Abstract

Acute stress has been shown to modulate the engagement of different memory systems, leading to preferential expression of stimulus–response (SR) rather than episodic context memory when both types of memory can be used. However, questions remain regarding the cognitive mechanism that underlies this bias in humans—specifically, how each form of memory is individually influenced by stress in order for SR memory to be dominant. Here we separately measured context and SR memory and investigated how each was influenced by acute stress after learning (Experiment 1) and before retrieval (Experiment 2). We found that postlearning stress, in tandem with increased adrenergic activity during learning, impaired consolidation of context memory and led to preferential expression of SR rather than context memory. Preretrieval stress also impaired context memory, albeit transiently. Neither postlearning nor preretrieval stress changed the expression of SR memory. However, individual differences in cortisol reactivity immediately after learning were associated with variability in initial SR learning. These results reveal novel cognitive mechanisms by which stress can modulate multiple memory systems.

INTRODUCTION

Stress can strongly influence how we learn and what we remember. One powerful effect of stress is to change the type of memory that is used to guide behavior, as demonstrated in situations where only one form of memory can be used. For example, if a rodent learns to navigate a “T”-shaped maze and is then tested from an unfamiliar starting point, two distinct outcomes can occur: Behavior can be guided by either memory for stimulus–response (SR) associations (i.e., the link between the maze and the motor response; dependent on the striatum) or memory for the spatial context (i.e., the layout of the maze as a whole; dependent on the hippocampus; Packard & McGaugh, 1996). Research using such dual solution tasks has shown that if rodents learn the task after experiencing acute stress, they consistently use SR rather than context memory (Leong & Packard, 2014; Schwabe, Schachinger, de Kloet, & Oitzl, 2010a; Packard, 2009; Packard & Wingard, 2004; Kim, Lee, Han, & Packard, 2001). Similar effects have been shown in humans (Vogel et al., 2016; Schwabe, Tegenthoff, Hoffken, & Wolf, 2013; Schwabe & Wolf, 2012; Schwabe et al., 2007; but see Schwabe, Oitzl, Richter, & Schachinger, 2009). This bias toward SR memory has been proposed to be a mechanism by which acute stress promotes maladaptive habitual behaviors, such as relapse in drug addiction (Packard, 2009). It can also serve an adaptive function. “Chunking” associations into efficient SR performance units may enable this form of memory to persist, despite distractions (Smith & Graybiel, 2016; Foerde, Knowlton, & Poldrack, 2006). Furthermore, engaging the striatum when the hippocampus was compromised by stress rescued performance in rodents (Vanelzakker et al., 2011). However, key questions remain unanswered.

In dual solution tasks, only one form of memory can be expressed at a time. Thus, the above experiments cannot tell us how each type of memory was influenced individually to cause SR memory to be preferentially expressed. There are several possible cognitive mechanisms that could result in this tendency to engage SR over context memory, each with empirical support: (1) stress could impair context and enhance SR memory; (2) stress could impair context but leave SR memory intact; (3) stress could impair both forms of memory, but to varying degrees; or (4) stress could enhance SR memory but leave context memory intact (for a review, see Goldfarb & Phelps, 2017). Research in nonhuman animals testing each form of memory separately have provided support for two of these mechanisms: Stress before training impaired hippocampal memory but left striatal memory intact (Schwabe, Schachinger, de Kloet, & Oitzl, 2010b), whereas stress after training impaired hippocampal and enhanced striatal memory (Packard & Gabriele, 2009; Wingard & Packard, 2008). Understanding which mechanism is driving this shift in humans has important implications for the malleability of these memory systems and the possibility of reversing this stress-induced bias. Here we examine the influence of acute stress on the consolidation and retrieval of both context and SR memory in humans.

One clue that there may be different mechanisms underlying the preferential engagement of SR over context memory is that, in rodents, this shift occurs regardless of when the stressor occurs in the memory cycle (before learning, after learning, or before retrieval; Perez, Serafin, Prado-Alcala, Roozendaal, & Quirarte, 2017; Leong, Goodman, & Packard, 2015; Hawley, Grissom, Patel, Hodges, & Dohanich, 2013; Elliott & Packard, 2008), whereas research examining these memory systems separately shows that changing the timing of the stressor can lead to opposite effects on hippocampal and striatal memory (Atsak et al., 2016; Guenzel, Wolf, & Schwabe, 2013, 2014; Roozendaal, 2002). This suggests that, by changing stressor timing, we could manipulate the cognitive mechanism by which stress results in the use of SR over context memory. To date, the only time-point that has been studied in humans is stress before learning (Vogel et al., 2016; Schwabe et al., 2007, 2009, 2013; Schwabe & Wolf, 2012). In the experiments presented here, we investigate two questions: How does acute stress immediately after learning (before consolidation) or before retrieval (after consolidation) influence memory for context and SR associations independently? Second, when only one type of memory can be used, will stressors at these different time-points also lead to a bias toward engaging SR over context memory in humans?

There are several methodological challenges associated with addressing these questions. First, because hippocampal memory and striatal memory are historically measured very differently in humans (e.g., declarative recall and motor sequences), it is difficult to compare the magnitude of stress effects using traditional tasks. Second, because stress can influence many cognitive functions in addition to memory (e.g., Arnsten, 2009), it is important to account for broader changes in task performance. Third, to determine if acute stress differentially impacts consolidation or retrieval of SR and context memory, it is critical to account for differences in initial learning. Finally, to target the consolidation or retrieval phases of memory, it is necessary to ensure that these memory phases are distinct and that a stress response after learning (during consolidation) does not remain elevated during retrieval.

We recently developed a task that enables us to address these challenges (Goldfarb, Shields, Daw, Slavich, & Phelps, 2017; Goldfarb, Chun, & Phelps, 2016). In this task, participants search for a target among distractors. There are two embedded memory cues—repeated contexts and probabilistic SR associations—that can facilitate performance, measured by RT. This form of context memory depends on the hippocampus (Chun & Phelps, 1999) and use of context and SR cues to guide attention has been associated with BOLD signal in the hippocampus and striatum, respectively (Goldfarb et al., 2016). This search task gives us comparable assays of these two forms of memory (both similarly reduce RT). Memory-guided attention is implicit and measured relative to performance on trials with no memory cues, helping account for nonmnemonic stress effects. By recording RT on every trial, we obtain a record of how each participant learns the cued associations. Last, to separate consolidation from retrieval processes, participants completed the search task over two consecutive days, enabling us to test memory retrieval 24 hr after learning. This way, any postlearning stress effects subside before retrieval.

Using this design, we can comparably assess the influence of acute stress on consolidation (Experiment 1) and retrieval (Experiment 2) of context and SR memory and test whether the cognitive mechanisms underlying the shift toward SR memory vary based on stressor timing.

EXPERIMENT 1: POSTLEARNING STRESS

In this experiment, we investigated whether stress immediately after learning would change consolidation of context and SR memories, resulting in differences in memory retrieval 24 hr later. To inform our hypotheses, we turned to previous studies examining effects of postlearning stress on hippocampal and striatal memory.

The effects of postlearning stress on hippocampal memory are mixed, with some studies reporting that stress at this time-point impairs hippocampal memory (Trammell & Clore, 2014; Packard & Gabriele, 2009; Wingard & Packard, 2008), whereas others have found enhanced memory, particularly for highly arousing information (Smeets, Otgaar, Candel, & Wolf, 2008; Okuda, Roozendaal, & McGaugh, 2004; Cahill, Gorski, & Le, 2003). Arousing memoranda can evoke an adrenergic response, which may interact with the stress-induced hypothalamic-pituitary-adrenal axis response to facilitate positive effects of postlearning stress (Roozendaal, Okuda, Van der Zee, & McGaugh, 2006; see McGaugh, 2004, for a review). Thus, we hypothesized that postlearning stress could enhance context memory, although because the learned contexts were affectively neutral, it was also possible that these memories would not be affected.

The literature on stress and striatal memory is sparser. All studies on postlearning stress have been conducted in rodents and report that stress enhances striatal memory (Goode, Leong, Goodman, Maren, & Packard, 2016; Leong, Goodman, & Packard, 2012; Packard & Gabriele, 2009; Quirarte et al., 2009; Wingard & Packard, 2008), leading us to predict that postlearning stress would enhance SR memory in humans as well.

