Social stress, performance after-effects and extra-role behaviour

Abstract

The article is concerned with the after-effects of social stress on work performance. In a lab-based experiment, seventy participants were assigned to either a stress condition or a no-stress condition. In the stress condition, participants received fake negative performance feedback and they were ostracised by two confederates of the experimenter. Participants carried out the following tasks: attention and divergent creativity. The effects of social stress were examined at three levels: performance after-effects on unscheduled probe tasks, extra-role behaviour and subjective operator state. The manipulation check confirmed that participants experienced social stress. The results showed after-effects of social stress for some forms of extra-role behaviour (i.e. spontaneous reactions) and for the accuracy component of attention. Furthermore, social stress was found to increase negative affect and to reduce self-esteem. The findings point to the importance of assessing different types of after-effects rather than limiting the methodological approach to instant effects on performance.

Practitioner summary: The study aimed to examine the multiple effects of social stress. Social stress resulted in increased negative affect and lower self-esteem. Furthermore, social stress was found to reduce the propensity of humans to show extra-role behaviour (i.e. providing spontaneous help to others).

Keywords:

• Social stress
• after-effects
• extra-role behaviour
• performance
• negative feedback
• ostracism

1. Introduction

Over recent years, the concept of social stress has gained some interest in the area of ergonomics and work psychology (e.g. Demerouti et al. 2019; Kluge et al. 2019; Gerhardt et al. 2021; Sauer et al. 2019). This increasing interest may be partly driven by the fact that the source of social stress may no longer be limited to human beings but may extend to complex machine agents. For example, when machine agents take over leadership functions from humans (e.g. Wesche and Sonderegger 2019), they may also induce social stress. A recent meta-analysis highlighted the many negative consequences associated with social stress, in particular with regard to job satisfaction, burnout, commitment and counterproductive work behaviour (Gerhardt et al. 2021). Such negative consequences are expected regardless of whether the source of social stress is a human or a machine (Sauer et al. 2019). Overall, all this makes it an important concept for research in ergonomics and work psychology.

1.1. Social stress

The concept of social stress subsumes a range of phenomena, which share the common feature of threatening the self-esteem (or social esteem) of the individual being exposed to the stressor (e.g. Semmer et al. 2007). The phenomena include negative performance feedback (e.g. Holbrook 2002), ostracism (Williams 2007), injustice (e.g. Judge and Colquitt 2004), bullying (e.g. Einarsen et al. 2020), and harassment (e.g. Bowling and Beehr 2006). Overviews of different listings of social stressors may be found elsewhere (e.g. Gerhardt, Semmer, Sauter et al. 2021; Sauer et al. 2019). Social stress is associated with devaluing social messages that are conveyed to the person exposed to the stressor, which may be in a direct form (e.g. negative performance feedback) and an indirect form (e.g. allocating an illegitimate task; Gerhardt et al. 2021).

While there is some evidence that social stress affects the subjective state of the operator in the form of reduced self-esteem (e.g. Krings et al. 2015) and increased negative affect (e.g. Porath and Erez 2007),the evidence is less clear when effects on performance are considered. Recent work has focussed on the question of performance by outlining different underlying mechanisms that may take effect when humans are exposed to social stress (Sauer et al. 2019). Three principal outcomes are conceivable in that social stress (a) has no effect on operator performance (‘blank-out’-mechanism), (b) decreases performance (‘rumination’-mechanism), or (c) improves performance (‘increased-motivation’-mechanism). Most empirical work on social stress was carried out in the field rather than in the laboratory, which makes it more difficult to answer the question about performance-related effects because it is more difficult to measure performance in the field. However, if we look at individual social stressors, negative performance feedback stands out in that considerable work included measures of performance (which perhaps is not surprising given that providing feedback is aimed at improving performance). The seminal meta-analysis of Kluger and DeNisi (1996) concluded that there was a mixed picture for the effectiveness of feedback interventions with over a third of the studies showing decreases in performance rather than the expected increases. Since then, research continued to show that negative feedback can also lead to performance decrements (e.g. Alder 2007; Raver et al. 2012). One of the reasons for the emergence of these mixed results is that feedback has an instrumental and an emotional component. For example, feedback includes helpful advice of what aspect of performance needs improving and how it can be improved (similar to instructions or other forms of training). Due to this instrumental advice, feedback may lead to performance improvements despite the presence of other devaluing elements associated with the feedback message (e.g. suggesting to an operator that he/she is not as good as their colleagues).

