △NS.SCRs and △SC) and the emotional arousal (b2b1).

Data exclusion: Prior to the main analysis, data from trials in which the time estimates exceeded three standard deviations from the mean of each condition were excluded (0.6% of trials). In addition, data from participants in which there was no linear upward trend in the mean of the interval estimates of 250 ms, 550 ms, and 850 ms for each condition were also excluded (Barlas, 2019). As a result, seven participants were excluded from the stress group and nine from the control group. At last, for the index of the SAM response to stress, HR increase (△HR) was calculated by subtracting the baseline from the maximum value among the other measurements (which was HR during the speech period of the TSST task)(Wu et al., 2017). △HR of the two groups was analyzed separately. In the stress group, five participants were excluded because their △HR was less than 0, which we considered as abnormal heart rates, representing participants whose heart rate did not increase but decreased during the TSST. In the control group, three participants were excluded using the boxplot of SPSS.26 (see Appendix A in supplementary materials). One participant was not included in analyses due to the data of salivary cortisol that was physiologically implausible (2 standard deviations above the response mean; Buchanan et al., 2014). Paired t-tests were conducted on the time estimates of the experimental and control groups to verify the differences in time estimates between voluntary and involuntary conditions. The temporal binding scores were calculated by subtracting the mean reported time interval in voluntary condition from involuntary condition (Barlas, 2019; Poonian & Cunnington, 2013).

A mixed-design ANOVA was conducted on the temporal binding scores of the two groups. The within-subjects factor was the stage (pretest and posttest) and the between-subjects factor was the group (experimental group and control group). The effect sizes and interaction were also checked, followed by a simple effects analysis if significant interaction occurred. All reported p-values were two-tailed, and the alpha value was set at 0.05.

Greenhouse-Geisser corrections were used when sphericity assumptions were violated. Post hoc comparisons were conducted using Bonferroni adjustments to correct for multiple testing. All the reported p-values were two-tailed, and the alpha was set at 0.05. In cases of (borderline) significance, ANOVAs were supplemented with partial eta squared in between-subject designs and generalized eta squared in mixed designs as measures of the effect size.

3. Results

3.1. Questionnaire results

3.1.1. T-AI and SoA scale scores

Trait anxiety scores and SoA scores were analyzed by independent samples t-test, and there was no significant difference between the two groups, ps > 0.05. Trait anxiety scores (59.53 ± 8.99) were significantly negatively correlated with SoPA (26.98 ± 5.27), r = −0.66, p < 0.01, while trait anxiety scores were significantly positively correlated with SoNA (7.77 ± 3.40), r = 0.53, p < 0.01.

3.1.2. SAM scale scores

Emotional valence was expressed as a2a1 and emotional arousal as b2b1. Independent sample t-tests were conducted for a2a1 and b2b1, with the group as the grouping variables. For emotional valence, the independent t-test showed no significant difference in the two groups (experimental group: −0.12 ± 1.45; control group: 0.65 ± 1.46), t (32) = −1.53, p = 0.14, Cohen’s d = −0.76, 95%CI [-1.78, 0.25]. For emotional arousal the independent t-test showed significant differences in the two groups (experimental group:1.82 ± 1.67; control group: −0.24 ± 1.60), t (32) = 3.67, p < 0.01, Cohen’s d = 2.06, 95%CI [0.91, 3.20].

3.2. Results of physiological indicators

3.2.1. Results of heart rate

For heart rate measures (see Table 1), ANOVA revealed a significant interaction between time and group, F(2.77, 88.79) = 16.88, p < 0.01, generalized ${\eta }_p^2$ = 0.35, a significant main effect of time, F(2.77, 88.79) = 29.72, p < 0.01, generalized ${\eta }_p^2$ = 0.48, and group, F(1, 32) = 13.22, p < 0.01, generalized ${\eta }_p^2$ = 0.29. Follow-up simple effects analysis highlighted a significant increase in HR occurring after TSST in the experimental group, with HR_spe, HR_Ari, and HR_0min all significantly higher than the control group, p < 0.05, See Figure 4(A).

Figure 4. Results of physiological indicators across both groups and times. Note: A represents the results of HR (BPM), B represents the results of SCL, C represents the results of NS.SCRs, and D represents the results of SC (nmol/L). where *represents p < 0.05, **represents p < 0.01. Error bars represent SEM.