Methods

Participants

Participants (n = 60) were randomly assigned to Stress (n = 30, 60% female) or No Stress (n = 30, 53.3% female) conditions. This sample size was set before the onset of data collection, based on previous research using this search task (Goldfarb, Shields, et al., 2017; Goldfarb et al., 2016) and stress manipulation (e.g., Raio, Brignoni-Perez, Goldman, & Phelps, 2014; Cahill et al., 2003). Participants were required to be over 18 years old, fluent in English, have normal or corrected-to-normal vision, have normal color vision, and not be pregnant. To reduce factors that could influence the stress response, participants were excluded if they were taking antidepressants, antianxiety medications, beta blockers, or corticosteroids or were over 35 years old. Female participants were included regardless of whether they were taking oral contraceptives (17.7%) or their menstrual phase. The Stress and No Stress groups did not differ in age or preexisting stress levels (Table 1, all ps > .25). Additional participants were excluded for not meeting performance criteria (see Multicued Search Task section below; high error rate: n = 5, significantly slower RT with either memory cue: n = 12), failing to complete the experiment (n = 4), taking antidepressants (n = 1), and experimenter error (n = 2). Two participants provided insufficient saliva to measure levels of cortisol (n = 2) or α-amylase (n = 1) at one time-point and are excluded from analyses with these measures. All procedures were approved by the New York University committee on activities involving human subjects.

Table 1. 

Experiment 1: Demographics and Overall Search Task Performance

 No Stress Stress Diff? 
Demographics 
Sex (% Female) 53.3 60 p > .25 
Age 22.53 [4.86] 21.97 [3.77] p > .25 
PSS 18.23 [6.22] 17.9 [6.63] p > .25 
 
Search Task: Day 1 
Accuracy 97.58 [1.74] 96.98 [1.7] p = .19 
RT (sec) 1.33 [.25] 1.42 [.22] p = .14 
 
Search Task: Day 2 
Accuracy 97.93 [1.5] 98.4 [1.39] p = .21 
RT (sec) 1.11 [.19] 1.14 [.18] p > .25 
 No Stress Stress Diff? 
Demographics 
Sex (% Female) 53.3 60 p > .25 
Age 22.53 [4.86] 21.97 [3.77] p > .25 
PSS 18.23 [6.22] 17.9 [6.63] p > .25 
 
Search Task: Day 1 
Accuracy 97.58 [1.74] 96.98 [1.7] p = .19 
RT (sec) 1.33 [.25] 1.42 [.22] p = .14 
 
Search Task: Day 2 
Accuracy 97.93 [1.5] 98.4 [1.39] p = .21 
RT (sec) 1.11 [.19] 1.14 [.18] p > .25 

There were no differences between groups in sex, age, or preexisting stress levels (PSS = Perceived Stress Scale; Cohen, Kamarck, & Mermelstein, 1983). The groups did not differ in their overall ability to perform the search task. Accuracy in terms of % correct (out of 576 trials on Day 1, 192 trials on Day 2). Values reported as mean [standard deviation].

Multicued Search Task

The search task has been described in detail elsewhere (Goldfarb et al., 2016; Goldfarb, Shields, et al., 2017). Participants searched for a target (rotated “T”) embedded among distractors (rotated “L”s), pressed a button based on the direction the “T” was pointing, and received feedback regarding their speed and accuracy. Unknown to participants, some trials included repeated spatial contexts (hereafter contextually cued, or CC trials; Figure 1), in which the layout of distractors provided a cue to the “T” location. Other trials included a probabilistic SR association, in which the color of the stimuli provided a cue (80% validity) to the approximate “T” location and the button-press response (Figure 1). Faster RT on trials with CC and SR cues compared with trials with no memory cue has been shown to involve the hippocampus and striatum, respectively (Goldfarb et al., 2016).

Figure 1. 

Example memory cues. On contextually cued trials (top), the layout of the stimuli and the appearance of the distractors provided a cue (100% validity) to the location of the target. On SR trials (bottom), the color of the stimuli probabilistically (80% validity) cued the location (quadrant) of the target and the button press response.

Figure 1. 

Example memory cues. On contextually cued trials (top), the layout of the stimuli and the appearance of the distractors provided a cue (100% validity) to the location of the target. On SR trials (bottom), the color of the stimuli probabilistically (80% validity) cued the location (quadrant) of the target and the button press response.

Participants completed 24 blocks of the visual search task (24 trials per block: eight unique CC contexts, each repeated 1× per block; five SR cues, 4/5 valid; 11 No Cue) on Day 1. For analyses, these are combined into epochs (four consecutive blocks per epoch). Only accurate, nonoutlier (RT <3SD outside the mean per cue per day) responses were included in analyses. We quantified learning and memory for CC and SR associations by comparing RT on trials with memory cues to RT on trials with no cues. To measure the maximum learning achieved on Day 1, we compared RT during the last two epochs or eight consecutive blocks (Goldfarb, Shields, et al., 2017).

Because we measure learning for each cue separately, we can test whether the associations have been learned and account for learning differences between groups. Previous research has shown that individuals vary in their ability to learn CC and SR associations, and we wanted to limit the influence of these baseline differences (Goldfarb, Shields, et al., 2017). Furthermore, as the goal of these experiments was to test the influence of stress on memory, it was critical that memories had been formed. Previous studies have addressed this by excluding participants who do not reach a learning criterion (Atsak et al., 2016; Guenzel et al., 2013; Roozendaal et al., 2006). Accordingly, we set two criteria for participant performance on Day 1, one for general performance and one for learning. For general performance, participants needed to have an error rate lower than 10% on No Cue, CC, and SR trials (for comparison, the average error rate in a previous experiment was 2.5%; Goldfarb et al., 2016). A high error rate would indicate difficulty performing the search task, making it difficult to assess memory-guided performance. For learning, memory cues could not impair visual search performance. As memory is quantified based on facilitated search with CC and SR cues relative to no cues, it would provide strong evidence that mnemonic associations were not formed if CC or SR cues impaired search. Specifically, participants could not have significantly slower RT on SR or CC trials relative to No Cue trials (last eight blocks of Day 1). These criteria ensured that participants had experience performing memory-cued responses before any stress intervention.

On Day 2, participants returned to complete eight additional blocks of visual search followed by tests probing explicit memory for the cued associations. Because both CC and SR trials provided a cue to the location of the “T,” we tested whether participants could explicitly report the quadrant of the screen in which the “T” had appeared on these trials. As the SR trials also cued the appropriate button press, we further tested whether participants could report the button press made on trials with SR cues. Finally, we tested which memory would be preferentially expressed using a probe test. In this phase, the CC and SR cues were combined (i.e., the repeated spatial context was shown in the SR color) and two “T”s were presented. On the basis of which “T” they chose, we could determine whether participants were preferentially expressing memory for the SR or context cue.

The specific features of the search task (e.g., the configuration of repeated context layouts and the cued responses on SR trials) were randomly generated for pairs of participants. That is, for each experiment, pairs of Stress (n = 27) and No Stress participants (n = 27) completed the same search task (same trials, same trial sequence, etc.). Because of experimenter error, six participants (Stress = 3, No Stress = 3) were not paired. This design aimed to restrict variability between Stress and No Stress conditions to exposure to stressor or the control condition.

Stress Manipulation

The cold pressor task (CPT) is a validated laboratory stressor that leads to increases in stress hormones (Lovallo, 1975). During the task, participants submerged their nondominant arm in a bucket of ice water for three continuous minutes (mean temperature = 2.2°C, SD = 0.96°C). In the control condition, participants underwent the same procedure, but with warm water (mean temperature = 38.07°C, SD = 1.09°C). Immediately afterward, participants rated the unpleasantness of the experience (0 = not at all unpleasant; 10 = extremely unpleasant).

The CPT elicits a stress response, resulting in activation of the sympathetic nervous system and the immediate release of catecholamines (e.g., adrenaline). It also triggers the hypothalamic-pituitary-adrenal axis, leading to the eventual release of glucocorticoids or cortisol in humans (Lupien, Maheu, Tu, Fiocco, & Schramek, 2007). Cortisol can cross the blood–brain barrier and directly influence brain structure and function, especially within the hippocampus (McEwen & Sapolsky, 1995). Importantly, glucocorticoids have specifically been implicated in the stress-induced shift to preferential engagement of striatal over hippocampal memory (Schwabe et al., 2010a, 2013). On the basis of these results, we waited to measure the stress response until 10 min after the stressor, when cortisol levels would be elevated.

Adrenaline does not cross the blood–brain barrier as readily but plays an important role in memory consolidation. Learning emotionally arousing information can evoke an adrenergic response, which then acts (indirectly) on the amygdala to enhance memory for that information (Roozendaal & McGaugh, 2011; Lupien et al., 2007; McGaugh, 2004). This adrenergic response during learning can also interact with stress-induced cortisol responses to modulate memory (e.g., Segal et al., 2014; Bryant, McGrath, & Felmingham, 2013).

We measured both of these neuroendocrine responses peripherally by taking saliva samples throughout the experiment (Figure 2A). After collection, salivary samples were stored at −20°C in sterile tubes. Samples were then shipped frozen to the laboratory of Dr. Andrea Gierens at the University of Trier for analysis of cortisol and α-amylase. We have previously shown that CPT leads to an increase in salivary cortisol levels, using a protocol similar to the one described here (Goldfarb, Frobose, Cools, & Phelps, 2017). α-Amylase provides a measure of sympathetic nervous system activity and has been specifically linked to noradrenaline (Thoma, Kirschbaum, Wolf, & Rohleder, 2012; van Stegeren, Wolf, & Kindt, 2008). As cortisol and α-amylase values were not normally distributed, these values were log-transformed before all analyses.