Similarly, a mixed picture has emerged for ostracism as a further important social stressor. When operators are socially excluded (e.g. an operator is not asked to join fellow operators for their lunch break), studies found that this form of social stress resulted in performance decrements (e.g. Lustenberger and Jagacinski 2010) but also performance improvements (e.g. Byrne et al. 2016). Further empirical work combined ostracism and negative feedback as a joint experience of social stress, which is not uncommon in work settings (e.g. an operator is ostracised by the team due to poor work performance). This work showed nil effects on performance (e.g. Thuillard et al. 2021; Sauer et al. 2021), hereby supporting the ‘blank-out’-mechanism. Of particular interest is the work of Peifer, Sauer, and Antoni (2020) who induced social stress in the context of a highly complex technical work environment, simulating process control with its multiple task activities and conflicting priorities. They did not find any indication of performance decrement, though social stress caused psychophysiological strain in the form of higher levels of cortisol and increased heart rate.

1.2. Instant effects and after-effects

The review of the research literature indicated that there was no strong evidence for performance-related effects of social stress, with a number of studies finding a nil effect. This raises the question of whether some performance decrements may be hidden (cf. Hockey 1997). The examination of hidden effects of social stress requires a suitable methodological approach that allows the detection of even subtle effects of such sub-optimal working conditions. For this purpose, it is important to make a distinction between instant effects and after-effects when examining the effects of stress at work. For some time now, the adoption of performance protection strategies has been considered a regulatory process, which allows operators to maintain their effectiveness in the advent of sub-optimal working conditions (Hockey 1996; Hockey 1997). This may be referred to as the ‘performance protection mode’. The adoption of such a ‘performance protection mode’ may make it difficult to observe immediate stress-related effects. However, if they are not obvious, it does not mean that they do not exist. Therefore, a broadband approach (Hockey 1983) is needed to detect such subtle effects. The broadband approach aims to assess a wide range of different outcome variables (e.g. primary and secondary task performance, information sampling behaviour, system control actions, psychophysiological operator state, subjective operator state and after-effects). Of particular relevance may be such after-effects because they may emerge after a time lag, at a time when operators had already left the ‘performance protection mode’ (e.g. pilot landed an aircraft and put it into parking position, surgeon finished an operation). One may make a distinction between three types of after-effects: performance after-effects (Hockey 1997), extra-role behaviour (Podsakoff et al. 2000) and affective after-effects (Meier and Semmer 2013). The first two are of particular relevance for the present study.

Performance after-effects refer to decrements in performance after the ‘performance protection mode’ had been left (e.g. surgeon does not detect a medication error on a sheet following the successful completion of a complicated operation). After the completion of this chief set of tasks (which may sometimes coincide with the end of the work shift), certain strategies to protect performance (e.g. increased effort expenditure, cutting corners, negligence of low-priority tasks; Hockey 1997) are no longer applied by the operator. When assessing such performance after-effects, the use of ‘unscheduled’ probe tasks may be promising. This is based on the concept of probe tasks (e.g. Hockey 1997; Earle et al. 2015). Unscheduled probe tasks are characterised by two features. First, they are presented without prior announcement. Second, they do not seem overly complicated or are of a highly critical nature in a job, which allows them to be carried out without expending much effort (e.g. asking an operator to carry out a routine check of a fellow operator’s machine shortly before the end of a shift).

The concept ‘extra-role behaviour’ refers to discretionary behaviour going beyond existing role expectations and providing advantages to the organisation (Van Dyne, Cummings, and Parks 1995). Its meaning is similar to the concept of ‘organisational citizenship behaviour’ (e.g. Podsakoff, et al. 2000). With regard to extra-role behaviour, a conceptual distinction has recently been made between spontaneous reactions and considered responses (Sauer et al. 2022). ‘Spontaneous reactions’ refer to situations in which the operator needs to take a decision within a very short time (e.g. if a fellow operator has technical problems with a printer, the operator needs to decide within seconds whether he/she is going to help or not). This is very different from ‘considered responses’, in which the operator has considerable time to decide of appropriate reaction to a given situation (e.g. a fellow operator leaves a note asking for some assistance in repairing a faulty machine before the start of tomorrow’s work shift). In the research literature, there are examples for both forms of extra-role behaviour. ‘Spontaneous reactions’ may be assessed by means of an experimenter dropping a stack of books in front of a participant (Porath and Erez 2007) while ‘considered responses’ refer to an experimenter requesting a participant to pre-test materials that is to be employed in a future study (Sherrod and Downs 1974). It appears to be important to make this distinction between the two forms of extra-role behaviour because they may differ with regard to sensitivity. It may be socially less acceptable to refuse an explicit request (e.g. helping to pre-test materials) than to ignoring a person in difficulty (e.g. stack of books fallen on the floor). Conversely, ‘spontaneous reactions’ may be more strongly influenced by habits (e.g. ‘It is a reflex of mine to pick up stuff from the floor that someone has inadvertently dropped.’) than ‘considered responses’ (‘Should I really help the experimenter who treated me so badly?’).