Here are four plots A, B, C, and D. A represents the results of HR (BPM). There is a significant increase in HR (eg., HRspe, HRAri, and HR0min) occurring after TSST in the experimental group. B represents the results of SCL. A significant increase in SCL during the mental arithmetic of TSST in the experimental group is observed. C represents the results of NS.SCRs. A significant difference is found in two groups during the preparation phase, with the experimental group being higher than the control group. D represents the results of SC (nmol/L). A significant difference between the two groups 15 minutes after the end of TSST is reported.

Table 1. Means and standard deviations for HR(BPM) in two groups and the simple effects of group.

	Experimental group	Control group	F (1,32)	p	${\eta }_p^2$
Baseline	74.27 (6.61)	74.22 (11.89)	0.00	0.99	0.00
TSST before	72.00 (6.23)	71.09 (11.42)	0.08	0.77	0.00
Preparation	75.81 (7.34)	70.28 (8.92)	3.89	0.06	0.11
Speech	97.61 (15.81)	77.64 (8.66)	20.87	0.00**	0.40
Mental arithmetic	98.01 (14.02)	70.11 (9.13)	47.24	0.00**	0.60
TSST 0 min	86.96 (14.91)	71.60 (9.15)	13.09	0.00*	0.29
TSST 15 min	72.10 (8.10)	68.47 (8.88)	1.55	0.22	0.05
TSST 30 min	71.29 (7.57)	67.34 (9.50)	1.79	0.19	0.05

*Represents p < 0.05, **represents p < 0.01.

3.2.2. Results of skin conductance levels (SCL)

For SCL measures (see Table 2), ANOVA revealed a significant interaction between time and group, F(4.18, 133.77) = 2.91, p = 0.02, generalized ${\eta }_p^2$ = 0.08, a significant main effect of time, F(4.18, 133.77) = 18.22, p < 0.01, generalized ${\eta }_p^2$ = 0.36, and group, F(1, 32) = 0.39, p = 0.54, generalized ${\eta }_p^2$ = 0.01. Follow-up simple effects analysis highlighted a significant increase in SCL during the mental arithmetic of TSST in the experimental group, with SCL_Ari, significantly higher than the control group, p = 0.003. Significantly differences were found in two groups before the TSST phase, with the control group being higher than the experimental group, p = 0.04, See Figure 4(B).

Table 2. Means and standard deviations for SCL in two groups and the simple effects of group.

	Experimental group	Control group	F (1,32)	p	${\eta }_p^2$
Baseline	0.18 (0.13)	0.19 (0.12)	0.10	0.75	0.00
TSST before	0.24 (0.17)	0.36 (0.17)	4.23	0.04*	0.12
Preparation	0.44 (0.18)	0.41 (0.16)	0.37	0.55	0.01
Speech	0.50 (0.13)	0.42 (0.19)	2.05	0.16	0.06
Mental arithmetic	0.52 (0.11)	0.38 (0.14)	10.49	0.00*	0.25
TSST 0 min	0.51 (0.11)	0.43 (0.21)	1.74	0.20	0.05
TSST 15 min	0.42 (0.21)	0.44 (0.20)	0.07	0.80	0.00
TSST 30 min	0.46 (0.17)	0.44 (0.24)	0.05	0.82	0.00

*Represents p < 0.05, **represents p < 0.01.

For NS.SCRs measures (see Table 3), ANOVA revealed a significant main effect of time, F(3.88, 124.24) = 4.23, p < 0.01, generalized ${\eta }_p^2$ = 0.12, the main effect of group was insignificant, F(1, 32) = 1.70, p = 0.20, generalized ${\eta }_p^2$ = 0.05, in absence of significant interaction effect (p = 0.29), See Figure 4(C).

Table 3. Means and standard deviations for NS.SCRs in two groups and the simple effects of group.