Figure 2. 

Experiment 1: Procedure and cortisol levels. (A) Timelines for the two sessions of Experiment 1 (Postlearning Stress). CTL = control task. t1–t5 indicate timing of saliva samples (analyzed for salivary cortisol and α-amylase). Numbers in italics indicate average duration across participants. (B) Cortisol levels. The CPT (between t2 and t3) led to a significant increase in cortisol, although these levels were no longer different before the retrieval tests (t4). Sample timing as described in A. Log-transformed cortisol values (original units: nmol/L) are shown.

Figure 2. 

Experiment 1: Procedure and cortisol levels. (A) Timelines for the two sessions of Experiment 1 (Postlearning Stress). CTL = control task. t1–t5 indicate timing of saliva samples (analyzed for salivary cortisol and α-amylase). Numbers in italics indicate average duration across participants. (B) Cortisol levels. The CPT (between t2 and t3) led to a significant increase in cortisol, although these levels were no longer different before the retrieval tests (t4). Sample timing as described in A. Log-transformed cortisol values (original units: nmol/L) are shown.

Procedure

The procedure is described in Figure 2A. We used a between-subject design, in which participants were randomly assigned to Stress (CPT) or No Stress (warm water) conditions. All participants came to the laboratory for two sessions, 24 hr apart. To control for circadian fluctuations in cortisol levels (Lupien et al., 2007), all sessions occurred between 12:00 and 6:00 pm.

On Day 1, all participants acclimated to the laboratory environment for 10 min and completed questionnaires. They then provided their first saliva sample and completed a practice block of the visual search task with the experimenter. After the practice, they completed 24 blocks of search alone and then provided a second saliva sample. Participants were next randomly assigned to complete the stress (CPT) or control (warm water) manipulation. They rested for 10 min and provided an additional saliva sample, completing Day 1.

On Day 2, participants again acclimated to the lab environment for 10 min before providing the first saliva sample and then began the visual search task. After the search task, participants completed tests for explicit memory and the probe test to check which form of memory would be preferentially expressed. All participants then provided a final saliva sample and were debriefed and compensated.

Results

Stress Response

The CPT led to significant subjective and objective stress responses. Participants in the stress group rated the water exposure (CPT) as significantly more unpleasant than participants in the No Stress group (Stress: M = 8, No Stress: M = 0.97, t(58) = 17.26, p < .001). The CPT also elicited a significant increase in salivary cortisol (Figure 2B). Focusing on time-points pre- and poststress (t2 to t3), we ran a repeated-measures ANOVA (rmANOVA) with Group as a between-subject factor and Time as a within-subject factor, finding a main effect of Group (F(1, 56) = 4.72, p = .034) and a Group × Time interaction (F(1, 56) = 41.36, p < .001). This was not due to baseline differences in cortisol, as the groups did not differ before the CPT (t(56) = 0.29, p > .25), although they did differ 10 min after the stressor (t(56) = 4.03, p < .001), driven by a significant cortisol increase in the Stress group (t(28) = 5.9, p < .001).

To ensure that the stress response specifically targeted consolidation, we ran further analyses to ensure that cortisol did not differ during learning or retrieval. There were no differences in cortisol levels on Day 1 pre- and postlearning (Group: F(1, 56) = 0.12, p > .25; Group × Time: F(1, 56) = 0.008, p > .25). Cortisol also did not differ on Day 2 (Group: F(1, 56) = 0.44, p > .25; Group × Time: F(1, 56) = 0.62, p > .25), indicating that stress did not interfere with retrieval.

Unlike cortisol, we did not observe significant differences between groups in α-amylase during any of these windows (pre- and postlearning: Group: F(1, 57) = 0.15, p > .25; Group × Time: F(1, 57) = 1.02, p > .25; pre- and poststress: Group: F(1, 57) = 0.27, p > .25; Group × Time: F(1, 57) = 0.87, p > .25; pre- and postretrieval: Group: F(1, 57) = 0.25, p > .25; Group × Time: F(1, 57) = 1.25, p > .25). The lack of CPT-induced change in α-amylase is likely due to the fact that saliva samples were taken 10 min after the stressor, at which point adrenergic responses have subsided (Plessow, Schade, Kirschbaum, & Fischer, 2012; Schoofs, Preuss, & Wolf, 2008). Finally, groups did not differ in adrenergic response during learning (Δα; t(57) = 1.01, p > .25), but there was a great deal of variability (range: −1.55 to 1.87). There were no differences in cortisol response to stress or adrenergic response during learning by sex (all ps > .25).

Learning Context and SR Associations, Day 1

Before any stress or control manipulations, participants did not differ in their ability to perform the search task (Table 1). To test whether the learning process differed between groups, we ran an rmANOVA with Group as the between-subject factor and Cue (No Cue, CC, SR) and Time (Epochs 1–6) as within-subject factors (Figure 3A). There was no main effect of Group (F(1, 58) = 2.02, p = .16), but there was a significant Group × Time interaction (F(1, 58) = 4.65, p = .035), showing that the groups differed in learning rate. We also found a significant main effect of Cue (F(3, 174) = 7.61, p < .001) and no Group × Cue interaction (F(3, 174) = 0.11, p > .25), indicating that, although the learning rate differed, both groups used mnemonic cues.

Figure 3. 

Experiment 1: CC and SR learning. (A) RTs for the different cue types throughout Day 1. Each epoch = average of 4 consecutive blocks. Error bars = +1 SE. (B) The percent difference between cued trials (CC: context cue, SR: stimulus-response) and trials with no memory cue at the end of Day 1 (highlighted by gray squares in A). Error bars = +1 SEM. ***p < .001, **p < .01, *p < .05.

Figure 3. 

Experiment 1: CC and SR learning. (A) RTs for the different cue types throughout Day 1. Each epoch = average of 4 consecutive blocks. Error bars = +1 SE. (B) The percent difference between cued trials (CC: context cue, SR: stimulus-response) and trials with no memory cue at the end of Day 1 (highlighted by gray squares in A). Error bars = +1 SEM. ***p < .001, **p < .01, *p < .05.

Importantly, by the end of Day 1, participants in both groups showed evidence of having learned CC and SR associations. Compared with trials with no cue, they were faster to respond on CC (No Stress: t(29) = 4.97, p < .001; Stress: t(29) = 4.83, p < .001) and SR trials (No Stress: t(29) = 2.1, p = .045; Stress: t(29) = 3.16, p = .004). To quantify the extent to which participants used these mnemonic associations, we computed the percent difference in RT between trials with memory cues (CC or SR) and trials with no cue at the end of learning (last two epochs or eight blocks of Day 1; Figure 3B). Despite the difference in learning rate, the groups did not differ in this final measure of learning (CCPercent Difft(58) = −0.11, p > .25, SRPercent Difft(58) = −0.78, p > .25). Female participants showed better context learning (t(58) = 2.76, p = .008), but this did not differ between groups (Group × Sex: F(1, 56) = 2.18, p = .15).

Retrieval of Context and SR Memories, Day 2

We hypothesized that postlearning stress would influence consolidation of context and SR memory. To test this, we examined the change between performance at the end of learning (last epoch, Day 1) and beginning of retrieval (first epoch, Day 2) and looked for differences between Stress and No Stress groups. We ran an rmANOVA with Group (Stress vs. No Stress) as the between-subject factor and Cue (CCPercent Diff vs. SRPercent Diff) and Day (Day 1 vs. Day 2) as within-subject factors. There were no effects of Group (all main effects and interactions p > .25), indicating that stress did not change consolidation overall. Throughout the two epochs of Day 2, both groups continued to show context and SR memory (No Cue vs. CC: No Stress: t(29) = 6.18, p < .001; Stress: t(29) = 5.32, p < .001; No Cue vs. SR: No Stress: t(29) = 2.79, p = .009; Stress: t(29) = 2.55, p = .016). Consistent with better learning, female participants showed better CC retention (t(58) = 2.48, p = .016), but, again, this did not differ between groups (Group × Sex: F(1, 56) = 0.04, p > .25).

Explicit Memory for Cued Associations

Despite showing significantly faster RT in the presence of CC and SR cues over the two days of the experiment, there was no evidence that participants in either group could explicitly recall the learned associations. They were no different from chance (25%) in their ability to identify the cued “T” location on CC (No Stress: 26.7%, t(29) = 0.58, p > .25; Stress: M = 22.1%, t(29) = −0.93, p > .25) or SR trials (No Stress: 20%, t(29) = −1.5, p = .14; Stress: 26%, t(29) = 0.26, p > .25). In addition, both groups showed below chance (50%) memory for the cued button press on SR trials (significantly below chance in the No Stress group; No Stress: 39.3%, t(29) = −3.15, p = .004; Stress: 42.7%, t(29) = −1.99, p = .056).

