Perceived control moderates the impact of academic stress on the attention process of working memory in male college students

Abstract

Academic stress is a common long-term stress among the student population and is known to impact working memory within the frontoparietal attention network. Perceived control is an individual variation that may play a buffering role between stress and overall adjustment. In this study, we addressed the moderating effects of perceived control between academic stress and working memory. Fifty-nine male college students participated in the study. Academic stress and perceived control were assessed before participants completed a working memory (n-back) task. Event-related potentials (ERPs) including P2 and P3 were analyzed to examine the attention and maintenance processes of working memory. A moderating effect of perceived control on the relationship between academic stress and working memory was found. For students with low levels of perceived control, academic stress was negatively associated with P2 amplitudes at the high workload (3-back) task, suggesting a negative impact on attention process of working memory. In contrast, academic stress did not affect students with high and moderate levels of perceived control. The results indicate that perceived control may serve as a buffer to protect the cognitive function from the disruption of academic stress.

Keywords:

• Academic stress
• Perceived control
• Working memory
• N-back
• ERP
• P2

1. Introduction

Accumulated researches have shown that chronic exposure to stress increases the risk of neuropsychiatric conditions such as depression and related mood and anxiety disorders (Liston, McEwen, & Casey, 2009). Particularly, stress disrupts the prefrontal cortex (PFC)-dependent functions including attention and executive functions (Liston et al., 2009; Moores et al., 2008). Such impacts are thought to result from a complex interplay among psychological state and function of neuroendocrine responses in the brain (de Kloet, Joëls, & Holsboer, 2005).

As a core executive function, working memory not only maintains information in mind but also operates with it. Working memory is closely associated with other core executive functions, i.e., inhibitory control and cognitive flexibility (for a review see Diamond, 2013). In the present study, we tend to define working memory as a cognitive ability involved in multitasking that requires orientating attention to the task-relevant stimuli, shifting attention, and updating the information held in the workspace. As working memory is necessary for many higher-order cognitive abilities, such as reasoning, language comprehension, and planning (Kane, Conway, Miura, & Colflesh, 2007), it has been a major concern in stress research (Evans & Fuller-Rowell, 2013; Lee & Goto, 2015). Previous results have shown impairments of working memory by chronic stress in animals (Mika et al., 2012; Mizoguchi et al., 2000), as well as in humans among the children (Park et al., 2014), elderly people (Mackenzie, Wiprzycka, Goldstein, & Hasher, 2009) and adults (Evans & Schamberg, 2009). However, the existing results are inconsistent. For example, Klein and Boals (2001) found that more life event stress (assessed with Life Experience Scale) was related to poorer performance on operation-word span working memory task. Lewis et al. (2008) tested the working memory with digit span test in undergraduates and found an improvement in aspects of manipulation of working memory under examination stress, which was assessed with the Perceived Stress Scale (Cohen et al., 1983). Philip et al. (2013) found that the performance of working memory (examined with n-back task) unchanged after exposure to prolonged stress, which was assessed with an Early Life Stress Questionnaire based on Child Abuse and Trauma scale (Sanders & Becker-Lausen, 1995). However, their functional MRI results showed that the group with early life stress demonstrated significantly greater default network deactivation. In addition to the above studies, Yuan et al. (2016) showed that the long period of examination stress had an impact on attention processes in working memory (examined with n-back task) by measuring the event-related potentials (ERPs), though the performance of working memory was not affected. A more recent study suggested that working memory (examined with n-back task), long-term memory and self-reported memory were all impaired by recent life stress exposure (assessed with Stress and Adversity Inventory for Daily Stress) (Shields et al., 2017).

These inhomogeneous results indicated that not all individuals are affected by stress to the same extent. For example, some students performed worse when they are stressed out, while others performed better under stress (Genc, 2017). Is there any individual difference that can help people to alleviate the negative impact of stress or even to get the advantage of stress? Giuliano et al. (2017) suggested that the physiological processes could be a factor moderating working memory capacity among people under chronic stress. However, in the current study we are interested in the factor that relates to coping strategy and can be enhanced through intervention. To our knowledge, perceived control is one such factor (Pallant, 2000).

