Acute stress and human spatial working memory strategy use

Abstract Acute stress can impair human working memory. Little is known, however, about the effects of acute stress on working memory strategies. The goal of this research was to investigate the effects of acute stress on use of a systematic spatial working memory search strategy. Participants (28 females and 20 males per group) completed the Trier Social Stress Test (TSST) or control tasks. Use of a systematic spatial working memory search strategy was measured through performance on the spatial working memory subtest of the Cambridge Neuropsychological Test Automated Battery (CANTAB). The TSST was effective at producing subjective and cortisol stress responses, but there was no significant stress effect on use of a systematic search strategy or working memory search errors. There were also no significant relationships between subjective and cortisol stress responses and use of a systematic search strategy or working memory search errors within the stress group. These results suggest that acute stress does not impair the self-generation or execution of a systematic spatial working memory search strategy.


Introduction
Research examining acute stress effects on working memory has shown that stress can impair working memory (see Shields et al., 2016 for a meta-analytic review), although null (Kuhlmann et al., 2005) and enhancing effects (Cornelisse et al., 2011) have also been reported.Self-initiated strategies play an important role in working memory task performance.For example, serial recall accuracy during operation and reading span tasks is significantly greater when resource-demanding memory strategies highly reliant on cognitive control (sentence, imagery, grouping) are used than when less resource-demanding memory strategies with relatively low reliance on cognitive control (reading, repetition) (Bailey et al., 2008(Bailey et al., , 2011;;Dunlosky & Kane, 2007) are used.Prior research suggests that acute stress exposure can lead to a switch from highly resource-demanding to lower resourcedemanding cognitive control strategies (Plessow et al., 2012;Steinhauser et al., 2007).Therefore, acute stress could decrease use of resource-demanding working memory strategies highly reliant on cognitive control.
While prior research has shown that acute stress can decrease use of spatial and simple associative learning strategies and increase use of stimulus-response and procedural learning strategies (Schwabe et al., 2007;Schwabe & Wolf, 2012), acute stress effects on working memory strategies have not been investigated.The current study examined the effects of acute stress on use of an objectively measured systematic search strategy during performance of the Cambridge Neuropsychological Test Automated Battery (CANTAB) spatial working memory task.We hypothesized that individuals who experienced acute stress would use a systematic search strategy less frequently, and make more spatial working memory errors, than individuals who did not experience acute stress.We also hypothesized that stressed participants who had the highest subjective and cortisol stress responses would use a systematic search strategy less frequently, and make more spatial working memory errors, than those with the lowest subjective and cortisol stress responses.

Participants
Ninety-eight young adults participated in this study (mean age ¼ 21.9, SD ¼ 4.6, range ¼ 18 -34, 56 female).Sample size was determined by results from a previous study (Luethi et al., 2009) that examined acute stress effects on working memory performance accuracy using a working memory task, reading span, for which working memory strategies have previously been assessed (Bailey et al., 2008(Bailey et al., , 2011)).The study showed an effect size of Cohen's d ¼ 0.84 for a stress-induced reduction in working memory performance.The required sample size to detect such a difference is n ¼ 24 per group, which we exceeded in the current investigation.An a priori power analysis to determine the appropriate sample size needed to test for group differences in working memory strategies could not be conducted due to lack of prior research.Participants were recruited from the Saint Louis University Psychology Department research participant pool and the St. Louis, Missouri community.Data from one participant were excluded due to equipment malfunction and from another participant due to lack of effort.Therefore, the analyzed dataset included 96 participants (28 females and 20 males per group).Participants were screened for neurological conditions, active psychiatric conditions, hormonal contraceptives, supplemental sex hormones, corticosteroid-based medications, and tobacco consumption.They were also screened for head injuries with loss of consciousness greater than five minutes.Females participated in the study in the luteal phase of their menstrual cycle.Women in the luteal phase of their menstrual cycle have comparable salvia cortisol stress responses to men whereas women in the follicular phase show significantly lower responses (Kirschbaum et al., 1999).Study procedures were approved by the Institutional Review Board of Saint Louis University, and written informed consent was obtained in accordance with its guidelines.

Stressor and control tasks
Stress group participants completed the Trier Social Stress Test (TSST) (Kirschbaum et al., 1993), which consisted of a five-minute speech preparation period, five-minute speech, and five-minute math task.The experimental panel consisted of one female and one male researcher.Control group participants completed tasks based on Het et al.'s (2009) placebo version of the TSST.

Subjective stress
Subjective stress was measured using a visual analogue stress (VAS) electronic slider scale generated using Qualtrics software (Qualtrics, Provo, UT).Endpoints were labeled "no stress at all" and "very high stress".Participants indicated on the slider how much stress they had experienced during the last 10 minutes.Responses were recorded on a numerical scale of 0 to 100, with higher scores indicating higher perceived stress.

