From family life to social life to work life, pressure and stress are so ubiquitous in modern life that it is no surprise that psychologists have taken great interest in the impact of pressure and stress on cognitive performance. Although the fields of pressure cognition and stress cognition research have proceeded somewhat independently, it is quite common to assume that pressure is stressful (Beilock & DeCaro, 2007; Masters, 1992; Staal, 2004). Despite the face validity of this assumption, there have been few direct tests of this prediction. This question is particularly important in light of recent studies demonstrating that cognitive performance can vary as a function of the pressure manipulation (DeCaro, Thomas, Albert, & Beilock, 2011) and the stress response (e.g., Ell, Cosley, & McCoy, 2011). In the present research, we have taken an initial step toward integrating pressure and stress research by examining whether pressure is experienced as stressful.

What is pressure?

Individuals experience pressure when they must perform to their potential in order to achieve a goal (Baumeister, 1984). This type of outcome pressure is often induced by increasing the difficulty of reaching some goal and/or by providing an incentive that is contingent on performance. Outcome pressure is thought to coopt working memory and attentional resources, resulting in impairment in cognitive tasks that are dependent on these processes (Beilock & Carr, 2005; Lewis & Linder, 1997; Markman, Maddox, & Worthy, 2006).

Pressure may also be induced by social evaluation or social monitoring (e.g., an evaluative other present and/or videotaping for later evaluation—DeCaro et al., 2011; Gimmig, Huguet, Caverni, & Cury, 2006). Monitoring pressure, in contrast, is more likely to encourage self-monitoring of task performance than to coopt working memory and attentional resources, resulting in impairment in cognitive tasks dependent on procedural knowledge (DeCaro et al., 2011). Consistent with these predictions, outcome pressure has been shown to impair performance on cognitive tasks dependent on working memory, and monitoring pressure impairs performance on more procedural cognitive tasks (e.g., Decaro et al., 2011). Importantly, however, many pressure situations are multifaceted including both aspects of outcome and monitoring pressure. Such combined pressure situations have been argued to negatively impact performance on both working-memory-dependent and procedural-knowledge-dependent tasks, although this prediction has yet to be tested (DeCaro et al., 2011).

What is stress?

As with pressure, stress is a multifaceted construct. Variability exists in individual responses to potential stressors (i.e., events, situations). Individuals experience more distress when the perceived demands of a situation exceed their resources to cope (e.g., Lazarus & Folkman, 1984). Higher levels of distress are marked by the psychological experience of threat, and activation of both the sympathetic nervous system and the hypothalamic pituitary adrenal axis (Blascovich & Tomaka, 1996; Dienstbier, 1989; Lazarus & Folkman, 1984; Lupien, Maheu, Tu, Fiocco, & Schramek, 2007; McEwen & Sapolsky, 1995). When, however, individuals perceive adequate resources to cope with the demands of the situation, they may experience less distress (e.g., Lazarus & Folkman, 1984).

Stressors that exhibit the most robust stress responses in the lab involve performance situations in which individuals are evaluated by others in a domain of personal importance, and in which they are motivated to do well (Blascovich & Tomaka, 1996; Dickerson & Kemeny, 2004). Such social-evaluative stressors share elements of both outcome pressure and monitoring pressure and, like pressure, have been shown to impair cognitive performance depending upon an individual’s stress response (Ell et al., 2011; Kassam, Koslov, & Mendes, 2009; Payne et al., 2007). Although a distress response is commonly associated with negative cognitive task performance, it may facilitate performance on more procedural cognitive tasks that are less dependent on working memory and attentional resources (e.g., Ell et al., 2011). For example, distress has been shown to impair working memory and attentional control (e.g., Plessow, Schade, Kirschbaum, & Fischer, 2012; Schoofs, Preub, & Wolf, 2008), but also to bias processing toward procedural knowledge (Schwabe & Wolf, 2012). These effects may be due in part to the “neuro-symphony” of neurotransmitters and stress steroids released in the stress response (e.g., norepinephrine and cortisol; Joels & Baram, 2009). In contrast to pressure research, in which the variability in the pressure situation (outcome, monitoring, combined) is argued to be of critical import to understanding the consequences of pressure for cognition, the consequences of a stressor on cognition have been argued to depend critically on variability in the stress response.

The present investigation: Is pressure stressful?