In summary, capturing after-effects may represent a promising way of detecting stressful states in operators, which do not manifest themselves openly. Together with the assessment of the subjective operator state (e.g. self-esteem and affect), these two types of after-effects (i.e. performance on unscheduled probe tasks and extra-role behaviour in discretionary tasks) represent the core concepts for a sensitive methodological approach that may allow assessing any emerging patterns of hidden decrement caused by social stress.

1.3. The present study

The present work represents a follow-up study of a preceding experiment, which examined the effects of social stress on performance (Sauer et al. 2021). Since the reference study provided little evidence for stress-related effects on performance, the present work extended the methodological approach by including further relevant outcome variables, which may be suitable to detect effects of social stress of a subtler nature. The additional outcome variables referred to measuring performance after-effects and extra-role behaviour in the form of spontaneous reactions and considered responses.

By and large, the same experimental procedure was used as in the reference study. This notably involved the use of two confederates who helped to create the presence of social stress in the form of ostracism, which was complemented by (fake) negative performance feedback by the experimenter based on the completion of a bogus task. This represented a combination of two social stressors, which is an ecologically valid situation present in real work environments (e.g. after a team member has been branded by negative performance feedback, other team members have begun avoiding his/her company because the team member’s poor performance may represent a threat to overall team performance). Furthermore, the combination of the two stressors increases the strength of the experimental manipulation, which in turn increases the probability that a social stressor will take effect in a participant in a laboratory situation. While the reference study made a comparison between human-induced and machine-induced social stress, the present study focussed on one of the two (i.e. human-induced stress) since little difference had emerged between the two sources of social stress. The same tasks were used (i.e. attention and creativity) as in the reference study but this time they were presented as unscheduled tasks (i.e. they were not announced during the experimental briefing of the participant). Finally, the subjective operator state was assessed with regard to affect and performance-related self-esteem.

Based on the review of the research literature, the following hypotheses were put forward. We expected affect to be negatively impaired by social stress compared to the control condition. Furthermore, we predicted that performance-related self-esteem was lower under social stress than in the control condition. Performance after-effects were expected to manifest themselves more strongly in stress condition than in the no-stress condition. Finally, we assumed that for both types of extra-role behaviour (i.e. spontaneous reactions and considered responses), participants showed less pro-social behaviour under social stress than in the control condition.

2. Method

2.1. Participants and experimental design

Seventy participants (58.6% female), aged between 18 and 62 years (M = 23.70, SD = 6.33), took part in the experiment. They were recruited through an email sent out to students at the University of Fribourg. They were required to have a good understanding of French. Students reading psychology, education, or special education were not allowed to take part because they might have been familiar with experimental procedures that make use of confederates. Participants were paid CHF 20 (about €18) for their participation. The study was approved by the Internal Review Board of the Department of Psychology at the University of Fribourg.

A between-subjects design was implemented, including the two groups in which social stress was either induced or not (i.e. control group). The participants were assigned to one of the two conditions in alternation to avoid any bias. Due to the random assignment of participants to the experimental conditions, there were no differences between groups with regard to age and gender.

2.2. Measures

2.2.1. Performance after-effects

2.2.1.1. Attention

In order to assess attention, we employed the D2-R test (Brickenkamp 1998). The instrument allows the assessment of the test takers’ ability to concentrate, measuring speed and accuracy as the two main components of attentional performance. The test consists of 14 lines, with each line containing 47 letters, which are either a 'd' or a 'p'. Each letter has two inverted commas, which are either placed below the letter, above the letter, or one below and one above it. The task of the test taker is to mark only the letter 'd' if it has both inverted commas either below or above it. Marking any of the other letters is considered to be a wrong response. The test lasted 4 minutes and 40 seconds in total, which makes it virtually impossible for the test taker to mark all target letters within the allocated time. For the analysis of the test results, we focussed on the following two performance indicators (Steinborn and Huestegge 2017): total number of characters processed and number of correctly marked characters (i.e. number of errors were subtracted from the total number having been marked).

2.2.1.2. Creativity

In order to measure creativity, the Alternative Uses Test by Guilford (1967) was chosen. In this test, participants were asked to write down as many alternative ways as possible of a how a spoon could be used. The allocated time was two minutes. The ideas did not have to be limited to the typical uses of a spoon but could be of a creative nature (e.g. using it as a mirror or for digging a hole). Performance was assessed by awarding one point for each proposition.

2.2.2. Extra-role behaviour

2.2.2.1. Spontaneous reaction (pen task)

To measure extra-role behaviour, we used a slightly adapted version of a task used by Porath and Erez (2007). In this task, the experimenter (here: a confederate) deliberately drops the ten pens next to the participant, pretending that she dropped them accidently. The number of pens picked up or shifted towards the confederate were counted, serving as an indicator of extra-role behaviour. We used the less strict criterion of shifting the pens (even by using an object like another pen) given the situation of the COVID pandemic, which might have discouraged some participants to touch and pick up the pens. Participants could obtain a score between 0 and 10.