	Experimental group	Control group	F (1,32)	p	${\eta }_p^2$
Baseline	8.00 (6.61)	7.41 (9.75)	0.03	0.86	0.00
TSST before	7.12 (8.03)	8.18 (7.82)	0.15	0.70	0.00
Preparation	21.76 (12.41)	11.18 (9.05)	8.07	0.00*	0.20
Speech	19.47 (15.30)	16.29 (16.04)	0.35	0.56	0.01
Mental arithmetic	15.29 (15.88)	9.29 (9.33)	1.80	0.19	0.05
TSST 0 min	15.65 (11.51)	13.12 (11.13)	0.42	0.52	0.01
TSST 15 min	14.41 (13.83)	11.12 (7.56)	0.74	0.40	0.02
TSST 30 min	112.65 (10.44)	14.59 (14.59)	0.20	0.66	0.00

*Represents p < 0.05, **represents p < 0.01.

3.2.3. Results of salivary cortisol (SC)

For SC measures (see Table 4), ANOVA revealed a significant main effect of time, F(2.18, 69.75) = 37.43, p < 0.01, generalized ${\eta }_p^2$ = 0.54; the main effect of group was insignificant, F(1, 32) = 1.83, p = 0.19, generalized ${\eta }_p^2$ = 0.05, in absence of significant interaction effect (p = 0.48), see Figure 4(D).

Table 4. Means and standard deviations for SC in two groups and the simple effects of group.

	Experimental group	Control group	F (1, 32)	p	${\eta }_p^2$
Baseline	0.83 (0.25)	0.73 (0.23)	1.38	0.25	0.04
TSST before	1.19 (0.32)	1.19 (0.35)	0.008	0.99	0.00
TSST 0 min	2.09 (0.85)	2.06 (0.72)	0.01	0.93	0.00
TSST 15 min	1.64 (0.35)	1.31 (0.36)	7.04	0.01*	0.18
TSST 30 min	1.09 (0.30)	1.14 (0.24)	0.30	0.59	0.01

*Represents p < 0.05, **represents p < 0.01.

Finally, to examine whether the stress-induced physiological indicators would be associated with the subjective SAM reports, we conducted Pearson’s correlations between the change rate of the physiological indicators (△HR, △SCL, △NS.SCRs and △SC) and the emotional arousal(b2b1) for all participants excluding abnormal physiological data (N = 50). Pearson correlations between △HR and emotional arousal(b2b1) were significant, r = 0.32, p = 0.03. In contrast, there was no significant correlation between the subjective SAM reports and the other physiological measures (SCL, NS.SCRs and SC), ps > 0.1.

3.3. Results of interval estimation

The mean time intervals reported for the different conditions in the two groups are shown in Table 5. The paired t-test for the reporting interval showed that the voluntary condition reported interval (M = 374.24 ms, SD = 80.34 ms) was significantly shorter than the involuntary condition (M = 398.44 ms, SD = 78.45 ms), t(33) = −2.69, p = 0.01, Cohen’s d = −0.46, 95% CI [−42.48, −5.91], indicating that the interval between behavior and subsequent outcomes in the voluntary condition was subjectively compressed to a greater extent than that in the involuntary condition and that there was a temporal binding effect in the voluntary condition (see Figure 5).

Figure 5. Paired t-test for reported time between voluntary and involuntary conditions. It shows that the reported interval for the voluntary condition is significantly shorter than that for the involuntary condition, p < 0.05.

Paired t-test charts of reported interval for all participants in the two different conditions, showed that the reported time in the voluntary condition was significantly shorter than that in the involuntary condition, p < 0.05.

Table 5. The reported intervals for the two groups with different conditions.

		Experimental group		Control group
		Mean	SD	Mean	SD
Pretest	Voluntary condition	364.52	83.97	368.04	77.64
	Involuntary condition	406.18	79.32	379.72	82.20
posttest	Voluntary condition	384.80	85.05	379.62	104.74
	Involuntary condition	402.65	86.72	405.21	97.04

A mixed-design ANOVA with between-subjects factor “group” (experimental and control groups) and within-subjects factor “stage” (pretest and posttest) was conducted for the temporal binding scores. To excluded the effect of heart rate variation on the temporal binding scores, we used HR increase (△HR) as a covariate. The ANOVA showed no significant main effects for stage, group, and △HR, ps > 0.05. The interaction effect between stage and group was significant, F (1,1) = 4.98, p = 0.03, ${\eta }_p^2$ = 0.14. No significant interactions were observed for HR increase and stage, p > 0.1. The three-way mixed ANOVA (between-factor “group,” within-factors “stage” and “time” (250 ms, 550 ms, and 850 ms)) showed significant main effects for time, p < 0.01. No significant interactions were observed for time and group, stage and time and group (see Appendix B in supplementary materials).