Many manipulations intended to increase pressure, particularly monitoring pressure manipulations, have included characteristics that might be expected to lead to a stress response. Indeed, the words “pressure” and “stress” are sometimes used interchangeably in the literature (e.g., Beilock & DeCaro, 2007; DeCaro et al., 2011; Staal, 2004). We propose that integrating the pressure cognition and stress cognition literatures may require consideration of the type of pressure, the stress response, and the cognitive system mediating task performance.

As a model task, we focused on category learning (i.e., the process of establishing a memory trace that improves the efficiency of assigning novel objects to different groups). Category learning has attracted the interest of both pressure cognition researchers (DeCaro et al., 2011; Markman et al., 2006; Worthy, Markman, & Maddox, 2009) and stress cognition researchers (Ell et al., 2011; Schwabe & Wolf, 2012), making it a particularly useful paradigm given our goals. Moreover, extensive evidence has suggested that processing can be biased toward different cognitive systems by simply manipulating the structure of the categories without any procedural changes or changes in how the dependent measure (i.e., the categorization response) is assessed (Ashby & Maddox, 2005). In particular, we focused on the rule-based (RB) and information-integration (II) tasks plotted in Fig. 1. RB and II tasks are argued to be probes for different category-learning systems that compete and vary in their dependence on working memory and attentional resources (with RB tasks being more dependent—Ashby, Alfonso-Reese, Turken, & Waldron, 1998; Ashby & Ell, 2001; Ashby & Maddox, 2005; but see Lewandowsky, Yang, Newell, & Kalish, 2012). For example, Markman et al. trained participants on the Fig. 1 tasks under either low outcome pressure (i.e., participants were instructed to do their best) or high outcome pressure (i.e., participants were led to believe that their performance would determine whether they and a [fictitious] partner would earn a monetary bonus). Consistent with the aforementioned research suggesting that outcome pressure coopts working memory and attentional resources, as well as the assumption of competition between category-learning systems, high outcome pressure impaired performance on the RB task and enhanced performance on the II task.

Fig. 1
figure 1

Scatterplots of the stimuli from the rule-based and information-integration tasks. The unfilled circles represent category A stimuli, the filled circles represent category B stimuli, and the solid lines are the optimal decision boundaries

Full size image

DeCaro et al. (2011) extended this work and tested the hypothesis that the effect of pressure on performance in these cognitive tasks is also dependent on the type of pressure. Using a different set of RB and II tasks (i.e., stimuli varying along four binary-valued dimensions), DeCaro et al. replicated the impairing effect of outcome pressure on an RB task, but observed no effect of outcome pressure on an II task. Monitoring pressure, in contrast, is argued to impair procedural tasks—tasks that are thought to be more sensitive to the effects of self-awareness and self-monitoring. Consistent with this pressure-type hypothesis, DeCaro and colleagues found that monitoring pressure impaired performance on an II, and not an RB, task. Although combined pressure, or pressure that contains elements of both outcome and monitoring pressure, was not investigated, DeCaro et al. argued that combined pressure would negatively impact performance in both RB and II tasks

Ell et al. (2011) investigated the impact of a social-evaluative stressor on the subsequent performance in the RB and II tasks depicted in Fig. 1. The social-evaluative stressor was adapted from the classic Trier Social Stress Test (TSST; Kirschbaum, Pirke, & Helhammer, 1993), which contains strong elements of both outcome and monitoring pressure (in addition to uncertainty and ego relevance elements). In contrast to the pressure manipulations described above, and in common with many investigations of the role of stress in cognition, this stressor occurred prior to the learning tasks (offline) and was not directly relevant to task performance yet participants’ stress remained elevated into the learning period. Ell et al. found that increased distress enhanced performance on the II task and tended to decrease (although not significantly) performance on the RB task. Moreover, distress was associated with increased use of a task appropriate II decision strategy in the II task suggesting a bias away from rule-guided behavior.

Thus, RB and II categorization tasks have been studied in the contexts of both pressure and stress and have yielded mixed results, particularly for performance in II tasks. To begin to integrate the pressure cognition and stress cognition literatures in this area, it will first be necessary to answer the basic question of whether pressure is stressful. In two experiments, we tested the basic question of whether pressure is experienced as stressful (in Exp. 1, outcome pressure; in Exp. 2, combined pressure) and the consequences of this pressure for performance in RB and II tasks.

Experiment 1

In Experiment 1, we replicated the procedure of Markman et al. (2006) while recording markers of the stress response (i.e., distress appraisals, heart rate, blood pressure). We first examined whether the outcome pressure manipulation was experienced by participants as being stressful. We then tested the hypothesis that outcome pressure would impair performance on the RB task and not impair performance on the II task (due to the absence of monitoring pressure; DeCaro et al., 2011), or even enhance performance on the II task (due to competition; Markman et al., 2006). Finally, a stress variability perspective (Ell et al., 2011) would predict enhancement of II only to the extent that the outcome pressure was distressing.