Cortisol
Cortisol was collected from saliva using Salivette collection tubes (Sarstedt, Germany).Samples were stored at À 20 � C until assayed.Cortisol was measured using a commercial immunoassay kit with chemiluminescence detection in the laboratory of Dr. Clemens Kirschbaum, Technical University of Dresden, Germany.Intra-assay and inter-assay coefficients of variation were less than 10% (see Heart Rate Supplement for heart rate measurement methods).

Spatial working memory task
Spatial working memory accuracy and strategy use were assessed using a high-function version of the CANTAB spatial working memory task (Cambridge Cognition, Cambridge, UK) that had two practice trials of three boxes, one trial of six boxes, one trial of eight boxes, three trials of 10 boxes, and three trials of 12 boxes.During each trial, participants tapped boxes on a touchscreen to search for tokens hidden inside them.The number of hidden tokens within a trial equaled the number of boxes.Only one token was hidden at a time.Once a token had been found within a box, a token was not hidden inside it again.A new search to locate the next hidden token then began.Spatial working memory accuracy was measured by summing between search errors, which occurred when a participant touched a box in which a token had previously been found within that trial.

Search strategy measures
Working memory strategy use was measured by assessing the order in which participants searched for hidden tokens within each CANTAB spatial working memory trial.Systematic search strategy use occurs when a participant starts each search within a trial with the same box and then searches through the remaining boxes in the same order.Frequent use of this systematic search strategy is associated with fewer between search errors than infrequent use (Owen et al., 1990).No other working memory strategies have been identified from young adults' search sequences in prior research.
Participants' use of a systematic search strategy was measured by two objective strategy calculations.Each participant's first box-based strategy proportion assessed how frequently they started each search within a trial with the same box.It was calculated by summing the number of search sequences within each trial in which the participant started with a box different from their original start box and dividing by the sum of the number of searches made after the original search for each trial until a token was found in the original start box.A high first box-based strategy proportion indicated infrequent use of a systematic search strategy.
Each participant's sequence-based strategy proportion assessed how frequently they searched boxes in the same order within each trial.To calculate this strategy proportion, a participant's original search sequence was identified.Then, each search sequence within a trial during which a participant deviated from their original search sequence was counted as a strategy sequence error.The number of strategy sequence errors made across all trials was summed and then divided by the total number of searches made across all trials.Skipping boxes in which tokens had already been found within a trial, or retouching boxes in which tokens had already been found within a trial, were not counted as strategy sequence errors if the boxes touched were touched in the same order as they had been touched during the initial search sequence.Therefore, strategy sequence errors were independent of between search working memory accuracy errors.A high sequence-based strategy proportion indicated infrequent use of a systematic search strategy.

Procedure
Participants were tested between the hours of 1200 and 1800 to control for the diurnal cycle of cortisol.They first completed a demographic questionnaire that collected their age, sex, years of education, and hours of sleep the previous night and filler tasks (paper mazes) to acclimate to the laboratory.The order and timing of study procedures are presented in Figure 1.

Data transformation
Subjective stress ratings and cortisol reactivity were not normally distributed and were transformed prior to data analyses using Templeton's (Templeton, 2011) two-step process or a log base 10 transformation, respectively.

Stressor efficacy
Figure 2(a) shows VAS subjective stress ratings at each of the three collection timepoints.Stress group participants reported greater subjective stress immediately after the speech and math tasks than controls (t 94 ¼ 2.18, p ¼ 0.032, Cohen's d ¼ 0.44).Stress group participants had lower subjective stress ratings than controls at baseline (t 94 ¼ À 2.43, p ¼ 0.017, Cohen's d ¼ 0.50).Subjective stress ratings did not significantly differ between groups immediately following completion of the CANTAB spatial working memory task (t 94 ¼ À 1.46, p ¼ 0.155, Cohen's d ¼ 0.30).
Figure 2(b) shows salivary cortisol levels at each of the two collection timepoints.Stress group participants had significantly higher cortisol levels immediately following completion of the CANTAB spatial working memory task than controls (t 94 ¼ 4.67, p < 0.001, Cohen's d ¼ 0.95), but not at baseline (t 94 ¼ 0.26, p ¼ 0.796, Cohen's d ¼ 0.05) (see Heart Rate Supplement and Supplementary Figure 1 for heart rate stress response results).