Perceived control is defined by (Wallston, Wallston, Smith, & Dobbins, 1987) as “the belief that one can determine one’s own internal states and behavior, influence one’s environment and/or bring about desired outcomes” (p.5). According to Pallant (2000), control over the event itself is not feasible, but it is important to believe that one can control the emotional reactions, thoughts and physical symptoms. Pallant (2000) divided the participants who suffered from major events into three groups on the scores of perceived control, and found that all adjustment measures (such as physical symptoms and negative affect) were significantly different across three levels of perceived control. Therefore, he argued that perceived control may play a buffering or moderating role between stress and overall adjustment. As suggested by Steptoe (1989), perceived control could be a common final pathway for various coping responses such as self-statements and cognitive restructuring, and these strategies won’t be effective any more without a belief in control. A study in middle-aged and older adults found that perceived control is related to cognitive performance such as memory, and the mechanisms might lie in adaptive behaviors (Lachman, 2006). The ostensible relationship between perceived control and cognitive performance might be that people who are high in perceived control are less vulnerable to the adverse effect of age on cognitive performance. A great number of clinical research showed that perceived control was associated with adjustments after traumatic stressful events and severity of posttraumatic stress disorders (PTSD) symptoms (Doerfler, Paraskos, & Piniarski, 2005; Frazier, Berman, & Steward, 2001). Perceived control has also been found to be relevant to other disorders, such as depression and anxiety (Brandao, Brites, Nunes, Pires, & Hipolito, 2018; Chorpita & Barlow, 1998). These findings implied that perceived control over stressors may be of evident importance to adaptation to stress. Nevertheless, these studies mainly focus on the role of perceived control in the situations regarding severe magnitude of chronic stress such as PTSD, whereas few studies have considered the role of perceived control in how naturalistic psychosocial stressors impact on the healthy population. To address this issue, we examined the academic stress in the current study, which is a naturalistic and moderate chronic stress.

The aim of the present study was to investigate the role of perceived control between stress and the core executive function, i.e., working memory. To investigate why and how the perceived control mediates the effects, the event-related potentials (ERPs) were analyzed. ERPs can measure neural activity with a high level of temporal resolution. N-back task was used to test working memory, as it is closely associated with the executive function (Kane et al., 2007) and it can elicit typical ERP components (Han, Liu, Zhang, Jin, & Luo, 2013). Two typical ERP components P2 and P3 are widely measured during working memory tasks. They reflect different cognitive aspects of working memory processing. The P2 component is most prominent at frontal-central sites and reflects the direction of attentional resources to stimuli (Yun et al., 2011), and is thought to be related to the early attention process (Luck, 2014). The P3 component is a centro-parietal component with a typical peak latency of 300 ms, and it is associated with the later maintenance of the updating process (Lenartowicz, Escobedo-Quiroz, & Cohen, 2010). P3 amplitudes are sensitive to working memory load, as they decrease with increasing load of n-back task (Watter, Geffen, & Geffen, 2001). Previous results showed that long-term stress affected the attention process of working memory rather than all processes and behavioral performance (Yuan et al., 2016). Thus, we adopted the ERPs technique in the current study to identify the alterations in the dynamic time course of cognitive neural activity during working memory processing. Two levels of working memory task (2-back and 3-back) were used to the test the moderating effect on different workloads.

The most common naturalistic stress, i.e., academic stress, was assessed among healthy college students. Two hypotheses were tested. One was that perceived control would moderate the relation between the academic stress and working memory, which meant that the working memory of individuals with low perceived control would be more vulnerable to academic stress, compared with those with high perceived control. Furthermore, we initially hypothesized that the moderate effect on working memory would be reflected on P2 and/or P3 components.

2. Methods

2.1. Participants

Fifty-nine healthy college students participated in the experiment. The participants in the current study were part of an ongoing large-scale study examining cortisol responses to stress, which was based only on the male population. The average age of the participants was 22.07 (SD = 2.14) years old. All participants signed informed consent and received financial compensation. This study was approved by the Ethics Committee of Human Experimentation in the Institute of Psychology, Chinese Academy of Sciences.

2.2. Procedure and measurements

Upon arrival at the laboratory, participants completed some questionnaires described below before taking the n-back working memory task.

2.2.1. Academic stress

The Hindrance and Challenge Stress Scale (LePine, LePine, & Jackson, 2004) was used to assess participants’ academic stress. The questionnaire consists of 10 hindrance and challenge academic stress items (e.g. “The degree to which your learning progression seems stalled”; “The difficulty of the work required in your classes”). Participants were required to rate the items using a Likert 5-point scale (ranging from 1 = no stress to 5 = a great deal of stress). The ratings were summed up to a total academic stress score. The Hindrance and Challenge Stress Scale has moderate internal consistency reliabilities. Cronbach’s alpha in the current sample was 0.88.