Method

Participants and design

Undergraduates (N = 116) with normal (20/20) or corrected to normal vision participated in a one hour session in exchange for course credit. Participants were randomly assigned to one of four experimental conditions (RB: low pressure = 23, high pressure = 36 participants; II: low pressure = 20, high pressure = 35 participants). One participant was excluded from the RB–low pressure condition for using only one response key throughout the experiment.

Task and procedure

All methods and procedures specific to the category-learning tasks replicated Markman et al. (2006), with one exception. Since the impairing/enhancing effects of outcome pressure were evident across the initial five blocks in Markman et al., participants were trained on either the RB or the II task for only five, instead of eight, blocks of 80 trials. The stimuli were sine-wave gratings weighted by a circular Gaussian filter that varied across trials in spatial frequency (cycles/degree of visual angle) and orientation (degrees of rotation counterclockwise from horizontal). On each trial a single stimulus was presented, and the participant was instructed to make a category assignment by pressing one of two response keys (labeled “A” or “B”) with the index and middle fingers of their dominant hand (blood pressure measurements were taken from their nondominant arm). No time limit was imposed for responses, and corrective feedback was provided immediately after each response (i.e., “Correct” or “Incorrect”). Two points were added to a point meter on the monitor following correct responses, but there was no change in points following incorrect responses. The stimuli were generated and presented using the Psychophysics Toolbox extensions (Brainard, 1997; Pelli, 1997) for MATLAB. The stimuli were displayed on a 17-in. LCD with 1,280 × 1,024 resolution in a dimly lit room.

Participants in the low-outcome-pressure condition were instructed to do their best. Participants in the high-outcome-pressure condition were instructed that they and a (fictional) partner would receive a monetary bonus if they both exceeded a performance criterion of 128 points (i.e., 80 % correct) at the end of training and that their partner had met this criterion. Thus, earning the monetary bonus depended solely on the participant’s performance. Participants were reminded of these instructions at the beginning of the final block. The performance criterion (128 points) was indicated by a line on the point meter.

Distress appraisal

Following the task, participants were asked to rate the extent to which they found the task stressful, demanding, effortful, and distressing, on a 0 (not at all) to 6 (very much) scale. The responses were averaged in order to form a reliable index of distress (α = .82).

Physical reactivity

Upon arrival, sensors to monitor cardiovascular and hemodynamic reactivity were applied (electrocardiogram and continual blood pressure [BP]). Participants then relaxed for a 5-min baseline period. Heart rate (HR) and mean arterial pressure (MAP) were recorded using the BioPac MP150 hardware and BioPac Acquire software. The data were ensemble averaged over relevant minutes using the Mindware software. We calculated the average HR (in beats per minute) and MAP [in mmHg: {2(diastolic BP) + systolic BP}/3] during baseline (last 2 min, most relaxed) and during the final block of trials (first 2 min, highest pressure).Footnote 1 Since earning the monetary bonus was contingent upon performance during the final block, pressure would be expected to be at its peak during the beginning of this block. Reactivity scores were computed by subtracting baseline from this peak level. Thus, positive or negative numbers indicate a rise or decline, respectively, in HR or MAP.

Results

Is outcome pressure stressful?

Distress appraisals

Participants did not find the outcome pressure particularly distressing as all participants reported levels well below the midpoint of the scale (high pressure: M RB = 2.10, SD RB = 1.26; M II = 1.97, SD II = 1.26; low pressure: M RB = 2.07, SD RB = .91; M II = 1.26, SD II = 1.10). Furthermore, a 2 (task) × 2 (pressure) ANOVA revealed no effect of outcome pressure on distress, F(1, 106) = 2.62, p = .11, η p 2 = .02, nor an interaction, F(1, 106) = 2.15, p = .12, η p 2 = .03. We found an effect of task, with participants rating the RB task as being more distressing (M = 2.09, SD = 1.13) than the II task (M = 1.71, SD = 1.24), F(1, 106) = 4.18, p = .04, η p 2 = .04.

Physiological reactivity

Consistent with the low levels of distress, participants did not evidence any physical reactivity to outcome pressure (see Fig. 2). Importantly, no significant differences emerged by pressure condition in baseline HR (all Fs < 1, ps > .40) or baseline MAP (all Fs ≤ 2.34, ps ≥ .13). No effects of pressure condition, task, or the interaction were observed for HR reactivity (all Fs < 1.20, all ps > .27).

