Stress and long-term memory retrieval: a systematic review

Abstract Introduction The experience of stressful events can alter brain structures involved in memory encoding, storage and retrieval. Here we review experimental research assessing the impact of the stress-related hormone cortisol on long-term memory retrieval. Method A comprehensive literature search was conducted on PubMed, Web of Science and PsycNet databases with the following terms: “stress,” “long-term memory,” and “retrieval.” Studies were included in the review if they tested samples of healthy human participants, with at least one control group, and with the onset of the stress intervention occurring after the encoding phase and shortly (up to one hour) before the final memory test. Results Thirteen studies were included in the qualitative synthesis (N = 962) and were classified according to the time elapsed between stress induction and memory retrieval (stress-retrieval delay), the stress-inducing protocol (stressor), the time of day in which stress induction took place, sex, and age of participants. Most studies induced stress with the Trier Social Stress Test (TSST) between 15 and 25 minutes before the final memory (mostly recall) test and showed significant increases in cortisol levels and memory impairment. Discussion The reviewed studies indicate that stress does impair retrieval, particularly when induced with the TSST, in the afternoon, up to 45 minutes before the onset of the final memory test, in healthy young men. These results may inform future research on the impact of stress-induced cortisol surges on memory retrieval.


Introduction
Stressful events are common in everyone's life, inducing physiological changes, such as increased heart rate and sweating. 1 Stressful events can also impact memory and reasoning. Due to their relevance, stress and its main hormone, cortisol, are widely investigated. [2][3][4][5][6][7] Here we review studies on the effects of stress on memory retrieval, highlighting inconsistencies found in the literature and presenting suggestions for future studies.

Stress response and memory retrieval
Stressful events activate both the sympathetic nervous system (SNS), leading to the release of noradrenaline and adrenaline (NA) into the bloodstream by the adrenal medulla, and also the hypothalamicpituitary-adrenal (HPA) axis, leading to the secretion of glucocorticoids (cortisol) into the blood by the adrenal cortex. 3 Stress-induced cortisol release can have a direct impact on the hippocampus and the amygdala, brain structures involved in memory and emotional processes. 8 Cortisol can cross the blood- 1 Departamento de Processos Psicológicos Básicos, Instituto de Psicologia, Universidade de Brasília, Brasília, DF, Brazil.
Submitted Aug 14 2019, accepted for publication Jan 30 2020. brain barrier and bind to glucocorticoid receptors in the hippocampus, thus modulating hippocampal function, and consequently, modulating encoding and retrieval of long-term memories. 9 A considerable amount of research has investigated the impact of stress-related cortisol release on memory processing. Empirical evidence, however, has been inconclusive, with studies showing that episodic memory can be enhanced, 10,11 impaired [12][13][14][15] or remain unaffected [16][17][18] after a stressful event depending on the timing of the event. When the stressor occurs just before the retrieval of consolidated information, memory performance is impaired, 7 as in the case of the student who knows the subject but cannot remember its contents during a high-stakes test.

Present study
In this study, we reviewed the literature published over the last 12 years on the impact of stress on memory retrieval, focusing on some key aspects, such as the time elapsed between stress induction and memory retrieval (stress-retrieval delay) and the protocol used to induce stress in the laboratory (stressor).
Stress-retrieval delay is crucial for understanding how stress modulates memory. 15 The fast stress response occurs seconds after the stressor and involves the release of adrenaline, which increases alertness and facilitates memory encoding. 6 The slow stress response, conversely, occurs several minutes after the stressor and involves the release of cortisol, which impairs the retrieval of consolidated memories. 6 In fact, a recent metaanalysis has concluded that acute stress shortly prior to retrieval can significantly impair memory retrieval. 7 Because stressful events prior to memory retrieval are fairly common (e.g., go blank during a speech), we restricted our systematic review to studies in which stress induction occurred shortly before retrieval.
Stressor type is also important, as some tasks are more effective than others at stimulating cortisol release and emphasize different aspects of the stress response (physiological vs. social). In one common protocol, the Cold Pressor Test (CPT), participants submerge their dominant hand into cold water for a short period of time. 19 The cold water triggers the HPA axis, inducing the release of cortisol. 9 By contrast, in the Trier Social Stress Test (TSST), the participant takes part in a task that includes giving a public speech and doing mental arithmetic aloud. 20 While stress induction in the CPT is driven mainly by physiology, in the TSST stress induction is influenced by social evaluation and unpredictability. 21 Given these differences in the nature of the stressors, we ask whether their impact on memory retrieval may also be different.
In addition to timing and stressor, we also classified studies according to the time of day in which the retrieval session took place (morning vs. afternoon), sex, and age of participants. These factors were deemed important because they are known to affect cortisol response and because previous studies reported conflicting results.
For example, cortisol levels are high in the morning and lower in the afternoon. 22 However, stress-related memory impairment appears not to vary with time of day. 23 Also, stress-related memory impairment seems to be sexand gender-specific, having been observed mostly in young male participants 16 : older participants are less affected, likely due to their lower responsiveness to circulating cortisol levels, 24 and women are less affected depending on the phase of their menstrual cycle. 18,25 We classified the search results along these factors (e.g., sex, age) in order to further explore their moderating influence on stress-related memory deficits.