2.2.2. Perceived control

The perceived control was measured using Pallant’s (2000) Perceived Control Of Internal States Scale (PCOISS). The PCOISS consists of 18 items in total, including statements such as “I don’t have much control over my emotional reactions to stressful situations,” and “I have a number of good techniques that will help me cope with any stressful situation (Reversed).” Statements on this scale were scored on a Likert 5-point scale (ranging from 1 = strongly disagree to 5 = strongly agree). The ratings were then summed to a total score. Higher total score indicated higher perceived control. The PCOISS had shown good internal consistency, with a Cronbach’s alpha coefficient of 0.92.

2.2.3. Neuroticism

As neuroticism was found to be highly correlated with daily stress and perceived control (Bolger & Schilling, 1991; Zellars, Perrewe & Hochwarter, 2000), including this variable in our analysis can help rule out the contamination from other related personality traits and reduce potential common method bias (Podsakoff, MacKenzie, Lee, & Podsakoff, 2003). The 8-item neuroticism scale from the Big Five Inventory (BFI: John, Donahue, & Kentle, 1991) was used. Participants rated their agreement or disagreement with each item on a Likert 5-point scale (ranging from 1 = strongly disagree to 5 = strongly agree). The scale has been widely used in many languages and possesses high internal consistency and retest reliability (Benet-Martinez & John, 1998; Zheng et al., 2008).

2.3. Working memory task

Working memory was assessed with a numerical n-back task (n = 2, 3). 2-back and 3-back conditions were chosen because 1-back have showed a high ceiling effect in accuracy in previous studies (Han et al., 2013; Yuan et al., 2016). In addition, they are the most frequently used versions and thus giving the best opportunity for comparison with results of others (Owen, McMillan, Laird, & Bullmore, 2005). Participants were presented with a series of white one-digit numbers (from 1 to 9) on a black background monitor 60 cm away from the participants’ eyes, at a visual angle of (1°×2°). Participants had to indicate whether the current number matched the target number from n-trials previous in the series, by pressing the “match” or “non-match” button with their right or left index finger. The match/non-match button was counterbalanced for the left/right hand. The stimuli were displayed for 500 ms with an inter-stimulus interval of 1300–1700 ms, with 50% match trials. The practice blocks consisted of 22 trials and 23 trials in the 2- and 3-back task respectively. The participants had to reach 80% accuracy for each practice before the experiment (Han, Liu, Zhang, Jin, & Luo, 2013). The experimental blocks consisted of 102 trials and 103 trials in the 2- and 3-back task respectively. The first two (in 2-back) and three (in 3-back) digits of each block were excluded from analyses. The experimental blocks lasted approximately 10 min in total. All participants completed 2-back first and then 3-back task. The stimuli were displayed using E-prime (Version 2.0, Psychological Software Tools Inc., Pittsburgh, PA).

2.4. ERP recordings

Electroencephalograms (EEG) were recorded from 64 sites according to the international 10–20 system using Ag/AgCl electrodes (Neuroscan Inc., USA), with an on-line reference to the left mastoid. The vertical electrooculogram (VEOG) was recorded using two electrodes located above and below the left eye. The horizontal electrooculogram (HEOG) was recorded at about 1 cm from the outer canthus of each eye. The impedance was kept below 5 kΩ. The EEG signals were band-pass filtered (0.05–100 Hz) and amplified with SynAmps2 (NeuroScan Inc., USA), and sampled at 1000 Hz. The EEG data was processed offline using Scan 4.3 (Neuroscan Inc., USA). The data was re-referenced to the average of the left and right mastoids through off-line computations. Eye blinks were identified and corrected using the built-in Ocular Artifact Reduction (OAR) transform algorithm of Neuroscan software (Semlitsch, Anderer, Schuster, & Presslich, 1986). The data was digitally lowpass filtered with 30 Hz ȨFIR filter, half-amplitude cutoff) and segmented into epochs of 200 ms pre-stimulus and 1000 ms post-stimulus. Trials with artifacts exceeding ±100 μV were rejected. And only the correct trials were included in the analyses to prevent ERPs from potential contamination of error-related negativity (Miller, Watson, & Strayer, 2012). The excluded trials for 2-back and 3-back were 16 ± 12 and 32 ± 14 (Mean±SD), respectively. The maximum of excluded trials (on an individual level) for 2-back and 3-back were 51 and 64, respectively. The average accepted trials (which were also correct trials) of 2-back and 3-back after the artifact rejection were 84 ± 12 and 68 ± 14 (Mean±SD), respectively, which were used for further processing.