2.2.2.2. Considerate response (participant feedback questionnaire)

The second type of extra-role behaviour was assessed by the experimenter asking the participant for advice of how she could improve the quality of her experimental study. This outcome measure referred to the extent to which participants were prepared to complete a questionnaire. Being developed by the authors of the article, the questionnaire consisted of five open questions and a closed one that required a yes/no answer. The items of the questionnaire referred to the procedure and the tasks of the experiment just completed (e.g. ‘If you enjoyed a particular task, could you tell us which one it was and why you enjoyed it?’). To assess the level of extra-role behaviour, the length of the participant’s verbal response was analysed. For each open question, a maximum score of 2 could be obtained (0 = no response, 1 = response containing 1-15 words, 2 = response containing more than 15 words). For the multiple-choice question (‘Would you like to be contacted again for taking part in another experiment?’) with a yes/no answer format, a maximum score of 1 could be obtained. This adds up to a total score of 11, which could be awarded.

2.2.3. Subjective state after-effects

2.2.3.1. Positive and negative affect

The Positive and Negative Affect Schedule (PANAS; Watson, Clark, and Tellegen 1988) was used in the French language version (Gaudreau, Sanchez, and Blondin 2006) to assess the participants’ emotional state. The PANAS consists of 20 adjectives, with half of them representing positive emotional states (e.g. enthusiastic) and the other half negative ones (e.g. anxious). Participants were asked to rate each adjective on a Likert scale ranging from 1 (‘Very little or not at all’) to 5 (‘Very much’).

2.2.3.2. State self-esteem

In order to measure the participant's self-esteem, the French version of the state self-esteem scale (SSES; Heatherton and Polivy 1991) was used. The questionnaire consisted of 7 items, which participants were asked to rate on a Likert scale ranging from 1 (not at all) to 7 (completely). The SSES included items such as ‘In this experiment, I felt confident about my abilities’, or ‘In this experiment, I felt frustrated about my performance’.

2.2.3.3. Perceived social stress

To assess perceived social stress the same purpose-built scale was used as in the reference study (Sauer et al. 2021). Serving as a manipulation check, it aimed to capture the specific experimental conditions experienced by participants. The following four items were used: ‘I felt excluded during the experiment’, ‘Auditory negative feedback showed me that I had lower task competencies than the other participants’, ‘I felt alone during the experiment’, and ‘I felt stressed during the experiment’. The four items employed a 7-point Likert scale, using exactly the same scale labelling as the self-esteem scale.

2.3. Manipulation of social stress

2.3.1. Confederates (ostracism)

In order to implement ostracism (and to a smaller extent negative feedback) in the experiment, two confederates of the experimenter adopted the role of participants. They were both female and aged 22 years, being present during all 70 testing sessions. The same experimenter ran all testing session, which had the advantage of the same 3-person team were present in all sessions. This reduced the variability of in the experimental procedure (though it represented a big organisational challenge).

2.3.2. Bogus task (negative feedback)

The element of negative feedback in social stress was induced in this experiment through a bogus task. It consisted of a face recognition task, in which it was difficult to assess one’s own performance and hence negative fake feedback could be given to participants without raising any doubts. The task has been used in a previous experiment (Sauer et al. 2021) and made use of a series of stimuli from a database generated by Biehl et al. (1997). First, a face (the primer) was presented to the participant, followed by two other faces, with the participant having to decide which face of the last two corresponded to the primer. The participant gave the response by pressing the fitting key on the computer. The faces were from three different ethnic backgrounds (Asian, African, and Caucasian) to provide a good match with the cover story (see below). To maintain participants’ focus on the task, a fixation cross was presented when no stimulus was shown.

After each response of the participant, visual feedback (i.e. displaying a happy or sad emoticon) as well auditory feedback (i.e. pleasant high-pitched sound for positive feedback, or unpleasant low-pitched sound for negative feedback). It is important to note that the auditory feedback was of sufficient volume that the two other ‘participants’ could not overhear it, allowing them to gain an impression of each other’s performance.

The bogus task consisted of three distinct phases. A short training phase (consisting of 4 stimuli with an interval of 0.35 s between the primer and the pair of faces) was followed by the two testing phases, with an experimental manipulation occurring in the intervening period between the last two phases (see also procedure below). The feedback in the training phase was always accurate whereas in the testing phase it was fake. In the second phase (i.e. testing prior to negative performance feedback), 17 stimuli were presented at an interval of 2.0 s between the primer and the pair of faces. In the stress condition, 64.7% of the feedback received by participants was negative (i.e. 11 stimuli out of 17) while the confederates received negative feedback for only 35.3% of the stimuli. In the no-stress condition, participants would receive 35.3% of negative feedback (i.e. 6 stimuli out of 17), which corresponded to the level of the confederates. In the third and final phase (i.e. testing following negative performance feedback by experimenter), another 30 stimuli were presented, with the interval between the primer and the pair of faces increasing to 2.7 s and the proportion of negative feedback being at a level of 70.0% (i.e. 21 stimuli out of 30). In the control condition, the proportion of negative feedback was at a level of 30.0% (i.e. 7 stimuli out of 30).