Method
The systematic review was conducted following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA). 26,27 PRISMA guidelines informed our search strategy, selection criteria, data extraction, and data analyses.

Literature search
A comprehensive literature search was conducted by the first author in January 2020 on three databases: Web of Science, PsycNet, and PubMed. We looked for articles published from January 2008 to December 2019. On the Web of Science and PsycNet databases, the following keywords and Boolean terms were used: ("stress" AND "long-term memory" AND "retrieval"). The search on the Web of Science database was conducted on SCI-EXPANDED, SSCI, A&HCI, CPCI-S, CPCI-SSH, and ESCI indexes. On PubMed, the search was conducted with the following Medical Subject Headings (MeSH terms): ("Stress, Psychological" OR "Stress, Physiological" OR "Stress") AND ("Memory, Long-Term" OR "Long-term memory") AND ("Mental Recall" OR "Retrieval").

Inclusion and exclusion criteria
The following inclusion criteria were taken into consideration: experimental studies published in English, from 2008 to 2019, with samples comprised of healthy human participants, with at least one control group, and with the onset of the stressor intervention occurring after the encoding phase and shortly (up to one hour) before the final memory test. Literature reviews, experimental studies with animals or with human patients (e.g., chronic stress disorders), studies without stress manipulation or with stress manipulation occurring at a time other than before the final memory test (e.g., before or during encoding) were excluded.

Data extraction and management
Eligible studies were selected based on title and abstract screening. Data from eligible studies were individually extracted and recorded in separate databases by the first author. Uncertainties were resolved through discussion with the second author. To improve objectivity, data from each study were organized through standardized forms containing the following categories: First author, year of publication, number of participants, participants' age and sex, stressor, stressretrieval delay, physiological measurement tools (e.g., cortisol saliva sampling), type of stimuli (e.g., word pairs, images, text), and type of final memory test (e.g., free recall).

Results
The initial search conducted on the three databases  As in the CPT, participants in the SECPT also submerge their dominant hand into cold water.

Systematic review findings
Unlike the CPT, however, participants in the SECPT are videotaped by the experimenter, who informs the participant that his/her reactions to cold water will be recorded for future analysis. 31 Only two studies used SECPT as a stressor, 15 Some studies reported a post-hoc analysis whereby participants were grouped as responders (showing an increase of at least 2.5 nmol/l in cortisol levels) and non-responders (see Table 1, last column). The aim was to assess whether cortisol responsivity (regardless of initial assignment to control vs. stress groups) was associated with memory impairment or enhancement.

Stress-retrieval delay
The time elapsed between stressor onset and the final memory test affected memory performance. In five experiments, memory was tested immediately after the stressor, and in 20 experiments there was a delay (Table 2). When the criterial test occurred immediately after the stressor, participants in the stress and control groups produced similar results, except for one study 10 that reported a significant memory enhancement for men. Another study 11 showed, in a post-hoc analysis, that cortisol non-responders also had their memories enhanced. When the criterial test occurred 10-45 minutes after the stressor, results varied with either no change or an impairment in memory performance (Table 2).