2.5. Data analysis

For behavioral data, the accuracy and RT were calculated after excluding trials with RT less than 100 ms. Then we checked outliers for accuracy and RT on an individual level. Since there are no data beyond the range of [−3,3] SD, they were all included for analysis. For ERP analysis, trials with RT less than 100 ms were also excluded. The largest amplitudes for P2 and P3 components were observed at electrodes Fz and Pz, respectively. The mean latency for the P2 component was 174 ms in the 2-back condition and 177 ms in the 3-back condition, and thus the mean amplitude of P2 was measured during the time interval from 160 to 190 ms at electrode Fz. The mean latency for the P3 component was 366 ms in the 2-back condition and 334 ms in the 3-back condition, and thus the mean amplitude of P3 was measured during the time interval from 300 to 450 ms at electrode Pz.

We first examined bivariate correlations between all variables (academic stress, perceived control, neuroticism, and age) and ERP index (P2 and P3 amplitudes). Next, in order to test our hypotheses, hierarchical multiple regressions were carried out to predict the amplitudes of ERP components under 2-back and 3-back tasks, respectively. For each regression, the covariates that need to be controlled for (age and neuroticism) were entered in Step 1, followed by academic stress and perceived control in Step 2, and an interaction term of academic stress and perceived control in Step 3. All independent variables were mean-centered prior to model estimation. If a significant moderating effect existed, we would further test the simple slopes (Frazier, Tix, & Barron, 2004). The significance level was set at 0.05.

3. Results

3.1. Descriptive statistics

The results of descriptive statistics and bivariate correlations were shown in Table 1. The average total score of academic stress was 26.15 (SD = 6.87), ranging from 12 to 40. Cronbach’s alpha coefficient was 0.88. The average total score of perceived control was 61.58 (SD = 9.01), ranging from 38 to 81. Cronbach’s alpha coefficient of perceived control was 0.91. The average score of neuroticism was 2.72 (SD = 0.60), ranging from 1.38 to 4.38. Cronbach’s alpha coefficient of neuroticism was 0.77.

Table 1. Descriptive statistics and bivariate correlations among study variables (N = 59).

Variable	Mean	Standard deviation	1	2	3	4	5	6	7	8	9	10	11
1. Academic Stress	26.15	6.87	—
2. Perceived Control	61.58	9.01	−.194	—
3. Neuroticism	2.72	0.6	.326*	−.791**	—
4. Age	22.07	2.14	−.113	.260*	−.147	—
5. 2-back ACC	0.92	0.05	−.261*	.123	−.115	−.129	—
6. 2-back RT (ms)	775	158	−.027	−.085	.081	−.066	−.056	—
7. 3-back ACC	0.78	0.08	−.110	.204	−.129	−.185	.672**	.018	—
8. 3-back RT (ms)	860	162	−.125	.106	−.042	.193	.054	.472**	−.047	—
9. 2-back P2 (μV)	4.79	4.23	−.160	.198	−.031	.260*	.093	−.253	.110	−.209	—
10. 3-back P2 (μV)	4.36	4.03	−.225^+	.279*	−.159	.157	.274*	−.181	.242	−.176	.849**	—
11. 2-back P3 (μV)	3.94	4.00	−.072	−.064	.133	−.271*	.027	−.256	.084	.027	.159	.143	—
12. 3-back P3 (μV)	3.25	3.09	.083	−.110	.128	−.374**	.126	−.045	.324*	−.144	.004	.084	.703**

The values presented here are not mean centered.

⁺p < 0.1, *p < 0.05, **p < 0.01.