2.4. Procedure

2.4.1. Recruiting and welcoming participants

The participants were recruited through an email sent out to students at the University of Fribourg. When a participant agreed to take part in the experiment, an appointment was made with the experimenter, who met the participant (together with the two confederates) near the cafeteria in the university building to walk over in a group of four to the laboratory, in which the experiment was to take place. In the laboratory, three tables had already been arranged next to each other, with a screen positioned between tables to prevent the ‘participants’ from observing each other. In line with the hygiene standard applied during the COVID-19 pandemic, each table was provided with some disinfectant, distances were respected, and masks were worn by all persons present in the lab. Upon entering the laboratory, the experimenter asked everyone to disinfect their hands and to take a seat at a specific computer, with the participant being requested to sit in front of the first screen in the lab (i.e. this was the one that was being prepared according to the experimental condition). Then, the experimenter explained that the aim of the study was to measure people's ability to recognise faces from different ethnicities. After having responding to any possible questions from the participant, the experimenter asked all three ‘participants’ to sign the consent form.

2.4.2. Completion of bogus task (face recognition)

Participants received oral and written instructions on how to carry out the face recognition task. They were told that they will see a fixation cross on the screen before the appearance of a face, and then another cross followed by a pair of faces placed next to each other. Their task was to indicate as quickly as possible whether the face previously seen was the one on the left or on the right by pressing the ‘l’ or ‘s’ key, respectively.

2.4.3. Social stress condition

In the stress condition, after the completion of the bogus task, all three ‘participants’ received some written feedback on their performance, which clearly showed that the participant performed more poorly than the two confederates. Following the negative feedback for the participant, the experimenter explained that, due to a technical problem, one of the three computers cannot be used for the following part of the experiment, which required that two participants had to work together for the second phase. Immediately, the two confederates expressed the desire to work together, emphasising the fact that they had been better at the preceding task than the participant. The experimenter accepted this request, inviting all three ‘participants’ to continue with the experiment completing the third phase, with the two confederates working together and the participant alone.

2.4.4. No-stress condition

In the no-stress condition, after the completion of the bogus task, all three ‘participants’ also received some written feedback, which this time showed the participant performed as well as the two confederates. Then, the experimenter asked all three ‘participants’ to continue the experiment on their individual computers, with the participant experiencing no social exclusion from the two confederates.

2.4.5. Completing performance tests and questionnaires

At the beginning of this part of the experiment, the two confederates were asked by the experimenter to cease working together since they would not need a computer for the rest of the experiment. At this point, the experimental procedure between the stress condition and the control condition was the same again. The experimenter announced to the ‘participants’ the main part of the experiment has already ended, which represented the most difficult part. She then declared that the ‘participants’ would have to complete a set of extra tasks in the form of performance tests. They had not been informed about these extra tasks during the experimental briefing. The two extra tasks were administered in the following order: creativity (Alternative Uses Test with a time limit of 2 min) and attention (d2-test with a time limit of 4 min 40 s). After having completed the two tasks, the following three questionnaires were given to ‘participants’: participant feedback questionnaire (extra-role behaviour in the form of considerate responses), affect and self-esteem. When the questionnaires were handed out, the pen task was administered, with the confederate sitting left to the participant dropping the 10 pens on the floor, waiting for 5 seconds to allow the participant to react, and then picking the remaining pens up. It was important to choose a moment when the participant was not busy. Before completing the participant feedback questionnaire, it was emphasised to the participant that the completion of this questionnaire was entirely voluntary for the participants. Afterwards, ‘participants’ completed the affect and self-esteem scales, followed by the perceived stress questionnaire, serving as a manipulation check.

2.4.6. Debriefing

In the debriefing, the experimenter first thanked the participant for taking part in the experiment, and then explained to her/him the real purpose of the experiment. In a first step, she introduced the confederates to the participant. The confederates spoke to the participants in a very friendly way, explaining their own role in the experiment and reassuring the participant that ostracising him/her was part of their role and was not meant to be personal. Both, the experimenter and the confederates gave detailed explanations about the incorrect feedback given as part of the face recognition task. The participant was reassured that his/her performance had actually been good. The participant was then informed why the cover story and the manipulation of negative feedback and the ostracism was required in the experiment. They also provided a concrete example of why research in this field was important and what pertinent research questions it could potentially answer. Once again, the experimenter and the confederates repeated the necessity to play a scripted role during the whole experiment to reassure the participant that the use of ostracism and the provision of negative feedback was not meant to be personal. Afterwards, the participant was given the opportunity to ask questions. On several occasions during the debriefing, the experimenter enquired with the participant about her/his emotional state to ensure that there was no persisting negative effect of the social stress manipulation. Finally, the participant completed a form so that she/he could be paid for their participation.