Preferential Expression of Context or SR Memory

To test whether postlearning stress led to preferential expression of SR memory, we examined performance in the first trial of the probe test, in which only one type of memory could be used. We focused on performance during the first trial in which CC or SR memory could be used (Goldfarb, Shields, et al., 2017). Overall, the groups did not differ in which cue they chose (SR cue – Stress: 63.3%, No Stress: 53.3%, p > .25).

Individual Variability in Postlearning Stress Effects

As mentioned earlier, we collected saliva samples to measure levels of cortisol and α-amylase throughout the study. Because previous work has shown that adrenergic activity during learning can interact with postlearning cortisol responses to change later declarative memory (Segal et al., 2014; Bryant et al., 2013), we tested whether individual variability in learning-related adrenergic responses would interact with postlearning stress to influence context memory. We computed the stress-induced cortisol response (ΔCortisol) as the difference in log-transformed cortisol from the samples pre- to poststressor (Figure 2, t3 − t2) and learning-related adrenergic responses (Δα) as the difference in log-transformed α-amylase from pre- to post-search task (t2 − t1).

Postlearning stress effects on context memory: Individual variability

We again tested whether postlearning stress led to changes in performance from Day 1 to Day 2 but focused on context memory (CCPercent Diff) and included individual variability in Δα (median split across all participants) to the model. We found a trend-level Group × Day × Δα interaction (Figure 4A; F(1, 56) = 3.76, p = .058). This was driven by differences in context memory on Day 2, as context memory differed between groups when including the adrenergic response during learning in the model (Group × Δα [hereafter continuous]: F(1, 56) = 5.85, p = .019). Participants exposed to stress after learning showed worse context memory on Day 2 if they had experienced higher adrenergic responses (Δα) during learning (r(28) = −.38, p = .036).

Figure 4. 

Influence of postlearning stress on multiple memory systems. (A) Acute stress after learning interacts with adrenergic responses during learning to impair context recall. Adrenergic response was computed as the change in levels of salivary α-amylase from pre- to postlearning on Day 1. Top, the percent difference in RT between trials with context cues and trials with no mnemonic cue during the last epoch of Day 1 and the first epoch of Day 2. Bottom, residual context-guided performance on Day 2 (as described above) after controlling for performance on Day 1. Bottom left, same median split as above; bottom right, full range of adrenergic responses plotted against residual search performance. Δα = change in α-amylase, pre- to postlearning. *p < .05, ∼p < .1. (B) Cortisol reactivity immediately after learning is associated with SR learning. Although learning occurred before the stress manipulation, participants in the Stress group who had higher cortisol responses to the CPT showed better SR learning (left). This relationship continued into SR memory 24 hr later (middle). Participants with higher cortisol responses were also biased toward using SR memory when only one form of memory could be used (right). Each dot = 1 participant. (C) Postlearning stress interacts with adrenergic response during learning to bias the expression of different memory systems. Among participants exposed to stress after learning, the majority of those who experienced a large adrenergic response during learning (as in A) used SR rather than context memory.

Figure 4. 

Influence of postlearning stress on multiple memory systems. (A) Acute stress after learning interacts with adrenergic responses during learning to impair context recall. Adrenergic response was computed as the change in levels of salivary α-amylase from pre- to postlearning on Day 1. Top, the percent difference in RT between trials with context cues and trials with no mnemonic cue during the last epoch of Day 1 and the first epoch of Day 2. Bottom, residual context-guided performance on Day 2 (as described above) after controlling for performance on Day 1. Bottom left, same median split as above; bottom right, full range of adrenergic responses plotted against residual search performance. Δα = change in α-amylase, pre- to postlearning. *p < .05, ∼p < .1. (B) Cortisol reactivity immediately after learning is associated with SR learning. Although learning occurred before the stress manipulation, participants in the Stress group who had higher cortisol responses to the CPT showed better SR learning (left). This relationship continued into SR memory 24 hr later (middle). Participants with higher cortisol responses were also biased toward using SR memory when only one form of memory could be used (right). Each dot = 1 participant. (C) Postlearning stress interacts with adrenergic response during learning to bias the expression of different memory systems. Among participants exposed to stress after learning, the majority of those who experienced a large adrenergic response during learning (as in A) used SR rather than context memory.

We ran follow-up analyses to determine if this interaction between postlearning stress and adrenergic response directly influenced context recall and was not driven by initial differences in context learning. To account for variability in initial learning (although learning did not differ between stress groups [F(1, 56) = 0.14, p > .25] and was not predicted by a Group × Δα interaction [F(1, 56) = 0.25, p > .25]), we modeled Day 2 context memory as a function of Day 1 context learning and compared the residuals. We still found a significant Group × Δα interaction predicting residual context memory (Figure 4A; F(1, 56) = 5.64, p = .021), indicating that the interaction between stress and adrenergic response was specific to context memory recall.

Although learning at the end of Day 1 did not differ between stress groups, we did find a trend-level Cue (No Cue vs. CC) × Δα interaction throughout Day 1 (F(1, 58) = 3.01, p = .088), suggesting that the rate of learning differed based on adrenergic response during learning. Participants with higher adrenergic responses during learning were slightly slower to learn context cues. To ensure that this learning rate variability did not account for the differences in the retrieval of context memory on Day 2, we ran another model with learning rate (slope of context learning on Day 1) as a covariate. In this revised model, which included Δα, Stress Group, and learning rate as predictors, the interaction of Group × Δα still significantly predicted context memory on Day 2 (F(1, 52) = 6.00, p = .018), demonstrating that the influence of stress and adrenergic response was not the result of different learning rates. Together, these analyses demonstrate that the adrenergic response during learning interacted with postlearning stress to impair the consolidation of context memory, leading to worse recall 24 hr later.

Next, we tested whether the observed interaction between stress and adrenergic response was specific by examining individual variability in other neuroendocrine measurements. We did not find that residual CC memory was predicted by Group (Stress vs. No Stress) and the change in α-amylase during retrieval (F(1, 56) = 0.53, p > .25), indicating that the adrenergic response during learning was uniquely important for postlearning stress effects. There was also no interaction between Group (Stress vs. No Stress) and the cortisol response to the stressor (F(1, 55) = 0.001, p > .25), demonstrating that the magnitude of the stress response did not drive these effects.

Finally, we examined whether expression of context memory changed over the course of Day 2. Unlike most memory tests, our task tests memory repeatedly by continuing to expose participants to memory cues (eight times each for CC trials). We use performance at the beginning of Day 2 (first epoch) as our index of memory retrieval to control for potential new learning effects. Thus far, we have shown that postlearning stress and learning-related adrenergic response interact to impair context memory at the beginning (first epoch) of Day 2, but it was possible that context memory would improve with repeated exposure. We found that stress after learning, combined with high adrenergic response during learning, blocked the expression of context memory throughout Day 2. Using Group and Δα as between-subject factors and Block (Day 2, all eight blocks) as a within-subject factor, we observed a trend-level Group × Δα interaction (F(1, 56) = 3.86, p = .054) and no Group × Δα × Block interaction (F(1, 56) = 1.87, p = .18), suggesting that the impairment in context memory did not resolve with repeated exposure. It is worth noting that, as discussed above, participants with high adrenergic responses during learning were also slower to learn context associations and did not show evidence of context memory during the first eight blocks of Day 1 (median split, high Δα: main effect Cue [No Cue vs. CC], F(1, 29) = 0.39, p > .25). Thus, although there was an opportunity to “relearn” context on Day 2, eight blocks may not have been sufficient for these participants.

Postlearning stress effects on SR memory: Individual variability

We first tested whether variability in adrenergic responses during learning also interacted with postlearning stress to influence SR memory. Unlike context memory, there was no evidence that adrenergic responses during learning predicted stress-induced changes in SR memory (Group × Day × Δα: F(1, 56) = 0.19, p > .25). Instead, we found a significant interaction with the cortisol response to the stressor (Group × Day × ΔCortisol: F(1, 55) = 4.55, p = .038), which we initially interpreted as a stress-induced change in consolidation of SR memory. However, despite the fact that the stressor occurred after learning, this interaction was driven by an association between cortisol reactivity and Day 1 learning.

We found that individual differences in cortisol reactivity were associated with differences in initial SR learning. We tested whether SR learning at the end of Day 1 (SRPercent Diff, last two epochs) was predicted by Group and ΔCortisol and found a significant main effect of ΔCortisol (F(1, 55) = 8.12, p = .006). There was no significant Group × ΔCortisol interaction (F(1, 55) = 2.04, p = .16), but the relationship was stronger in the Stress group (Figure 4B, ΔCortisol × SR learning: Stress – r(28) = .5, p = .005; No Stress – r(27) = .01, p > .25). Although there was a trend-level relationship between basal cortisol levels (at the beginning of Day 1) and SR learning (r(57) = −.24, p = .07), the relationship between ΔCortisol and SR learning persisted even when including both ΔCortisol and basal cortisol (and their interaction) in the model (F(1, 54) = 5.91, p = .018). In fact, the relationship between basal cortisol and SR learning appeared to be driven by a relationship between basal cortisol and performance on No Cue trials (r(57) = −.35, p = .007), such that higher basal cortisol was associated with faster RT.