For working memory performance (see Table 1), paired t-tests showed that the accuracy of 2-back was significantly higher than that of 3-back (t(58) = 18.152, p < 0.001), and the response time of 2-back was also significantly shorter than that of 3-back (t(58) = −3.987, p < 0.001). For ERP data, Figure 1 showed the grand average waveforms and topographies in 2-back and 3-back tasks. The mean amplitudes of P2 and P3 were listed in Table 1. P2 was not significantly different between 2-back and 3-back (p > 0.14), but P3 was significantly larger in 2-back than that in 3-back (t(58) = 4.199, p < 0.001). The load effect of P3 component (the larger the load, the smaller the P3 amplitudes) was consistent with previous studies (Azizian & Polich, 2007).

Figure 1 Stimulus-locked grand average ERP amplitude (uV) in function of time (milliseconds) elicited during the 2-back (black line) and 3-back (red-orange line) conditions. The topographies are the comparisons of the scalp distribution of the fronto-central P2 (160–190 ms) and the posterior P3 (300–450 ms) between 2-back and 3-back conditions.

3.2. Correlation analysis

We examined the bivariate correlations among all the variables prior to building the regression models. We found that neuroticism was positively associated with academic stress (r = 0.325, p = 0.012), and was negatively associated with perceived control (r = −0.791, p < 0.001). Besides, age was positively associated with perceived control (r = 0.260, p = 0.047). Neuroticism and age were later included in the regression modeling to control for.

3.3. Moderator effects

To test the moderating effect of perceived control, we conducted a series of hierarchical multiple regressions, in which covariates (age and neuroticism) entered in Step 1, academic stress and perceived control in Step 2, and an interaction term in Step 3. For 2-back, we didn’t find any significant results in regression models (ps > 0.1). For 3-back, we found that perceived control moderated the effect of academic stress on P2 amplitudes (p = 0.039), as shown in Table 2, but not on P3 amplitudes (ps > 0.1). The post-hoc power analysis conducted in G*power 3 (Faul, Erdfelder, Lang, & Buchner, 2007) showed that the power of the final multiple regression on P2 amplitudes at 3-back was 81.46% and effect size was 0.208.

Table 2. Results of multiple regression analyses predicting P2 amplitudes at 3-back.

		P2 amplitudes at 3-back
	Variable	B	SE	β	B	SE	β	B	SE	β
Step 1	Age	0.258	0.249	0.137	0.108	0.254	0.058	0.154	0.247	0.082
	Neuroticism	−0.939	0.892	−0.139	1.475	1.463	0.218	1.672	1.421	0.248
Step 2	Academic stress				−2.378	1.951	−0.168	−2.823	1.902	−0.199
	Perceived control				10.386	5.509	0.403^+	10.836	5.344	0.421*
Step 3	Academic stress × Perceived Control							29.872	14.117	0.266*
R^2		0.044			0.116			0.185
△R^2		0.044			0.072			0.069
Model F		F (2, 56) = 1.276			F (4, 54) = 1.773			F (5, 53) = 2.405*

The scores of perceived control and academic stress were log-transformed because of non-normal distributions. Then all independent variables were mean centered before model estimation.

⁺p < 0.1, *p < 0.05, **p < 0.01.

To interpret the significant moderator effects of perceived control, we conducted a simple slope test. As shown in Figure 2, for students who reported high (m = 1.85, p = 0.503) and mean (m = −2.8, p = 0.144) levels of perceived control, there was no significant correlation between academic stress and P2 amplitudes in the 3-back task. However, for students reported low level of perceived control, there was a significant negative effect of academic stress on P2 amplitudes at 3-back (m = −7.5, p = 0.018).

Figure 2 Visual depiction of the interaction between academic stress and perceived control. SD = standard deviation. We computed predicted values of the P2 amplitudes at 3-back for three groups: (1) 1 standard deviation below the mean of perceived control; (2) the mean of perceived control; (3) 1 standard above the mean of perceived control. The simple slope test showed that only for students reported low level of perceived control, there was a significant negative effect of academic stress on P2 amplitudes at 3-back (m = −7.5, p =.018).

4. Discussion

The present study investigated the moderating effect of perceived control between stress and working memory processes. Our results found that perceived control moderated the impact of academic stress on the attention process of working memory as indicated by P2 amplitudes. Furthermore, after probing the interaction, our results showed that the negative association between academic stress and P2 amplitudes was only significant for individuals with low level of perceived control. This result indicated that individuals with low level of perceived control were more vulnerable to the impact of academic stress.