Group analyses of acute stress effects on spatial working memory errors and search strategy
Mean between search errors, first box-based strategy proportions, and sequence-based strategy proportions are shown in Figure 3.There was no significant group difference in between search errors (t 94 ¼ 0.30, p ¼ 0.767, Cohen's d ¼ 0.06).There was also no significant group difference in use of a systematic search strategy as assessed by first boxbased (t 94 ¼ 0.56, p ¼ 0.578, Cohen's d ¼ 0.13) and sequencebased (t 94 ¼ 0.46, p ¼ 0.649, Cohen's d ¼ 0.09) strategy proportions.Closer adherence to a systematic search strategy, as

Analyses of the relationships between subjective and cortisol stress responses and spatial working memory errors and search strategy
Stress group participants' subjective and cortisol stress responses were not significantly correlated with their between search errors, first box-based strategy proportions, or sequence-based strategy proportions (Table 1).

Discussion
This study investigated the effects of acute stress on use of a systematic spatial working memory search strategy.Robust stress responses were produced, as evidenced by higher subjective stress and salivary cortisol following the TSST than control tasks.Closer adherence to a systematic search strategy was associated with fewer working memory errors.However, acute stress did not significantly affect use of a systematic search strategy or working memory error frequency.
The lack of a significant acute stress effect on use of a systematic search strategy in this study suggests that acute stress may not negatively impact self-generation and execution of systematic working memory strategies.Successful use of a systematic search strategy during the CANTAB spatial working memory task requires relatively rigid, repetitive processing.Prior research has shown that acute stress can decrease cognitive flexibility (Shields et al., 2016).A potential explanation for the lack of an acute stress effect on use of a systematic search strategy in this study is that acute stress may impair flexible cognitive processing more than rigid, repetitive cognitive processing.Future research should explore this possibility.Future research should also examine whether acute stress affects working memory strategy use during operation and reading span tasks.Young adults have self-reported spontaneous use of multiple types of working memory strategies (sentence, imagery, grouping, reading, repetition) that engage diverse cognitive processes to varying degrees during these tasks (Bailey et al., 2008(Bailey et al., , 2011;;Dunlosky & Kane, 2007).Analyzing the effects of acute stress on use of these strategies will further elucidate whether acute stress can alter working  memory strategy use, and if so, the cognitive processes that characterize the types of working memory strategies that are and are not impacted.Prior research has not examined the effects of acute stress on performance of the CANTAB spatial working memory task.Acute stress exposure did not lead to the hypothesized increase in working memory errors.The lack of a stress-induced alteration of systematic search strategy use in this study could partially explain the lack of a stressinduced change in working memory accuracy.In addition, while the CANTAB spatial working memory task does require significant updating of memory for the locations of boxes in which tokens have previously been found within a trial, there is relatively little requirement for attentional shifting, inhibition, or manipulation of information relative to other working memory tasks for which significant acute stress effects have been reported (e.g., operation span, nback, backward digit span; Schoofs et al., 2008Schoofs et al., , 2009)).Therefore, the results of this study suggest that acute stress may not strongly impair the updating subcomponent of working memory.
A major strength of this study is the use of a working memory task that allowed self-initiated working memory strategy use to be objectively measured.The first box-based and sequence-based strategy measures used in this study do not require strategy descriptions, strategy self-reports, or retrospective strategy data collection after task performance.Limitations of this study are the lack of a salivary cortisol measurement immediately before, and the lack of a heart rate measurement immediately after, performance of the CANTAB spatial working memory task.
There was no prior research on acute stress effects on working memory strategy use that we could use to conduct an a priori power analysis.Importantly, robust subjective, cortisol, and heart rate stress responses were produced in this study.This study's failure to reject null hypotheses suggests that acute stress does not have large effects on the self-generation or execution of a systematic spatial working memory search strategy, as tested in the current study.

Figure 1 .
Figure 1.Timeline of study procedures.HR: heart rate recording interval; VAS: visual analogue scale subjective stress rating; S: saliva sample; TSST: Trier Social Stress Test; Control: control tasks; CANTAB: spatial working memory subtest of the Cambridge Neuropsychological Test Automated Battery.

Figure 2 .
Figure 2. (a) Untransformed mean visual analogue scale (VAS) subjective stress ratings made at baseline (VAS1), immediately following the Trier Social Stress Test or control tasks (VAS2), and immediately following the CANTAB spatial working memory task (VAS3).(b) Untransformed mean salivary cortisol levels at baseline (S1) and immediately following the CANTAB spatial working memory task (S2).All participant data points are also plotted.Control: control group participants; Stress: stress group participants.

Figure 3 .
Figure 3. (a) Mean spatial working memory between search errors.(b) Mean first box-based strategy proportions.(c) Mean sequence-based strategy proportions.All participant data points are also plotted.Control: control group participants; Stress: stress group participants.

Table 1 .
Pearson product-moment correlations between stress group participants' subjective stress responses, cortisol stress responses, between search errors, first box-based strategy proportions, and sequence-based strategy proportions.
VAS: Visual Analogue Scale rating of subjective stress; S: salivary cortisol.a All ps > 0.050.