A pairwise t-test showed that the temporal binding scores of the pretest (M = 124.98 ms, SD = 172.21 ms) were significantly greater than the posttest (M = 53.54 ms, SD = 179.66 ms) in the stress group, t(16) = 2.85 p = 0.01, Cohen’s d = 0.69, 95% CI [18.28, 124.60], whereas pretest temporal binding scores in the control group (M = 35.04 ms, SD = 178.14 ms) was not significantly different from posttest (M = 76.78 ms, SD = 151.43 ms), t(16) = 1.19, p = 0.25, Cohen’s d = 0.28, 95% CI [-116.30, 32.28] (see Figure 6). An independent t-test showed no significant difference for the temporal binding scores of the pretest (M = 124.98 ms, SD = 41.77 ms) in the stress group and the temporal binding scores of the pretest (M = 53.54 ms, SD = 178.14 ms) in the control group, t(32) = −1.50, p = 0.14, Cohen’s d = 0.51, 95% CI [-32.46, 212.35].

Figure 6. Temporal binding scores for experimental and control groups. Temporal binding scores of the pretest was significantly greater than posttest in the experimental group, while there was no difference in the control group.

Temporal binding scores for the two groups. The temporal binding scores of the experimental group in posttest were significantly lower than the scores in pretest , p < 0.05. There was no difference in the control group.

4. Discussion

The present study examined the effect of acute psychosocial stress on the sense of agency using a persuasive examination manipulation of stress. Generally, the results revealed that stressed participants had significantly lower temporal binding scores, whereas non-stressed participants kept relatively steady temporal binding scores, indicating that acute psychosocial stress leads to a decrease in the sense of agency.

4.1. The validity of the TSST manipulation

No difference in trait anxiety and sense of agency was found between experimental and control groups by examining the results of the T-AI Questionnaire and SoAs questionnaire. While we integrated the data from the two groups, it was found that trait anxiety scores were significantly negatively correlated with SoPA scores and positively correlated with SoNA scores, suggesting a potential association between anxiety and the sense of agency. The relationship between trait anxiety and SoPA/SoNA could be further investigated by larger sample sizes.

In order to draw a persuasive conclusion, massive objective and subjective indicators were carefully examined to make sure the validity of the evoking of stress. Using the SAM scale, we tested the positive change in emotional arousal before and after TSST in the experimental group while negative in the control group, indicating that the TSST manipulation succeeded in arising participants’ emotional reactions.

As supplements to self-reported results, physiological indicators were also analyzed, including heart rate, skin electricity, and salivary cortisol. A total of eight time periods of heart rate were collected, each for 5 min. Independent t-tests revealed a significant difference between the heart rates of the two groups during the two phases of TSST and the 5 min after the end of TSST. The difference can be found in all three time periods, which is consistent with the results of previous studies (Lin et al., 2020).

The same independent t-test for skin conductance level (SCL) was performed for both groups, and the results demonstrated a significant difference between the two groups during the final mental arithmetic phase of TSST. The SCL was measured by the amount of sweat secreted from fingertips contributing to the change of conductance. There was a gradual accumulation of sweat secretion from the fingertips, and the SCL of the experimental group started to rise at the beginning of TSST and reached a peak at the end of the cardio phase (Posada-Quintero et al., 2017). The results of the NS.SCRs of the two groups showed significant differences in the preparation phase of TSST. Participants in the experimental group had significantly higher skin conductance responses in the preparation phase, indicating that the autonomic nervous system was activated at the beginning of the TSST, where the volatility was greatest.

Salivary cortisol was also collected on five occasions, and the results showed a significant difference between the two groups 15 min after the end of TSST. Salivary cortisol is a product of the activation of the HPA axis in humans undergoing acute stress. Therefore, its concentration rises slowly, usually reaching its peak secretion 10–30 min after the end of TSST (Liu et al., 2017). We detected salivary cortisol concentration in the experimental group 15 min after the end of TSST, indicating that the HPA axis was activated in participants under acute psychosocial stress.

Combining the above subjective scores on the SAM scale and objective data on a range of physiological indicators, we conclude that our TSST setting is reasonable and well-validated for the acute psychosocial stress state of the experimental group.