The current study suggested a moderating effect of the perceived control between the academic stress and working memory on the ERP level. This moderating effect is manifested in P2, a component reflecting the early attention process of working memory (Choi et al., 2014; Luck, 2014). Our finding on the P2 is consistent with a previous study, which showed that long-term exam stress affected the P2 but not the P3 component during working memory task (Yuan et al., 2016). P2 is generally believed to be associated with attention allocation and stimulus detection (Luck, 2014). Potts (2004) described P2 as a “frontal selection positivity” in attention studies. Lenartowicz et al. (2010) proposed that these attentional processes relied on context updating in PFC and that P2 reflected the operation of these mechanisms. Considering working memory tasks, a couple of other studies suggested that P2 reflect early stages of information selection (Crowley & Colrain, 2004; Marchand, Lefebvre, & Connolly, 2006). In the present study using an n-back task, we tend to interpret the P2 as an indicator of effectively allocating attention to the task-relevant stimulus. Different from P2, some researchers suggested that the P3 component reflects the later maintenance of the updating process (Lenartowicz, Escobedo-Quiroz, & Cohen, 2010; Watter, Geffen, & Geffen, 2001). Therefore, the result suggested that the moderating effect was on the attention process rather than the maintenance process of working memory.

This study tested two levels of working memory load. Interestingly, we only found the moderate effect at 3-back but not at 2-back load. One potential reason could be that 2-back was not as challenging enough as 3-back. This suggests that the moderate effect only exists in the more demanding task. When the task is easy to handle with, perceived control won’t benefit the process. However, as the task gets challenging, the individual differences in perceived control are reflected. The high accuracy and short response time of 2-back might suggest a ceiling effect (Yuan et al., 2016).

Further analysis with the slope test showed that for individuals with low level of perceived control, the academic stress was negatively associated with P2 amplitudes. This result confirms our hypothesis that the working memory of individuals with low perceived control would be more vulnerable to academic stress. Liston et al. (2009) suggested that long-term stress selectively exerts disruption on attentional control and functional connectivity, within the frontoparietal network that mediates attention shifts. When students are under academic stress, perceived control serves as a buffer to reduce the negative impact of stress on the attention network. However, for those students with low level of perceived control, their attention network is more vulnerable to stress without the buffering protection. Therefore, their P2 amplitudes elicited during working memory task are more sensitive to academic stress.

One possible explanation of the mechanism underlying the moderating effect of perceived control might be related to the effective adjustment. Perceived control helps individuals to adjust their internal states such as emotions and thoughts adaptively (Thompson, Nanni, & Levine, 1994), and thus may lower the risk of suffering from the disruption of stressful events. For example, previous study found that for those women with depression during pregnancy, intervention of enhancing perceived control minimized the child-birth related PTSD (Brandao et al., 2018). Another study found that perceived control was associated with lower levels of worry and symptoms of generalized anxiety disorder (Frala, Leen-Feldner, Blumenthal, & Barreto, 2010). A study with patients having coronary artery bypass grafts found that those with stronger perceived control were less anxious or depressed (Gallagher & McKinley, 2009). In addition, impacts of excess emotion like anxiety is found to be closely related to the frontoparietal attention network, as indicated by P2 component (Bar-Haim, Lamy, & Glickman, 2005). For those with low levels of perceived control, their inner states are more likely to be disrupted by academic stress. Therefore, their attention-executive processes of working memory tend to at risk of academic stress. In contrast, if a person feels capable of taking control over emotions and thoughts, they may be able to adjust to the situation more effectively (Pallant, 2000). And this effective adjustment protects them from disruption to cognitive abilities.

In addition, previous clinical psychology studies have confirmed that interventions of enhancing clients’ control of internal states (including physical reactions and thought processes) is useful in stress management (Pallant, 2000). Perceived control could be one possible mechanism underlying the effectiveness of those interventions, as Steptoe (1989) argued that the coping strategies are only effective as much as people develop a belief in control. The concept of perceived control in the current study includes three main domains according to Pallant (2000), i.e., the emotions, thoughts, and physiological reactions. In the current study, if a college student under high academic stress feels capable of controlling over their feelings (to snap themselves out of a bad mood), thoughts (to distract themselves from obsessions over troubles), and reactions (to do something to relax when they get uptight), they will be in a better position to effectively cope with the stressful encounter.