Because the SR cue is probabilistic, another way to quantify SR learning is by taking the difference in RT between trials with valid SR cues (80%) and invalid cues (20%). If participants have learned that the SR cue is associated with a particular response (valid cues), they should be slower to respond when this expectation is violated (invalid cues). Again, there was a significant main effect of ΔCortisol predicting this metric of SR learning (F(1, 55) = 6.91, p = .01) driven by the Stress group (r(28) = .44, p = .016; No Stress: r(27) = .06, p > .25). There was no relationship between basal cortisol and this index of SR learning (r(57) = −.13, p > .25). Thus, despite the fact that both measures of SR recall on Day 2 were predicted by ΔCortisol (SRPercent Diff, Figure 4B: F(1, 55) = 5.06, p = .029; Stress: r(28) = .34, p = .06; No Stress: r(27) = .17, p > .25; SR Valid vs. Invalid: F(1, 55) = 6.65, p = .013; Stress: r(28) = .396, p = .03; No Stress: r(27) = .17, p > .25), this was not due to stress-induced changes in consolidation. Rather, higher cortisol reactivity after learning was associated with greater initial SR learning, and those participants who learned better initially also showed better recall on Day 2.

Explicit memory: Individual variability

Although we found differences in implicit memory when considering variability in neuroendocrine responses, explicit memory was not affected. Including the adrenergic response had no effect on explicit memory for cued locations (Group × Δα, CC trials: F(1, 56) = 0.38, p > .25; SR trials: F(1, 56) = 0.57, p > .25) or cued button press (SR trials: F(1, 56) = 0.13, p > .25). There were also no effects of cortisol reactivity on explicit memory (Group × ΔCortisol, CC location: F(1, 55) = 0.78, p > .25; SR location: F(1, 55) = 1.9, p = .18; SR button: F(1, 55) = 0.03, p > .25).

Preferential expression of context or SR memory: Individual variability

The above analyses demonstrated that postlearning stress interacted with individual variability in adrenergic activity during learning to impair consolidation of context memory. We tested whether these two factors would then lead to preferential expression of SR memory during the probe test. Looking specifically at participants with high adrenergic responses during learning (defined by median split of Δα-amylase across all participants) who were then exposed to postlearning stress, we found that the majority used SR memory on the first probe test trial (Figure 4C, 75% of participants; difference from chance: Χ2(df = 1) = 4, p = .046). This bias was stronger in male participants (binomial general linear model [GLM]: choice of SR predicted by Group × Δα × Sex: β = 7.31 [2.91], p = .012). These results show that postlearning stress, when accounting for variability in adrenergic response during learning, can cause expression of SR over context memory.

To test whether this bias was related to stress-induced changes in context memory, we ran a binomial GLM and found that use of the SR cue was predicted by the interaction of Group, Δα, and residual context memory (β = −25.13 [12.76], p = .049). When we included residual SR instead of context memory in the model, this interaction was not significant (β = −2.63 [5.24], p > .25), providing evidence that stress, in tandem with the adrenergic response during learning, drove the preferential expression of SR memory by modulating context memory.

We also showed that individual variability in cortisol responses was associated with differences in initial SR learning, raising the question of whether this variability would also predict preferential expression of SR memory. Indeed, use of the SR cue on the first trial was also predicted by higher cortisol reactivity (GLM, binomial: choice of SR predicted by ΔCortisol, β = 3.89 [1.96], p = .048; ΔCortisol × Group: β = −3.6 [2.15], p = .09). This pattern persisted throughout the probe test, with ΔCortisol predicting the proportion of trials on which participants used the SR cue (F(1, 55) = 8.17, p = .006), driven by a correlation between ΔCortisol and bias toward SR memory in the Stress group (r(28) = .44, p = .016; No Stress: r(27) = .13, p > .25).

EXPERIMENT 2: PRERETRIEVAL STRESS

In Experiment 1, we found that stress after learning interacted with adrenergic responses during learning to disrupt consolidation of context memory. This impaired context memory led participants to preferentially use SR memory when only one form of memory could be expressed. We also found that individual differences in cortisol reactivity immediately after learning were associated with variability in initial SR learning and later preferential use of SR memory. In Experiment 2, we investigated whether stress administered immediately before retrieval, but after intact consolidation, would also alter expression of these memory systems.

Preretrieval stress has frequently been shown to impair hippocampal memory (human: Smeets, 2011; Tollenaar, Elzinga, Spinhoven, & Everaerd, 2009; Buchanan & Tranel, 2008; Smeets et al., 2008; de Quervain, Roozendaal, Nitsch, McGaugh, & Hock, 2000; rodent: Dorey, Pierard, Chauveau, David, & Beracochea, 2012; Park, Zoladz, Conrad, Fleshner, & Diamond, 2008; Diamond et al., 2006; de Quervain, Roozendaal, & McGaugh, 1998; see Gagnon & Wagner, 2016, for a recent review), leading us to hypothesize that it would also impair context memory in our search task. As mentioned earlier, relatively few studies have examined stress and striatal memory. Recently, two articles reported that preretrieval stress impaired retrieval of SR memory in rodents and humans (Atsak et al., 2016; Guenzel et al., 2013). These findings are surprising, given that stress at this time-point also resulted in the preferential expression of SR over context memory (Elliott & Packard, 2008). Thus, we hypothesized that preretrieval stress would impair SR memory, but to a lesser extent than context memory, thus enabling SR memory to be preferentially expressed.

Methods

Participants

Participants (n = 60) were randomly assigned to Stress (n = 30, 50% female) or No Stress (n = 30, 50% female) conditions. The exclusion criteria were the same as in Experiment 1. The groups did not differ in age or preexisting stress levels (Table 2, all ps > .25). Additional participants were excluded for not meeting performance criteria (high error rate: n = 5, significantly slower RT with either memory cue than no cue: n = 10), failing to complete the experiment (n = 6) and experimenter error (n = 1). One participant provided insufficient saliva at one time-point to measure α-amylase and is excluded from analyses with this measure. All procedures were approved by the New York University committee on activities involving human subjects.

Table 2. 

Experiment 2: Demographics and Overall Search Task Performance

 No Stress Stress Diff? 
Demographics 
Sex (% Female) 50 50 p > .25 
Age 21.97 [3.47] 23 [4] p > .25 
PSS 18.17 [6.09] 16.7 [7.18] p > .25 
 
Search Task: Day 1 
Accuracy 96.98 [1.63] 97.33 [1.85] p > .25 
RT (sec) 1.36 [.28] 1.32 [.26] p > .25 
 
Search Task: Day 2 
Accuracy 97.76 [1.52] 97.78 [2.16] p > .25 
RT (sec) 1.15 [.23] 1.04 [.16] p = .04 
 No Stress Stress Diff? 
Demographics 
Sex (% Female) 50 50 p > .25 
Age 21.97 [3.47] 23 [4] p > .25 
PSS 18.17 [6.09] 16.7 [7.18] p > .25 
 
Search Task: Day 1 
Accuracy 96.98 [1.63] 97.33 [1.85] p > .25 
RT (sec) 1.36 [.28] 1.32 [.26] p > .25 
 
Search Task: Day 2 
Accuracy 97.76 [1.52] 97.78 [2.16] p > .25 
RT (sec) 1.15 [.23] 1.04 [.16] p = .04 

The groups did not differ in sex, age, or preexisting stress levels, or their overall ability to perform the search task, on Day 1. The group exposed to preretrieval stress became significantly faster than the No Stress group on Day 2. Accuracy reported as in Table 1.

Procedure

Participants completed the same multicued search task as in Experiment 1. Again, we used a “paired” design, where sets of Stress (n = 30) and No Stress (n = 30) completed the same task (same trials, same trial sequence, etc.) to limit potential variability between Stress and No Stress conditions.

The procedure was the same as Experiment 1, except the timing of the stress manipulation (Figure 5A). On Day 1, participants completed the visual search task and provided saliva samples before and after learning. On Day 2, after acclimating to the laboratory environment for 10 min, they provided a baseline saliva sample and were then randomly assigned to complete the CPT (mean temperature = 1.29°C, SD = 0.64°C) or warm water manipulation (mean temperature = 38.06°C, SD = 0.94°C). They then rested for 10 min, allowing cortisol levels to increase, and provided a saliva sample before continuing with the search task and memory tests as in Experiment 1.

Figure 5. 