Fig. 2
figure 2

Average physiological reactivity at the beginning of the final block during Experiment 1 for heart rate (HR, in beats/minute) and mean arterial pressure (MAP, in mm Hg) as a function of the categorization task and pressure condition

Full size image

Although a significant effect of pressure condition on MAP reactivity did emerge, F(1, 86) = 7.12, p < .05, η p 2 = .08, this effect was not a result of increased MAP in the high-pressure condition (see Fig. 2). No other effects were observed for MAP (Fs < 0.30, ps > .55).Footnote 2

Does outcome pressure impair cognitive performance?

Accuracy analyses

Consistent with the predictions, outcome pressure significantly impaired accuracy in the RB task (see Fig. 3). We examined the effect of pressure by blocks separately for each task condition (with Block as the within-subjects factor, Greenhouse–Geisser corrected for violation of sphericity). As is evident in Fig. 3, participants in all conditions improved in accuracy over blocks [RB: Block, F(2.65, 148.55) = 39.92, p < .05, η p 2 = .42; II: Block, F(3.10, 164.06) = 30.98, p < .05, η p 2 = .01].

Fig. 3
figure 3

Average accuracy across blocks in Experiment 1 as a function of the categorization task and pressure condition

Full size image

Outcome pressure only negatively impacted performance in the RB task [RB: Pressure, F(1, 56) = 5.96, p < .05, η p 2 = .10; II: Pressure, F(1, 53) = 0.69, p = .41, η p 2 = .01], and this effect was consistent across blocks [RB: Block × Pressure, F(2.65, 148.55) = 0.65, p = .57, η p 2 = .01; II: Block × Pressure, F(3.10, 164.06) = .48, p = .71, η p 2 = .01]. Consistent with a pressure-type perspective (DeCaro et al., 2011), performance on the II task was not impaired. We found no evidence of enhancement of II performance in the high-outcome-pressure condition.

Model-based analyses

Analysis of the accuracy data does not directly address the question of what decision strategies were used to perform the categorization tasks. For instance, does the impairment in the RB task reflect a shift to a less optimal decision strategy or an increase in guessing? The following analysis represents a quantitative approach to investigating these questions. Three different types of models were evaluated, each based on a different assumption concerning the participant’s strategy. RB models assume that the participant sets decision criteria on one (or both) stimulus dimensions (e.g., if the bars are wide, respond “A”; otherwise, respond “B”). II models assume that the participant integrates the stimulus information from both dimensions prior to making a categorization decision. Finally, random-responder (RR) models assume that the participant guessed. Each of these models was fit separately to the data from every response block for all participants using a standard maximum likelihood procedure for parameter estimation (Ashby, 1992b; Wickens, 1982) and the Bayesian information criterion for goodness of fit (Schwarz, 1978; see Appendix A for a more detailed description of the models and fitting procedure).

For brevity, we will focus on the results from the final block (Table 1). In the RB task, we observed a reduction in the dominance of RB strategies (and an increase in guessing) in the high-pressure condition, but this shift in the distribution of best-fitting models was not significant [χ 2(2) = 2.49, p = .29]. Nevertheless, the increase in guessing likely contributed to the reduced accuracy in the RB–high-pressure condition, as the subset of participants best fit by RR models (M = 55.83, SD = 4.42) performed much worse than did those best fit by RB models (M = 82.14, SD = 6.85) or II models (M = 76.67, SD = 6.88). In the II task, pressure had little effect on the distribution of best-fitting strategies [χ 2(2) = 0.51, p = .78]. In sum, pressure was neither psychologically nor physiologically stressful, but consistent with previous work, pressure impaired performance on a RB task.

Table 1 Proportions of participants best fit by each model type during the final block
Full size table

Discussion

Our data suggest that the answer to the basic question of whether outcome pressure is stressful is “no.” At the group level, participants did not find outcome pressure to be psychologically or physiologically distressing. Consistent with DeCaro et al. (2011), outcome pressure impaired performance in the RB task and did not impair performance in the II task. Collectively, these data are most consistent with the hypothesis that outcome pressure coopts working memory and attentional resources, leading to selective impairment in the RB task. It may be the case that higher levels of distress are required to evidence the enhancement of II performance demonstrated by Ell et al. (2011). Indeed the distress appraisals in Experiment 1 were quite low, and significantly lower in the II than the RB task. Thus, in Experiment 2 we augment our outcome pressure with monitoring pressure in an effort to increase distress and to test the hypotheses for combined pressure set forth by DeCaro et al. (2011; i.e., impairment in both II and RB tasks).