2.5. Data analysis

A t-test for independent samples was used to analyse the differences between groups for the different dependent variables. Cohen’s d was calculated as a parameter for the effect size.

3. Results

3.1. Manipulation check

In order to verify whether the stress induction was effective in our participants, we analysed the difference between the two conditions for the items on perceived social stress (see Table 1). The data indicated that participants in the stress group perceived social stress to be significantly higher than in the control group, which was confirmed by a t-test for independent samples, t(34) = −4.93, p < .001, d = 1.12. This showed that the implementation of the experimental manipulation was successful, with Cohen’s d indicating a large effect.

Table 1. Effects of social stress on outcome measures.

Outcome measure	Social stress	No stress
	Mean (standard error)	Mean (standard error)
Manipulation check
Perceived social stress (1–7)	3.67 (.24)	2.34 (.15)*
Performance after-effects
Attention: speed (no. of characters processed)	519.4 (14.3)	554.94 (12.7)^ns
Attention: accuracy (no. of correctly marked characters)	198.17 (8.40)	223.97 (6.48)*
Divergent creativity (no. of alternative uses)	8.77 (0.46)	7.69 (0.39)^ns
Extra-role behaviour
Spontaneous reaction (% of pens moved)	4.3 (2.8)	42.3 (7.9)*
Considered response (0–11)	4.2 (0.44)	3.6 (0.39)^ns
Subjective state
Positive affect (1–50)	31.10 (.84)	31.50 (.95)^ns
Negative affect (1–50)	17.50 (.98)	14.10 (.69)*
Performance-related self-esteem (1–7)	4.48 (.21)	5.52 (.15)*

*Significant effect at p < .05; ns: non-significant effect.

3.2. Correlations between study variables

Table 2 shows correlation coefficients between all study variables. The analysis revealed some association between spontaneous reactions as one form of extra-role behaviour and two of the subjective state measures, that is, negative affect and perceived stress. Furthermore, a cluster of subjective state variables emerged as being associated with each other, which included perceived stress, self-esteem and negative affect. Although participants were randomly assigned to experimental conditions, we carried out an analysis to ensure that participants in the two experimental groups did not differ with regard to age and sex, with t-tests and Chi²-tests confirming this to be the case (all p > .47).

Table 2. Intercorrelations for study variables.

	Sex	Age	Perceived social stress	Attention: speed	Attention: accuracy	Divergent creativity	Spontaneous reaction	Considered response	Positive affect	Negative affect	Performance related self-esteem
Sex	–
Age	.36*	–
Perceived social stress	−.18	−.15	–
Attention: speed	.01	−.24*	−.29	–
Attention: accuracy	−.08	−.23	−.19	.81*	–
Divergent creativity	.06	.17	−.05	.12	.06	–
Spontaneous reaction	.13	−.07	−.34*	.18	.22	−.07	–
Considered response	.12	−.14	.08	.13	015	−.17	−.07	–
Positive affect	.18	.11	−.21	.22	.14	.14	.10	−.13	–
Negative affect	−.17	−.17	.69*	−.18	−.10	.02	−.32*	.12	−.08	–
Performance related self-esteem	.22	.23	−.72*	.19	.17	.12	.34	−.11	.28	−.57*	–

*Significant effect at p < .05; Sex was coded: 1 = female, 2 = male.

3.3. Performance

3.3.1. Attention

The d2 test was used to measure speed and accuracy as the two dimensions of attentional performance. Attentional speed (measured by counting the total number of characters processed) showed no effect of social stress (see Table 1). This was confirmed by a non-significant t-test, t(34) = 1.96, p = .059, d = 0.44. In contrast, the accuracy dimension of attentional performance was affected, with stressed participants marking fewer characters correctly than non-stressed participants (see Table 1). This difference between groups was found to be significant, t(34) = 2.39, p = .023, d = 0.58.

3.3.2. Creativity

The scores for divergent creativity assessed by the Alternative Uses Test are reported in table 1. Counting the number of alternative uses of a spoon proposed by participants showed no significant difference between the two groups, t(34) = −1.78, p = .085, d = 0.43.

3.4. Extra-role behaviour

3.4.1. Spontaneous reactions

The results showed that the exposure to social stress resulted in less extra-role behaviour being displayed towards the fellow participant (i.e. fewer pens being picked up or pushed towards her) than in the no-stress condition (see table 1). This was confirmed by a significant t-test, with the effect size indicating a large effect, t(34) = 4.595, p < .001, d = 1.09.

3.4.2. Considered response

For considered responses as the second form of extra-role behaviour, no effect of stress exposure was found, with the amount of detailed feedback provided to the experimenter being very similar across conditions (see table 1). The t-test confirmed the absence of a significant difference between the control group and experimental group t(34) = −1.06, p = .297, d = 0.43.