Experiment 2: Procedure and cortisol levels. (A) Timelines for the two sessions of Experiment 2 (Preretrieval Stress). CTL = control task. t1–t5 indicate timing of saliva samples (analyzed for cortisol and α-amylase). Numbers in italics indicate average duration across participants. (B) Cortisol levels. The CPT (occurring between t3 and t4) led to a significant increase in cortisol, which remained elevated at the end of the memory tests (t5). Sample timing as described in A. Log-transformed cortisol values (original units: nmol/L) are shown.

Figure 5. 

Experiment 2: Procedure and cortisol levels. (A) Timelines for the two sessions of Experiment 2 (Preretrieval Stress). CTL = control task. t1–t5 indicate timing of saliva samples (analyzed for cortisol and α-amylase). Numbers in italics indicate average duration across participants. (B) Cortisol levels. The CPT (occurring between t3 and t4) led to a significant increase in cortisol, which remained elevated at the end of the memory tests (t5). Sample timing as described in A. Log-transformed cortisol values (original units: nmol/L) are shown.

Results

Stress Response

As in Experiment 1, the CPT was effective in inducing a stress response. Participants in the Stress group rated the CPT as significantly more unpleasant than participants in the No Stress group rated the warm water (Stress: 7.62, No Stress: 1.15; t(58) = 17.68, p < .001). The preretrieval stress manipulation also led to a significant increase in salivary cortisol (Figure 5B; pre- to poststress: Group: F(1, 58) = 6.72, p = .012; Group × Time: F(1, 58) = 20.17, p < .001). The groups did not differ at the beginning of Day 2 (t3: t(58) = 0.91, p > .25) but were significantly different at the start of retrieval (t4: t(58) = 3.91, p < .001). Cortisol levels remained elevated in the Stress group throughout the retrieval tests (t5, Stress vs. No Stress: t(58)= 3.72, p < .001). The stress-induced cortisol response (ΔCortisol) did not differ between the Stress groups in Experiment 1 and Experiment 2 (t(57) = 0.78, p > .25).

Further analyses revealed that the stress-induced cortisol response specifically targeted retrieval processes, and not learning or consolidation, as there were no differences in cortisol on Day 1 (Group: F(1, 58) = 0.1, p > .25; Group × Time: F(1, 58) = 0.22, p > .25).

As in Experiment 1, we did not observe significant differences between groups in α-amylase during any of the above time windows (pre- and postlearning: Group: F(1, 57) = 0.13, p > .25; Group × Time: F(1, 57) = 0.72, p > .25; pre- and poststress: Group: F(1, 57) = 0.15, p > .25; Group × Time: F(1, 57) = 1.02, p > .25; pre- and postretrieval: Group: F(1, 57) = 1.02, p > .25; Group × Time: F(1, 57) = 1.66, p = .2). We also did not observe any differences between groups in the adrenergic response during learning (Δα: t(57) = 0.85, p > .25), and there was a great deal of variability (range: −1.91 to 1.89). There were no differences in the adrenergic response during learning between Experiments 1 and 2 (t(116) = 1.03, p > .25). Cortisol response to stress and adrenergic response during learning did not differ by sex (all ps > .25).

Learning Context and SR Associations

As in Experiment 1, we ran an rmANOVA with Group as the between-subject factor and Cue (No Cue, CC, SR) and Time (Epochs 1–6) as within-subject factors to examine the learning process on Day 1. There were no significant interactions with or main effects of Group (Figure 6A; all ps > .25), and there was a significant main effect of Cue (F(3, 174) = 14.71, p < .001), indicating that the groups learned similarly and used the embedded mnemonic cues. By the end of Day 1, both groups were significantly faster to respond on CC (No Stress: t(29) = 2.91, p = .007; Stress: t(29) = 4.64, p < .001) and SR trials (No Stress: t(29) = 2.8, p = .009; Stress: t(29) = 3.37, p = .002) than trials with no cue. There were no differences between groups in the extent to which they used these cues (Figure 6B; CCPercent Difft(58) = 0.11, p > .25; SRPercent Difft(58) = 0.17, p > .25). Unlike Experiment 1, there were no differences between sexes in CC learning (t(58) = 0.53, p > .25) or in any of the stress effects reported below (all ps > .18).

Figure 6. 

Experiment 2: CC and SR learning. (A) RTs for the different cue types throughout Day 1. Error bars = +1 SE. (B) The percent difference between cued trials (CC, SR) and trials with no memory cue at the end of Day 1 (highlighted by gray squares in A). Error bars = +1 SEM. ***p < .001, **p < .01, *p < .05.

Figure 6. 

Experiment 2: CC and SR learning. (A) RTs for the different cue types throughout Day 1. Error bars = +1 SE. (B) The percent difference between cued trials (CC, SR) and trials with no memory cue at the end of Day 1 (highlighted by gray squares in A). Error bars = +1 SEM. ***p < .001, **p < .01, *p < .05.

Retrieval of Context and SR Memories, Day 2

To test whether stress influenced memory retrieval, we ran an rmANOVA with Group (Stress vs. No Stress) as the between-subject factor and Cue (CCPercent Diff vs. SRPercent Diff) and Day (last epoch of Day 1 vs. first epoch of Day 2) as within-subject factors. We found a significant Group × Cue × Day interaction (F(1, 58) = 5.81, p = .019), where the Stress group appeared to become worse at using CC associations but showed no change in use of SR associations from Day 1 to Day 2 (Figure 7A). However, the Stress group also became faster on No Cue trials from Day 1 to Day 2 (t(29) = 3.03, p = .005), more so than the No Stress group (Group × Day: F(1, 58) = 4.58, p = .037). It is possible that this general change in task performance was driving this apparent difference in memory. To account for this, we ran the ANOVA again, allowing for potential interactions with the change in No Cue RT, and still found a significant Group × Cue × Day effect (F(1, 56) = 5.67, p = .021), indicating that stress changed memory independent of changes in overall task performance. Below we discuss the effects of preretrieval stress on each memory system separately.

Figure 7. 

Influence of preretrieval stress on multiple memory systems. (A) Preretrieval stress impairs memory for contexts learned 24 hr earlier, but does not change SR memory. Top, the percent difference in RT between trials with memory cues (green = context, red = SR) and trials with no mnemonic cue during the last epoch of Day 1 and the first epoch of Day 2. Bottom, residual memory on Day 2 after controlling for learning on Day 1. (B) Although preretrieval stress impaired initial recall of context associations, these participants began to show context memory by the end of Day 2. *p < .05, ∼p < .1. The “*” symbol above a bar indicates that this bar is significantly different from zero.

Figure 7. 

Influence of preretrieval stress on multiple memory systems. (A) Preretrieval stress impairs memory for contexts learned 24 hr earlier, but does not change SR memory. Top, the percent difference in RT between trials with memory cues (green = context, red = SR) and trials with no mnemonic cue during the last epoch of Day 1 and the first epoch of Day 2. Bottom, residual memory on Day 2 after controlling for learning on Day 1. (B) Although preretrieval stress impaired initial recall of context associations, these participants began to show context memory by the end of Day 2. *p < .05, ∼p < .1. The “*” symbol above a bar indicates that this bar is significantly different from zero.

Preretrieval stress effects on context memory

Further analyses revealed that preretrieval stress significantly impaired context memory. Although the groups did not differ in context learning at the end of Day 1 (t(58) = 0.3, p > .25), the No Stress group showed significantly better context memory than the Stress group on Day 2 (t(57) = 2.46, p = .017). To account for differences in initial learning, we regressed out context learning on Day 1 from context memory on Day 2 and found that the No Stress residuals were significantly higher than those in the Stress group (Figure 7A, bottom: t(58) = 2.45, p = .017).

As mentioned earlier, the memory test on Day 2 involved repeated exposure to contexts. We thus tested whether the Stress group, although they were initially impaired at recalling context associations (during the first epoch or four blocks), would eventually be able to use context memory (Figure 7B). We ran an rmANOVA with Group as a between-subject factor and Block (during Day 2) as a within-subject factor and found a significant Group × Block effect (F(1, 58) = 5.24, p = .026). The No Stress group did not change throughout Day 2 (Main effect of Block: F(1, 29) = 1.18, p > .25), but the Stress group significantly improved (Block: F(1, 29) = 4.77, p = .037). The participants in the Stress group had also shown significant context memory during the first eight blocks of Day 1 (Main effect of Cue [No Cue vs. CC]: F(1, 29) = 4.19, p = .0497), suggesting that the eight blocks of Day 2 could be sufficient for them to “relearn.”

As an exploratory analysis, we tested whether the change in context memory over Day 2 was similar to initial learning on Day 1. We compared RT on context relative to No Cue trials (i.e., change in CCPercentDiff as a function of Block) over the eight blocks of Day 2 to the change in CCPercent Diff over the first eight blocks of Day 1. We observed a trend-level Day × Group interaction (F(1, 58) = 3.81, p = .056). The slope on Day 2 in the Stress group did not differ significantly from their initial slope on Day 1 (Day 1 vs. Day 2; t(29) = −0.69, p > .25), although these marginally differed in the No Stress group (t(29) = 1.98, p = .057). Thus, changes in context on Day 2 in the Stress group resembled “relearning,” rather than eventual recall of contexts learned on Day 1.