Experiment 2

In Experiment 2, we tested predictions regarding the consequences of combined pressure for stress reactivity and cognitive performance. To create our combined pressure condition, we augmented the outcome pressure manipulation from Markman et al. (2006) with monitoring pressure by adding elements of social evaluation. Social evaluation is a key component of many classic stress manipulations (see Dickerson & Kemeny, 2004, for a review) and may lead to greater stress reactivity than outcome pressure alone. In addition we examined the theoretical predictions from DeCaro et al. (2011) that combined pressure would impair performance on both RB and II tasks.

Method

Participants and design

Undergraduates (N = 103; normal or corrected-to-normal vision) participating for course credit were randomly assigned to one of the four experimental conditions (RB: low pressure = 28, high combined pressure = 26 participants; II: low pressure = 23, high combined pressure = 26 participants).Footnote 3

Task and procedure

The tasks and procedure were identical to Experiment 1 with the exception that participants in the high-combined-pressure condition experienced monitoring pressure in addition to outcome pressure. To subtly heighten social evaluation, participants engaged in a brief interaction over an intercom with the fictional partner. Following the task instructions, participants introduced themselves to their partner over an intercom. The partner then stated, “I did meet the criterion, so it is all up to you . . . good luck.” The experimenter reinforced the social evaluation of the partner by stating, “Remember, you will have a chance to discuss your performance with your partner afterward” (see Appendix B for the full script). Participants in the low-pressure condition were simply asked to do their best. Physical reactivity and distress appraisals (α = .78) were measured as described in Experiment 1.

Results

Is combined pressure stressful?

Distress appraisal

Although participants in the high-combined-pressure condition (M high = 2.29, SD high = 1.23) reported significantly higher distress appraisals than did participants in the low-pressure condition (M low = 1.71, SD low = 1.23), F(1, 98) = 5.82, p < .05, η p 2 = .06, it should be noted that as in Experiment 1, distress remained below the midpoint of the scale. We observed no effect of task, F(1, 98) = 1.23, p = .27, η p 2 = .01, nor of the interaction, F(1, 98) = 3.34, p = .07, η p 2 = .03.

Physiological reactivity

Adding the subtle manipulation of monitoring pressure to the outcome pressure from Experiment 1 resulted in a modest, but significant, increase in HR and MAP relative to both baseline and the low-pressure condition (see Fig. 4). Importantly, no significant differences in MAP or HR were observed at baseline, although the effect of pressure condition approached significance for HR (M High = 75.88, SD = 11.28; M Low = 71.42, SD = 11.32) [F(1, 85) = 3.33, p = .07, η p 2 = .04; all other Fs < 1.54, ps > .21].

Fig. 4
figure 4

Average physiological reactivity at the beginning of the final block during Experiment 2 for heart rate (HR, in beats/minute) and mean arterial pressure (MAP, in mm Hg) as a function of the categorization task and pressure condition

Full size image

The addition of monitoring pressure to the outcome pressure used in Experiment 1 resulted in significant main effects of pressure for both HR reactivity, F(1, 85) = 13.18, p < .05, η p 2 = .13, and MAP reactivity, F(1, 84) = 9.34, p < .05, η p 2 = .10.Footnote 4 Moreover, the modest increase in HR and MAP observed in the combined-pressure conditions was significantly different from baseline [i.e., 0; HR, t(42) = 6.34, p < .05, d = 1.96; MAP, t(43) = 6.44, p < .05, d = 1.96]. Although the interaction approached significance for MAP, F(1, 84) = 3.69, p = .06, η p 2 = .04, no significant effect of task, or moderation by task, was observed, Fs < 2.14, ps > .14. Thus, adding a subtle manipulation of monitoring pressure to the outcome pressure from Experiment 1 resulted in a modest, but significant, increase in HR and MAP relative to both baseline and low pressure.

Does outcome pressure impair cognitive performance?

Accuracy analyses

Consistent with the predictions from DeCaro et al. (2011), combined pressure impaired performance on both the RB and II tasks (see Fig. 5). As in Experiment 1, the participants in all conditions improved their accuracy over blocks [RB, F(2.88, 149.73) = 50.79, p < .05, η p 2 = .49; II, F(3.27, 147.31) = 24.02, p < .05, η p 2 = .35]. Participants under high combined pressure evidenced significantly lower performance in the RB [F(1, 52) = 4.84, p < .05, η p 2 = .09] and the II [F(1, 45) = 8.36, p < .05, η p 2 = .16] tasks, relative to the low-pressure condition, and these effects were not moderated by block [RB, F(2.88, 149.73) = 0.59, p = .68, η p 2 = .009; II, F(3.27, 147.31) = 1.08, p = .36, η p 2 = .02].