4.2. Examination of the temporal binding effect

As for the implicit measure of the temporal binding effect, the pairwise t-test revealed that the reported interval was significantly shorter under voluntary conditions than involuntary conditions, identifying that the interval between action and subsequent outcome in voluntary conditions was subjectively compressed.

A mixed-design ANOVA was conducted on the temporal binding scores of the two groups, revealing a significant interaction between pre- and post-measures and groups. Simple effect analysis found a significant difference between the pretest and posttest in the experimental group, but not in the control group. Considering that the temporal binding scores after TSST significantly decreased, it can be inferred that acute psychosocial stress has a negative effect on the sense of agency.

Currently, there are two main approaches in empirical research on how affective information moderates SoA. One is to manipulate the affective validity of action effects by presenting positive or negative action outcomes (Christensen et al., 2016; Wen et al., 2015), and the other is to directly manipulate participants’ emotional states (Caspar et al., 2016; Christensen et al., 2019; Saarikallio et al., 2019). Our study falls into the latter category, and the findings are consistent with the results of experiments on sexual arousal (Render & Jansen, 2021), fear and anger (Christensen et al., 2019) where a diminished sense of agency was detected by experimental manipulation that placed participants in specific emotional states. Our results are different from another study that reported that stress did not impair SoA (Stern et al., 2020). There are three main reasons for this difference between the two studies. First, we adopt a group design instead of inducing stress in different blocks to exclude the interference from each other. Second, we adopted the Pretest-posttest designs that we can directly explore the influence of stress on sense of agency. Third, we are in a normal situation rather VR context, the induced stress in our study may be relatively large. Additionally, SoA contains two forms called embodied SoA and non-embodied SoA respectively. The former focuses on the immediate connection between the intention to perform the action and the action itself, and the latter focuses on the link between the action performed by the agent and the intended outcome. Based on the experimental manipulation, it is obvious that our study applies the TSST paradigm to the effect of non-embodied SoA, conversely, the Stern team focus on examining embodied SoA in VR context. Further studies are need to clarify these issues.

Another issue should be mentioned. The sense of agency in our study differs from the concept of the locus of control. The locus of control was considered a personality trait. There are two types of the locus of control: internal and external. The first is characterized by the participant’s belief that the things happening to him or her are a consequence of his or her actions, and the second is because the subject belief that things occurring are due to external factors (de Dios-Duarte et al., 2022). The important aspect of sense of agency is the experience of how one’s action affects the external environment. Further studies are needed to testify whether there is a difference of SoA between two types of the locus of control.

5. Conclusion

The present study investigated the effect of acute psychosocial stress on the sense of agency using the Trier Social Stress Test paradigm. The finding suggests that acute psychosocial stress can lead to strong emotional arousal, while significantly attenuating the time-compression effect as well as the sense of agency. With the support of physiological data, our research points out the potential association between acute psychosocial stress and the sense of agency, inspiring further exploration of physiological impacts through the change of the two variables.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

All raw data obtained in this experiment including questionnaires, physiological measurements and time estimates, and data used for statistical analysis in the article (excluding excluded data), are publicly available on the figshare website: https://doi.org/10.6084/m9.figshare.21758537.v2.

Additional information

Funding

This research was financially supported in part by the National Natural Science Foundation of China (62061136001, 32071055).

Notes on contributors

Yayun Chu

Yayun Chu,Postgraduate students at the Institute of Psychology, Chinese Academy of Sciences, Beijing, China.

Guanhua Huang

Guanhua Huang,Postgraduate students at the Institute of Psychology, Chinese Academy of Sciences, Beijing, China.

Yunyun Li

Yunyun Li, PhD, Postdoctorat the Institute of Psychology, Chinese Academy of Sciences, Beijing, China.

Qin Chen

Qin Chen,Postgraduate students at the Institute of Psychology, Chinese Academy of Sciences, Beijing, China.

Jiajia Liu

Jiajia Liu, Postgraduate students at the Institute of Psychology, Chinese Academy of Sciences, Beijing, China.

Ke Zhao

Ke Zhao,PhD,Associate professor at the Institute of Psychology, Chinese Academy of Sciences, Beijing, China.

Xiaolan Fu

Xiaolan Fu, PhD,Professor at the Institute of Psychology, Chinese Academy of Sciences, Beijing, China.