Preretrieval stress effects on SR memory

Participants in the Stress and No Stress groups did not differ in SR memory on Day 2 (t(58) = 1.13, p > .25) or in their residual SR memory on Day 2 after regressing out variability in Day 1 learning (t(58) = 1.5, p = .14). Within the Stress group, the impairment of context memory did not extend to SR memory. Participants exposed to preretrieval stress showed significantly better SR than context memory on Day 2 (t(29) = 2.14, p = .022), even when accounting for variability in learning (residuals: t(29) = 2.17, p = .039), although they did not differ in learning on Day 1 (t(29) = 1.37, p = .18).

Explicit Memory for Cued Associations

As in Experiment 1, there was no indication that either group had explicit memory for context or SR associations. They were no different from chance in memory for cued “T” locations (chance = 25%; CC: No Stress: M = 26.3% accurate, t(29) = 0.45, p > .25; Stress: 23.3%; t(29) = −0.56, p > .25; SR: No Stress: 25.3%, t(29) = 0.10, p > .25; Stress: 26%, t(29) = 0.3, p > .25) or cued button press responses (chance = 50%; No Stress: 45.3%, t(29) = −1.06, p = .3, Stress: 38%, t(29) = −3.19, p = .003).

Preferential Expression of Context or SR Memory

On the first trial of the probe test, where only one memory cue could be used, slightly more than half of the participants from the Stress group used the SR cue to find the target (53.3%). This proportion was not significantly different from the No Stress group (Used SR: 46.7%; p > .25), demonstrating that preretrieval stress did not bias participants to preferentially use SR memory. This is likely due to the eight blocks of context reexposure, which led to the eventual expression of context memory in the Stress group.

Individual Variability in Preretrieval Stress Effects

We tested whether individual variability in neuroendocrine responses interacted with preretrieval stress to influence memory. Unlike postlearning stress (Experiment 1), we did not find any interactions between Group (Stress vs. No Stress) and adrenergic response during learning (Δα) that influenced residual context memory (F(1, 55) = 0.001, p > .25), although the main effect of Group remained significant (F(1, 55) = 4.87, p = .032). There were also no interactions with adrenergic response during retrieval (F(1, 56) = 0.002, p > .25) or CPT-evoked cortisol responses (F(1, 56) = 0.02, p > .25), although the main effect of Group remained significant throughout (all ps < .05), indicating that the effects of preretrieval stress on context memory were not driven by variability in these neuroendocrine responses.

We also tested whether cortisol reactivity was associated with initial SR learning, as in Experiment 1, although the stressor in Experiment 2 occurred 24 hr after learning. There was no correlation between any metric of SR learning and CPT-induced cortisol response 24 hr later (all ps > .25), suggesting that this was specific to cortisol reactivity near the time of learning.

Explicit memory and preferential memory expression: Individual variability

We did not find effects of variability in these neuroendocrine measures on explicit memory for cued associations for context (Group × Δα: F(1, 55) = 1.8, p = .19, Group × ΔCortisol: F(1, 56) = 2.4, p = .13) or SR trials (Location – Group × Δα: F(1, 55) = 0.03, p > .25; button press – Group × Δα: F(1, 55) = 0.89, p > .25; Group × ΔCortisol: F(1, 56) = 1.14, p > .25). There was a marginal Group × ΔCortisol interaction predicting SR location memory (F(1, 56) = 2.98, p = .09), driven by a trend-level negative correlation between ΔCortisol and memory in the No Stress group (r(28) = −.33, p = .07). Finally, variability in cortisol and adrenergic responses were not associated with preferential expression of context or SR memory (all ps > .25).

DISCUSSION

In the experiments presented here, we investigated the influence of acute stress on the consolidation (Experiment 1) and retrieval (Experiment 2) of context and SR memory. Although the cortisol response to stress did not differ between studies, changing the timing of the stressor led to distinct effects on memory. We found that postlearning stress, combined with high adrenergic responses during learning, led to the preferential expression of SR over context memory. This was due to stress-induced impairments in retrieval of context memory. Preretrieval stress did not lead to a bias in which memory was preferentially expressed. Although stress at this time-point also impaired the expression of context memory, this effect was transient. Neither of these stressors changed SR memory. Finally, we found that individual differences in cortisol reactivity immediately after learning were associated with better SR learning.

Our finding that postlearning stress could bias memory expression is consistent with recent reports in nonhuman animals (Perez et al., 2017; Leong et al., 2015). In addition to extending this effect to humans, the current results provide insight into the cognitive mechanism underlying the preferential expression of SR memory. We found that this bias was related to a stress-induced change in context, rather than SR memory, and that the effect depended on the degree of adrenergic response (a proxy for arousal) experienced during learning.

The negative effect of postlearning stress (combined with the adrenergic response during learning) on context memory is in contrast to previous studies showing that postlearning stress selectively enhances memory for arousing items (e.g., Smeets et al., 2008; Cahill et al., 2003). However, in the experiments presented here, we assessed memory for associations rather than items. The vast majority of research in humans has focused on the influence of stress on item-level memory (e.g., Cadle & Zoladz, 2015; McCullough, Ritchey, Ranganath, & Yonelinas, 2015; McCullough & Yonelinas, 2013; Schwabe, Bohringer, Chatterjee, & Schachinger, 2008; Kuhlmann & Wolf, 2006; Payne et al., 2006; Abercrombie, Kalin, Thurow, Rosenkranz, & Davidson, 2003; Cahill et al., 2003; de Quervain et al., 2000, and others). Because item and context memory have different neural substrates (reviewed in Davachi, 2006), it is possible that stress would differentially impact these forms of memory. Furthermore, item and associative memory can be differentially affected by arousal (reviewed in Mather, 2007). This raises the possibility that high adrenergic responses, during and after learning, disrupted the binding and consolidation of these contextual associations. We also assessed memory for neutral, rather than arousing, information. Thus, despite the fact that some participants showed increased adrenergic activity during learning, this general adrenergic response would not necessarily “tag” the neutral contexts for preferential consolidation (e.g., Dunsmoor, Murty, Davachi, & Phelps, 2015).

Other research examining general (not item-linked) adrenergic responses have reported results that are more consistent with the present findings. These suggest a possible mechanism by which the adrenergic response during learning could, combined with postlearning stress, impair memory. An early study (Sandi, Loscertales, & Guaza, 1997) found that high levels of arousal during learning blocked the beneficial effects of postlearning stress. Rats who learned a water maze under moderately arousing conditions (water temperature = 25°C) showed enhanced memory with postlearning corticosterone, but there was no benefit for rats who learned under highly arousing conditions (19°C; associated with larger adrenergic response—see Mabry, Gold, & McCarty, 1995). In line with this “too much arousal” theory, studies that have reported negative effects of postlearning stress on hippocampal memory used pharmacological interventions to increase the adrenergic response (rather than the glucocorticoid response; Packard & Gabriele, 2009; Wingard & Packard, 2008; reviewed in Leong et al., 2012). Again, these results suggest that elevated adrenergic responses can have negative effects on memory (see Packard & Gabriele, 2009, for discussion). Thus, in our data, it is possible that high adrenergic responses during learning combined with an adrenergic response to the CPT to impair hippocampal memory. One limitation of our design is that the post-CPT saliva sample was taken 10 min after the stress manipulation, at which point α-amylase levels had already returned to baseline. In addition, our measures of salivary α-amylase reflect peripheral, rather than central, adrenergic activity. Although previous work has demonstrated peripheral mechanisms by which adrenergic responses modulate memory (Roozendaal & McGaugh, 2011), these measures limit our ability to understand central mechanisms. Thus, further work is needed to test the hypothesis that “too much arousal” impairs consolidation of context memory.