Fig. 5
figure 5

Average accuracy across blocks in Experiment 2 as a function of the categorization task and pressure condition

Full size image

Model-based analyses

In order to investigate any differences in the decision strategies used by participants, the models described in Experiment 1 were also fit to these data. For brevity, we will focus on the results from the final block (Table 2). In both tasks, we observed an increase in the dominance of guessing models in the combined-pressure condition. Although the difference in the distribution of best-fitting models between pressure conditions was statistically significant only in the II task [RB, χ 2(2) = 1.42, p = .49; II, χ 2(2) = 9.46, p < .05], the increase in guessing likely contributed to the pressure impairments in both tasks, as the subset of participants best fit by RR models (M RB = 52.68, SD RB = 3.78; M II = 53.75, SD II = 5.59) performed much worse than those best fit by RB models (M RB = 81.48, SD RB = 8.96) or II models (M II = 70.25, SD II = 3.47).

Table 2 Proportions of participants best fit by each model type during the final block
Full size table

Is distress associated with cognitive performance at these more modest levels?

Previous work has demonstrated enhancement of performance in an II task with higher levels of distress (Ell et al., 2011). As is shown in Fig. 6, the stress reactivity in the combined-pressure condition of Experiment 2 was orders of magnitude lower than that observed from the social-evaluative stressor used by Ell et al. at peak stress (i.e., a modified version of the TSST; reactivity continued into the learning period in Ell et al., 2011).Footnote 5

Fig. 6
figure 6

Average physiological reactivity and distress appraisals from the high-pressure conditions (averaged across tasks) of Experiment 1 (E1), Experiment 2 (E2), and the final 2 min. of a modified version of the TSST that was administered in Ell et al. (2011). HR, heart rate in beats per minute; MAP, mean arterial pressure in mm Hg

Full size image

It is important to note, however, that although all participants faced the same social-evaluative stressor in Ell et al. (2011), individuals showed considerable variability in stress reactivity: The more distress that participants evidenced, the higher their II performance and the lower their RB performance (although the RB effects were not significant). Thus, although mean differences in reactivity by pressure condition may have occurred in Experiment 2, we would not expect the relationship between reactivity and cognitive performance to differ by pressure condition or experiment. Accordingly, we examined the relationships between distress (appraisals, HR reactivity, and MAP reactivity) and cognitive performance within each task condition, collapsed across conditions and experiments.Footnote 6

Although the level of distress in the present studies was more modest than that observed in Ell et al. (2011), the more distress that participants reported, the lower their accuracy in the RB task [r(110) = –.47, p < .001] but not in the II task [r(100) = –.15, p = .17]. This impairment effect was replicated when examining the point-biserial correlation between task-appropriate strategy use (appropriate vs. inappropriate; see Appendix A) and reported distress. The more distress that participants reported, the less likely they were to use task-appropriate rule-based strategies in the RB task [r(110) = –.42, p < .001]. In contrast, distress did not predict appropriate strategy use in the II task [r(100) = –.04, p = .67]. Consistent with the perspective that distress may enhance II performance, even at this modest level of distress, higher heart rate reactivity was associated with better accuracy in the II task [r(86) = .29, p < .01], but not in the RB task [r(91) = .16, p = .13]. This pattern of enhancement in the II task was also evident in the association between heart rate reactivity and the use of task-appropriate strategies [r II(86) = .25, p < .05; r RB(91) = .02, p = .88]. No effects were observed for blood pressure reactivity at these modest levels of distress [accuracy: r II(83) = .06, p = .57; r RB(93) = .16, p = .11; appropriate strategy use: r II(83) = .15, p = .17; r RB(93) = .02, p = .85].

Discussion

In Experiment 2, we added monitoring pressure to the outcome pressure manipulation used by Markman et al. (2006). The addition of social evaluation did indeed raise the intensity of the stress response. Although the result was more modest than the stress response observed in Ell et al. (2011; see Fig. 6), participants did perceive the combined pressure as being more distressing and reported higher HR and MAP reactivity relative to the control condition.