Age
Only two studies addressed age as a factor, including older participants (56-76 years of age 16,17 ). Both studies reported similar memory performance for both men and women under both conditions (stress or control). The level of cortisol increase in older participants was smaller than in younger participants, which could account for the lack of stress-related memory impairment in the old age groups.

Sex
Only three studies reported significant differences in performance in the final memory test between men and women. 10,11,16 When both sexes were in groups with successfully increased levels of cortisol, young men's scores on the final memory test were lower than young women's. Women in the luteal phase of the menstrual cycle are less responsive to surges in cortisol levels Numbers in parentheses next to author name represent an experimental condition or group in the corresponding study. -= not available; CG = control group; CPT = Cold-Pressor Test; F = female; LD = participants with low diastolic blood pressure; M = male; NR = cortisol nonresponders; RD = cortisol responders; RP = retrieval practice condition; SECPT = Socially Evaluated Cold-Pressor Test; SG = stress group; TSST = Trier Social Stress Test; TSST-G = TSST group version. * After controlling for changes in energetic and tense arousal, the association between increase in salivary cortisol and memory retrieval was no longer significant. † Only for negative items. ‡ Marginally significant. and their memory is less impaired than men's, despite reliable cortisol increase following stress induction via TSST (the results reported by Schoofs & Wolf 18 contrast with earlier results from the same laboratory 33 ).

Time of day
Cortisol levels are usually higher in the morning when compared to the afternoon. Most studies held the retrieval session in the afternoon, with only two studies having taken place in the morning. 18

Discussion
We reviewed studies assessing the impact of stressinduced cortisol increases on memory retrieval. In particular, we looked at differences in stressors, stress- These stressors, however, were less efficient with older participants of both sexes. 16 The reason why older participants are less affected by these stressors is still an open question. One possibility relates to the fact that cortisol receptors in older participants are less densely distributed and less responsive in memory-relevant areas (hippocampus and prefrontal cortex) than in younger participants. 34,35 This could partly account for the lower stress-induced memory impairments observed so far in older populations. Another possibility is that older participants are more habituated than younger participants to psychosocial stress. This possibility, however, is at odds with findings that older participants under stress (e.g., TSST) release more salivary alpha amylase, a SNS activity marker, than younger participants. 21 In addition, older participants may experience higher levels of subjective stress than younger participants, possibly due to the unfamiliarity and unpredictability of the testing environment, which is more familiar to younger university students. 9 29 particularly if they are in the follicular phase of the menstrual cycle and not taking oral contraceptives. 7,21 By contrast, women taking contraceptives or in the luteal phase of the menstrual cycle have their memories less affected by surges in cortisol levels. 7,18 The lack of sex differences in stressrelated memory impairments is consistent with the conclusions of a recent, focused review. 35 Even though stress impairs retrieval, 7 it does so more strongly for materials that have been encoded via repeated study than for materials that have been encoded via repeated retrieval. 29,42 The latter strategy (retrieval practice) results in more enduring memories, 43 even when compared to active encoding strategies. 44 Because both memory and stress are ubiquitous in learning contexts, 45 it is important to understand how they interact to improve or impair student outcomes.
One direction for future research involves assessing whether the protection against the deleterious effects of stress afforded by retrieval practice is similar to different types of study materials (e.g., easy-vs. hard-to-learn materials), as difficult materials may be particularly affected by stress during high-stakes tests.
One limitation of the present review was its restricted scope. The focus on stress at retrieval was motivated by the ubiquity of stressors during recall tasks (e.g., during exams) and by the promising protective effects of retrieval practice on stress-related memory deficits.
The restricted scope of our review, and consequent small number of eligible studies, also prevented us from conducting a quantitative analysis on the studies returned by the search. As the field progresses, future reviews may be able to quantify the individual roles of age, sex, and type of stressor on stress-related memory deficits.
In conclusion, we reviewed research assessing the impact of stress on memory retrieval. The reviewed studies indicate that stress impairs retrieval, particularly when induced with the TSST in healthy young men. These results may inform future research on the impact of stress-induced cortisol increases on memory retrieval.