3.5. Subjective state

3.5.1. Positive and negative affect

The PANAS scale distinguishes between positive and negative affect (see Table 1 for the descriptive data of both dimensions). Regarding positive affect, the analysis of the scores revealed no significant differences between the groups, t(34) = 0.33, p = .746, d = .08. However, the analysis of negative affect scores showed that exposure to social stress had a significant negative impact, t(34) = −2.9, p = .007, d = .68.

3.5.2. Self-esteem

The analysis of the SSES scores showed that participants who had not experienced social stress rated their self-esteem significantly higher than stressed participants (see Table 1). This difference between the two groups was statistically highly significant, t(34) = 4.01, p < .001, d = .98.

4. Discussion

The present study aimed to examine the effects of social stress by adopting a sensitive, multi-level methodology that allows detecting even subtle effects of strain in the form of performance after-effects and extra-role behaviour. The results showed that the experimental manipulation of social stress was successful. Performance after-effects were partly observed, that is, for the accuracy component of attentional performance but not for the speed component or for divergent creativity. For extra-role behaviour, a significant effect was recorded for spontaneous reactions but not for considered responses. Finally, most subjective measures (i.e. negative affect and self-esteem) were negatively affected by social stress.

The first important finding refers to the partly confirmed hypothesis that performance after-effects are the result of social stress. Compared to the reference study in which exactly the same tasks were used (with none of these showing any performance impairment), one performance measure showed a decrement in the present study. This may suggest that the vulnerability of task performance during the presence of social stress may be slightly increased when the tasks are unscheduled with a view to capturing after-effects rather than instant effects. It was the accuracy component of attentional performance that showed a clear decrement as a function of stress but not the speed component. This corresponds to a pattern that represents a shift in the speed-accuracy trade-off, which is sometimes observed in the form of faster but less accurate responding when simple tasks (e.g. choice reaction time) are performed under stress (e.g. noise or time pressure; Hockey 1997). In other words, the accuracy component of task performance may be more vulnerable than the speed component. For creative performance, no impairment was observed. Although this result implied that our hypothesis for performance after-effects had to be partly rejected, a closer look at the findings of a meta-analysis (Byron, Khazanchi, and Nazarian 2010) may provide a post-hoc explanation for the hypotheses not being supported. The meta-analysis revealed a rather complex non-linear relationship between the level of stress and creative performance, indicating that low-stress situations caused increases in creative performance whereas high-stress situations resulted in performance decrements. With regard to the present study, we would consider the level of stress to be at a medium level within the intensity range, which might have resulted in the observed nil effect (i.e. stress level was too high to produce an increase in creative performance but too low to result in a performance decrement). Overall, in comparison to the reference study measuring instant effects on performance, performance measures in the present experiment showed higher sensitivity to stress exposure when unscheduled probe tasks were used rather than customary tasks.

The second important finding refers to the partly confirmed hypothesis showing that extra-role behaviour in the form of spontaneous was negatively affected by social stress whereas considered responses were not. For spontaneous reactions, the hypothesised result emerged in that participants not being subjected to social stress were more inclined to help by picking up or shifting the pens towards their fellow participant. This is in line with some research examining this form of extra-role behaviour as a function of stress (e.g. Porath and Erez 2007). Interestingly, such extra-role behaviour did not occur in the form of considered responses, with the amount of helpful advice the participants offered to the experimenter being unaffected by social stress. It may be that for considered responses, social expectations may take effect more strongly (e.g. being helpful to the experimenter in an academic context) than for spontaneous reactions, which are characterised by the need to show an immediate reaction without being able to ponder over the matter. Furthermore, the explicitness of the request for considerate responses makes it more difficulty to refuse it (although the voluntary nature of the request was stressed). The situation associated with spontaneous responses represents a more implicit request, which can be ignored more easily (e.g. by pretending that one has not noticed it). We would also like to point out that the pen task showed a very large effect according to Cohen’s classification (Cohen 1988), being the largest effect size found in the experiment (apart from the manipulation check). This provides some support for the idea of including extra-role behaviour in the methodological approach. This may be due to the voluntary nature of extra-role behaviour, which may increase the probability that the performance protection mode does not take effect.

The third important finding refers to the relative stability of effects for subjective state variables. Exactly as in the reference study, negative affect and self-esteem were affected by social stress in the present experiment whereas positive affect was not. This represents an exact replication of the results. Among the subjective state variable, self-esteem is of particular importance because it is a key concept in the definition of social stress and its underlying mechanisms, with the basic assumption being that the social self is threatened during the exposure to social stress (Gruenewald et al. 2004). The devaluation of the person targeted by social stress is more likely to produce a negative affective reaction rather than having a negative effect on positive emotions. This is in line with the developers of the PANAS scale who claim that ‘… the NA scale (but not the PA scale) is significantly related to perceived stress’ (Watson et al. 1988, p. 1069). Similar arguments have also been put forward by Porath and Erez (2007), favouring the negative affect scale over the positive affect scale in stress research.