Consistent with the theorizing of DeCaro et al. (2011), combined pressure impaired both RB and II tasks relative to the control condition. The addition of monitoring pressure to the outcome manipulation from Experiment 1 led to an impairment of II performance that was not evident with outcome pressure alone. Model-based analyses suggest that this impairment may be due to an increase in random responding in the II task under combined pressure.

We also found mixed evidence for the relationships between distress and task performance at these modest levels of stress reactivity. As hypothesized, accuracy and appropriate strategy use in RB tasks were impaired by distress—but only self-reported distress. In contrast, accuracy and appropriate strategy use in II tasks were enhanced by distress—but only as marked by heart rate reactivity.

General discussion

The present research represents a modest first step toward integrating the pressure cognition and stress cognition literatures. We believe the concepts of pressure, stress, and cognition to be multifaceted. When examining the effects of a stressor on cognitive performance, it is important to consider the type of pressure exerted (outcome, monitoring, or combined), stress reactivity, and the cognitive system mediating task performance. As an initial step toward examining these broader questions, we first addressed the basic question of whether pressure is indeed stressful. Across two experiments, we examined the stress response to common pressure manipulations, and the consequences for categorization tasks thought to depend on different learning systems.

Is pressure stressful?

In Experiment 1, we found that outcome pressure was not experienced by participants as stressful. This is particularly important for future attempts at reconciling stress cognition and pressure cognition findings as the terms “stress” and “pressure” are often conflated in the literature. In Experiment 2, the addition of a very subtle manipulation of social evaluation increased monitoring pressure and resulted in significant, but modest, stress reactivity to this combined pressure. Participants evidenced significantly higher heart rate reactivity, blood pressure reactivity and distress appraisals in the combined-pressure condition relative to the low-pressure condition. As is noted in Fig. 6, however, this stress response was substantially lower than that observed by Ell et al. (2011). As we work toward integrating the stress cognition and pressure cognition literatures, it may be important to distinguish when, and to what degree, pressure is experienced as distressing.

Does pressure impair cognitive performance?

Consistent with the findings from DeCaro et al. (2011) and Markman et al. (2006), outcome pressure and combined pressure impaired performance on the category-learning task thought to be more dependent upon working memory and attentional resources (i.e., a RB task). These data are consistent with the hypothesis that outcome pressure serves as a distraction/divides attention, impairing performance on tasks dependent on working memory. Indeed, this may explain why participants tended to engage in more random responding in the RB task in the high-pressure conditions relative to the no-pressure conditions, although this effect was not significant in either study.

Our findings for the effects of pressure on II task performance were consistent with the theorizing and findings of DeCaro et al. (2011). Outcome pressure did not impair performance on a more procedural-based category-learning task (i.e., an II task). Combined pressure, in contrast, led to impairments in accuracy and more random responding in the II task relative to the control condition, arguably due to increased self-awareness resulting from monitoring (DeCaro et al., 2011).

We did not find that outcome pressure enhanced II performance, contrary to the findings of Markman et al. (2006). Since the prominent theoretical perspective on the learning of RB and II tasks (Ashby et al., 1998) does not distinguish between types of pressure, the consequences of adding monitoring pressure (i.e., combined pressure) to the predictions of Markman et al. are unclear. If adding monitoring pressure is perceived as simply increasing the intensity of the pressure, perhaps enhancement on the II task might have been expected. We have no easy explanation for our failure to replicate the enhancement observed by Markman et al. Perhaps the participants in Markman et al. found the outcome pressure more distressing than our participants did (i.e., higher levels of distress may be associated with enhancement of II; Ell et al., 2011). In Experiment 1, participants reported low levels of distress overall, and significantly lower distress in the II than in the RB task.

Distress and cognitive performance: More questions than answers?

How do the present findings relate to those observed by Ell et al. (2011)? Ell and colleagues found that distress enhanced performance on an II task. For both accuracy and appropriate strategy use, the higher was a participant’s distress (i.e., appraisals, total peripheral resistance, cardiac output), the better the participant’s II performance. We did not observe enhanced II performance in the only condition that evidenced significant stress reactivity above baseline: combined pressure. Yet, from the stress-variability perspective of Ell et al., individuals vary in their responses to pressure, and those that respond with higher levels of distress (regardless of pressure condition) may evidence enhanced II performance.