The possible physiological mechanisms underlying the above effects may be associated with the stress-related responses systems, such as Hypothalamic-Pituitary-Adrenal axis (HPA axis) and Sympathetic Nervous System (SNS). For example, Duan et al. (2013) found that long-term exposure to academic stress decreases the cortisol awakening responses. The disruption of stress hormone function can induce structural remodeling in the prefrontal cortex (McEwen, 2007). The P2 component is most prominent at frontal-central sites (Yun et al., 2011), and thus we assume that it could be more sensitive to the alterations in the neural activity in the prefrontal cortex, compared with P3 component. Evidence from source localization suggested the activity of P2 may be located on the parietal lobe and frontal lobe related to attention (Maeno, Gjini, Iramina, Eto, & Ueno, 2004). In the case of P3, studies have shown that differences in the P3 ERP are related to differential modulations in inferior parietal cortex, including the temporo-parietal junction (TPJ, BA40) (Muckschel, Stock, & Beste, 2014; Wolff, Mückschel, & Beste, 2017), which may relate to response selection (Karch et al. 2010) and has been suggested to play a role in dual-task and sustains executive control (Collette et al. 2005). In the current study, perceived control reduces the academic stress burden on working memory as reflected in the P2 component. Therefore, we assume it helps to maintain homeostasis and the normal function of physiological responses, which benefits the executive functions. Future study would measure the physiological index that can reflect the impact of long-term stress to verify our assumption.

There are several contributions of the current study. Previous studies based on stress and perceived control have mainly focused on patients suffering from PTSD or depression (Brandao et al., 2018; Doerfler et al., 2005; Eways et al., 2018; Frazier et al., 2001). Our attempt with healthy young adults suggests that perceived control is not only a buffer to traumatic stress, but it also serves as a buffer to moderate chronic stress such as academic stress. In addition, to our knowledge, the existing studies on perceived control were constrained to its relationship with subject-wellbeing (Doering et al., 2018; Imel, MacPherson, Schreiber, Lisech, & Dautovich, 2018; Mooney, Elliot, Douthit, Marquis, & Seplaki, 2018; Wallston et al., 1987). The present study investigated the buffering effect of perceived control for cognitive functions, in this case, the working memory.

Our study has several limitations that need to be addressed in future studies. Firstly, we only recruited male college participants. Therefore, cautions must be taken when generalizing the findings in this study to other population. Future studies need to be done to investigate whether the results of this study fit for female students and teenagers. Secondly, concepts such as self-regulation and coping strategies could have similar effect or show an overlap to perceived control, which weakens our conclusion on the specificity of academic stress and perceived control. Future studies should measure and control for those similar factors. In addition, although the n-back task is a standard working memory measure in cognitive neuroscience, it mainly reflects attention processes but not the maintenance or storage-capacity of working memory (Kane et al., 2007). In fact, based on the cognitive framework of working memory, it combined both executive control and temporary maintenance of information (Baddeley, 1986; D'Esposito, 2007; Robbins, 2007). Therefore, future studies need to use paradigms such as digit-span task, to examine whether the moderate effect we found in this study also works for the capacity of working memory. Finally, we only measured working memory in the current study, but it is possible that the moderating effect of perceived control is not special for working memory but also exists in attention processes of other executive functions, such as inhibitory control, and thus further studies should be taken to address this issue.

In conclusion, the current study examined the moderating effect of perceived control between academic stress and the attention processes of working memory. Our result showed that for students with high or mean levels of perceived control, academic stress does not significantly influence working memory. In contrast, for students with low levels of perceived control, academic stress was negatively associated with P2 amplitudes when task load was challenging. The result indicates that perceived control may have buffering effects on academic stress by helping students to adapt to stressful encounters and to alleviate the excess emotion brought by academic stress. This effective adjustment may protect the attentional control and functional connectivity of the frontoparietal network from the disruption of academic stress. Taken together, our study demonstrates that the impact of stress on cognitive functions (in this case, working memory) varies among individuals. For those who believe that they can control over their feelings, thoughts, and reactions, they are less vulnerable to stress in high demanding tasks, and vice versa. Colleges in the future might identify those students with low levels of perceived control and provide academic assistance to deal with stress.

Disclosure statement

The authors declare no conflicts of interest.

Additional information

Funding

This work was supported by the National Natural Science Foundation of China [U1736220] and National Key Research and Development Plan under Grant 2018YFC0831101 and Grant 2016YFB1001201.