It is interesting to note that for all three types of after-effects, at least one outcome measure was affected by social stress, indicating the promising nature of capturing after-affects. However, there appeared to be differences between the three types of after-effects with regard to their sensitivity. The results pattern seems to suggest that extra-role behaviour and subjective operator state was more strongly affected than performance. This was indicated by the number of outcome measures affected for each type of after-effects as well as the effect size found for each outcome measure. Such a difference in sensitivity may also be supported by the theoretical assumption that the performance protection mode is geared towards core tasks of the operator, but does not include activities associated with extra-role behaviour or shielding the subjective operator state. In addition to the performance protection mode, the impediment of important recovery processes such as sleep (e.g. due to rumination) may also play a role in the occurrence of after-effects (e.g. Kottwitz et al. 2021). These after-effects may even be of a longer duration, that is, they extend to the following day.

The study has two main limitations, with both referring to the nature of the experimental tasks (i.e. task duration and nature of task). First, task duration in this experiment (during which the ‘performance protection mode’ may take effect) fell short of typical durations of work shifts. If the probe tasks had been administered after a full work shift (e.g. eight hours of demanding work), levels of fatigue would have been higher, with operators being more likely to leave the performance protection mode completely when finishing the shift. Therefore, the experimental set-up in the current study may have underestimated the impact of social stress on the after-effect measures. Second, a series of static tasks was employed rather than an integrated dynamic environment comprising multiple tasks. The static nature of such static tasks does not fully reflect the dynamic nature of typical work of operators in human-machine systems. Furthermore, it is possible that the tasks used for assessing after-effects were not considered to be sufficiently simple or routine by participants to leave the performance protection mode completely. Instead, participants attached more importance to the task than the experimenter would have liked them to, which again may have underestimated the impact of social stress on the after-effect measures. As a third limitation, one may consider the fact that the experimenter and her two confederates were inevitably aware of the nature of the experimental conditions, which might have influenced their behaviour towards the participants. This in turn might have influenced the behaviour andinternal state of participants, though we did not become aware of such an influence.

In future research, there is a need to extend the exposure of participants to social stressors, paralleled by a longer phase of task completion (i.e. allowing participants to adopt the performance protection mode to the full whereas they completely leave this mode at the end of the shift). Furthermore, future research should model more closely the dynamic features of many real technical work environments (e.g. advanced manufacturing, process control). Such a set-up may also allow a direct comparison between instant effects and after-effects on performance, which was in the current study achieved by making a comparison with a reference study. Both methodological approaches might find more pronounced after-effects at all three levels than the present study, which however (and this is important to note) already showed some effects using the current experimental-up. Lastly, the operation of the performance protection mode was only indirectly assessed (i.e. through the absence of performance decrements). Future research may complement this objective data by administrating a scale that aims to assess the degree to which participants were in a performance protection mode, using subjective data.

Being the first study that made full use of a methodological approach assessing different types of after-effects in the context of social stress, several conclusions can be reached from this work together with the reference study. First, performance became increasingly vulnerable when the task status was moved from a regular task to an unscheduled probe task. Second, there was overall a remarkable consistency in the results pattern between the reference study and the present experiment with exactly the same subjective state variables being affected and showing highly similar effect sizes (e.g. both studies showed a large effect size for self-esteem and a medium effect size for negative affect). This indicates some stability of these strain-related patterns across experimental studies. Third, the conceptual distinction between spontaneous reactions and considered responses appears to be important. The present study provided first evidence that there might be some underlying differences between the two with regard to their sensitivity to stressors, though there is clearly a need to explore this issue further still in future research. Finally, considering that we used a lab-based set-up for the experimental work (which may make it more difficult to induce the performance protection mode in participants than it is possible in real work environments), the current experimental set-up might even underestimate the role of after-effects. Overall, there is a need to continue this line of research on the multiple effects of social stress while adopting a comprehensive methodological approach. If future research were able to confirm this pattern of findings, it would suggest a need for organisations to pay particular attention to the negative effects of social stress, in particular, as any such effects may not be easily discernible by supervisors or managers. This is because the negative effects might be limited to extra-role behaviour and undesirable operator states, which may be less well noticeable than performance decrements in core tasks. This is problematic since organisations benefit strongly from the willingness of their members to engage in such extra-role behaviour.

Abbreviations
PANAS	=	positive and negative affect schedule
SSES	=	state self-esteem scale

Disclosure statement

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

Additional information

Funding

We would like to acknowledge the financial support of the Swiss National Science Foundation (SNSF) for funding this piece of research [No. 173344].