Accordingly, we analyzed the associations between our markers of stress reactivity and task performance collapsed across conditions and experiments. Our results were mixed. Consistent with the findings that Ell et al. (2011), observed with different distress markers, higher levels of heart rate reactivity were associated with greater accuracy, as well as with more appropriate strategy use in the II but not in the RB task. It is particularly intriguing to observe this association at these lower levels of stress reactivity when behavior is often unassociated with reactivity (Lupien et al., 2007; Roozendaal, 2002). Indeed, II and RB task performance was unassociated with blood pressure reactivity. In contrast to the physical distress variable predictive of II performance, RB performance was related to the self-report marker of distress. The more that participants reported feeling that the task was stressful, effortful, demanding, and distressing, the lower were their accuracy and appropriate strategy use in the RB task. Self-reported distress was unrelated to performance in the II task.

Reconciling the present mixed findings with those of Ell et al. (2011) is further complicated by differences in the levels of participants’ stress reactivity. Although distress rose with the addition of monitoring pressure in Experiment 2, as can be seen in Fig. 6, distress reactivity was more modest than that observed in Ell et al. In addition to the aspects of outcome and monitoring pressure, the TSST incorporates a number of other factors that are associated with a more robust stress response (e.g., uncontrollability—Dickerson & Kemeny, 2004). Furthermore, the TSST is a 20-min social-evaluative stressor developed to activate the hypothalamic pituitary adrenal axis (HPA) and measurable increases in cortisol (Kirschbaum et al., 1993). Although combined pressure resulted in significant increases for both MAP and HR (relative to baseline), these effects were relatively modest when compared to those observed with the TSST. Thus, it may be that consistent associations between stress reactivity and task performance are more evident with higher levels of reactivity.

Yet, whereas the mixed findings between distress reactivity and performance could be due to the lower intensity of the response, other important factors should be considered. Our analyses assume a linear relationship between distress and cognitive performance. It is conceivable that this relationship could vary across levels of distress. Certainly the “neuro-symphony” (Joels & Baram, 2009) of the stress response is likely to vary (e.g., cortisol is more evident at higher levels of intensity). A broader representation of data across the continuum of the stress response would help clarify whether the relationship between distress and cognitive performance in the RB and II tasks is nonlinear.

In addition, pressure has been implemented in the laboratory setting in methodologically diverse ways. Some researchers have simply used time pressure or performance bonuses, whereas others have incorporated a social-evaluative component (see Staal, 2004, for a review). In addition, pressure manipulations are typically administered online (i.e., the pressure manipulation and task are concurrent), and often the manipulation is directly relevant to the task at hand (i.e., a partner’s bonus is contingent on the participant’s performance). This stands in contrast to the majority of stress cognition research, which has used offline stressors (i.e., the stressor precedes the task, but reactivity may remain high during the task, as in Ell et al., 2011), stressors unrelated to the cognitive task (e.g., a cold pressor; Smeets, Otgaar, Candel, & Wolf, 2008), and/or more intense prolonged stressors known to activate a cortisol response (i.e., TSST). These differences in method may stem from differences in the proposed mediators of the role of pressure and stress in cognitive performance. Much of the pressure literature has posited cognitive explanations, but in contrast, much of the stress literature has posited more biological ones. Our findings may raise more questions than answers, but the questions are important for future work integrating the stress cognition and pressure cognition literatures. Our data suggest that research examining factors related to the intensity, task relevance, type (outcome, monitoring, combined), and timing (online vs. offline) of stressors/pressure manipulations is needed.

Conclusions

The goal of this research was to take an initial, modest step toward integrating the pressure cognition and stress cognition literatures. Although the terms “pressure” and “stress” are often used interchangeably in the literature, our data suggest an important distinction: Pressure is not always stressful. Our findings regarding task performance support the hypotheses of DeCaro et al. (2011) demonstrating that differentiating pressure types may be important for understanding the relationship between pressure and performance in cognitive tasks dependent on different learning systems. Outcome pressure is assumed to impact cognitive performance by coopting working memory and attentional resources, thereby impairing performance on tasks dependent upon executive control processes. Monitoring pressure, in contrast, is assumed to increase awareness of the subcomponents of tasks, thereby impairing performance on more procedural tasks. In our experiments, outcome pressure selectively impaired performance on the RB task, whereas the addition of monitoring pressure led to impairment in both RB and II tasks. The latter finding is novel and contributes to our understanding of the role of combined pressure in cognitive performance. These data also contribute to the growing appreciation in both the stress cognition and pressure cognition literatures of distinguishing the cognitive systems mediating task performance when considering the consequences of pressure or stress. Although our findings for the role of distress in cognitive performance were less clear at the low levels of reactivity that we observed, including measurement of the physiological stress response in future studies of the pressure–cognition relationship will be critical in elucidating this relationship, given the possible interplay between stress response variability and the cognitive systems mediating task performance.