It is possible that postlearning stress did not enhance context memory because performance on contextually cued trials was at ceiling and could not be improved. However, previous studies (as well as specific features of the current design) suggest that this is not the case. For example, repeated context trials followed by reward were learned better, and more quickly, than repeated context trials that were not followed by an outcome (Tseng & Lleras, 2013). This demonstrates that posttrial interventions can enhance context memory within a learning session and raises the possibility that context memory could also be facilitated between sessions. However, although several studies have reported that context memory is retained over a delay (Zellin, von Muhlenen, Muller, & Conci, 2014; Jiang, Song, & Rigas, 2005; Chun & Jiang, 2003), few have directly tested postlearning manipulations (with the exception of rest; Mednick, Makovski, Cai, & Jiang, 2009) or compared the magnitude of contextual cueing at the end of learning to the beginning of retention. Several features of the current experiments suggest that context memory could be enhanced after learning. First, we reduced the amount of search on Day 1 from 30 blocks (as in Zellin et al., 2014; Jiang et al., 2005; Chun & Jiang, 2003) to 24 blocks to prevent learning reaching asymptote on Day 1. Second, we used complex displays to increase task difficulty, featuring 16 stimuli per screen and distractors (offset “L”-shapes) that closely resembled the target (differing from Zellin et al., 2014; Mednick et al., 2009; Jiang et al., 2005). This enabled us to avoid ceiling effects of search across the two sessions, which would preclude the possibility of memory-guided search benefits (discussed in Chun & Jiang, 2003). Finally, in the analysis, we compared the percent difference between memory-cued and no cue trials, enabling us to account for decreases in RT across days, which necessarily reduce absolute differences (Jiang et al., 2005). Using this approach, we did observe a marginal increase in contextual cueing between sessions in the No Stress group of Experiment 2 (Figure 7A). This suggests that it is possible to improve context memory after learning, although further research is needed to explore the circumstances under which these enhancements occur.

Unlike postlearning stress, we did not find that preretrieval stress led to a bias in which form of memory was preferentially expressed. This is in contrast to previous reports in nonhuman animals, which report a bias toward expression of SR memory (Hawley et al., 2013; Elliott & Packard, 2008). This lack of bias may be due to the fact that the effects of preretrieval stress on memory were transient. Although preretrieval stress impaired context memory initially, after repeated exposure to the same contexts on Day 2, participants were eventually able to use these memories to guide attention. Thus, by the time of the probe test, participants exposed to preretrieval stress no longer showed impaired expression of context memory. It is possible that these stress effects were transient because our learning procedure protected against the negative effects of stress. Recent work has shown that retrieval practice attenuates the deleterious effects of preretrieval stress (Smith, Floerke, & Thomas, 2016), and over the course of learning the search task, participants are continually retrieving context memories and using them to guide attention. However, it is not clear whether the stressed participants were eventually able to recall what they learned on Day 1, or if they were learning the context associations as if for the first time. Although there is a preliminary indication that participants were “relearning,” given the similarity between the slopes on the first eight blocks of Day 1 and the eight blocks of Day 2, our data cannot answer this question.

We did not find that acute stress, at the timings and dosage used in these experiments, led to significant changes in SR memory. This is not because SR memory in this task is not vulnerable to stress, as we have previously shown that lifetime stress exposure modulates expression of SR memory (Goldfarb, Shields, et al., 2017). Rather, these results dissociate the susceptibility of these types of memory to different types of stress. In Experiment 2, we found that participants exposed to stress before the memory test showed significantly better SR than context memory. The fact that moderate increases in cortisol led to changes in hippocampus-dependent memory, but not striatum-dependent memory, is consistent with work showing lower binding capacity for glucocorticoids in the striatum than the hippocampus (Defiore & Turner, 1983). As this research represents the first investigation of postlearning stress and SR memory in humans, further work is needed to investigate whether higher levels of stress would be able to modulate the expression of this form of memory.

Our finding that preretrieval stress did not impair SR memory is in contrast to recent reports (Guenzel et al., 2013). One reason may be the timing of the stressor. In the current study, participants began the memory test 12 min after the stress manipulation, whereas the study that found an impairing effect of stress in humans had a 25-min delay between the stressor and the task (Guenzel et al., 2013). Variability in timing of preretrieval stress is important for effects on recognition memory (Schwabe & Wolf, 2014), although future studies are needed to characterize the effects of stress timing on striatal memory. Another difference is that the previous study in humans used a navigation task to measure striatal memory. In such tasks, distinct regions of the pFC are involved in engaging hippocampal and striatal strategies (Dahmani & Bohbot, 2014). Because pFC is highly sensitive to stress (Arnsten, 2009), it is possible that stress-induced disruption of pFC function could impair navigation performance. As the measure of striatal memory used in the current study is implicit, the involvement of strategy selection (and potential pFC-mediated processes) is limited and thus unlikely to drive effects.

We also unexpectedly found an association between individual differences in cortisol reactivity to stress and expression of SR learning, although learning occurred before the stress manipulation. These results emphasize the importance of measuring learning when interpreting the effects of postlearning stress interventions. The fact that the correlation with cortisol reactivity was specific to stressors on the day of learning and did not occur for participants exposed to stress on the following day (despite comparable demographic features and cortisol responses) indicates that SR expression was associated with “state”-level cortisol reactivity at the time of learning, rather than “trait”-level variability. More research is needed to understand the mechanism by which such state-level propensities could influence learning.

In summary, these data provide insight into the cognitive mechanisms by which stress can modulate multiple memory systems. By accounting for variability in learning and separately measuring context and SR memories, we demonstrate that acute stress can time-dependently impair context memory, thus causing a bias toward preferential expression of SR memory.

Acknowledgments

This work was supported by NIH grant MH097085 (E. A. P.), a National Science Foundation Graduate Research Fellowship (E. V. G.), and a NYU Dean's Undergraduate Research Fund grant (Y. M.).

Reprint requests should be sent to Elizabeth A. Phelps, Department of Psychology, New York University, 6 Washington Place, 8th floor, New York, NY 10003, or via e-mail: liz.phelps@nyu.edu.

REFERENCES

REFERENCES
Abercrombie
,
H. C.
,
Kalin
,
N. H.
,
Thurow
,
M. E.
,
Rosenkranz
,
M. A.
, &
Davidson
,
R. J.
(
2003
).
Cortisol variation in humans affects memory for emotionally laden and neutral information
.
Behavioral Neuroscience
,
117
,
505
516
.
Arnsten
,
A. F.
(
2009
).
Stress signalling pathways that impair prefrontal cortex structure and function
.
Nature Reviews Neuroscience
,
10
,
410
422
.
Atsak
,
P.
,
Guenzel
,
F. M.
,
Kantar-Gok
,
D.
,
Zalachoras
,
I.
,
Yargicoglu
,
P.
,
Meijer
,
O. C.
, et al
(
2016
).
Glucocorticoids mediate stress-induced impairment of retrieval of stimulus–response memory
.
Psychoneuroendocrinology
,
67
,
207
215
.
Bryant
,
R. A.
,
McGrath
,
C.
, &
Felmingham
,
K. L.
(
2013
).
The roles of noradrenergic and glucocorticoid activation in the development of intrusive memories
.
PLoS One
,
8
,
e62675
.
Buchanan
,
T. W.
, &
Tranel
,
D.
(
2008
).
Stress and emotional memory retrieval: Effects of sex and cortisol response
.
Neurobiology of Learning and Memory
,
89
,
134
141
.
Cadle
,
C. E.
, &
Zoladz
,
P. R.
(
2015
).
Stress time-dependently influences the acquisition and retrieval of unrelated information by producing a memory of its own
.
Frontiers in Psychology
,
6
,
910
.
Cahill
,
L.
,
Gorski
,
L.
, &
Le
,
K.
(
2003
).
Enhanced human memory consolidation with postlearning stress: Interaction with the degree of arousal at encoding
.
Learning and Memory
,
10
,
270
274
.
Chun
,
M. M.
, &
Jiang
,
Y.
(
2003
).
Implicit, long-term spatial contextual memory
.
Journal of Experimental Psychology: Learning, Memory, and Cognition
,
29
,
224
234
.
Chun
,
M. M.
, &
Phelps
,
E. A.
(
1999
).
Memory deficits for implicit contextual information in amnesic subjects with hippocampal damage
.
Nature Neuroscience
,
2
,
844
847
.
Cohen
,
S.
,
Kamarck
,
T.
, &
Mermelstein
,
R.
(
1983
).
A global measure of perceived stress
.
Journal of Health and Social Behavior
,
24
,
385
396
.
Dahmani
,
L.
, &
Bohbot
,
V. D.
(
2014
).
Dissociable contributions of the prefrontal cortex to hippocampus- and caudate nucleus-dependent virtual navigation strategies
.
Neurobiology of Learning and Memory
,
117
,
42
50
.
Davachi
,
L.
(
2006
).
Item, context and relational episodic encoding in humans
.
Current Opinion in Neurobiology
,
16
,
693
700
.
de Quervain
,
D. J.
,
Roozendaal
,
B.
, &
McGaugh
,
J. L.
(
1998
).
Stress and glucocorticoids impair retrieval of long-term spatial memory
.
Nature
,
394
,
787
790
.
de Quervain
,
D. J.
,
Roozendaal
,
B.
,
Nitsch
,
R. M.
,
McGaugh
,
J. L.
, &
Hock
,
C.
(
2000
).
Acute cortisone administration impairs retrieval of long-term declarative memory in humans
.
Nature Neuroscience
,
3
,
313
314
.
Defiore
,
C. H.
, &
Turner
,
B. B.
(
1983
).
[3H]Corticosterone binding in the caudate-putamen
.
Brain Research
,
278
,
93
101
.
Diamond
,
D. M.
,
Campbell
,
A. M.
,
Park
,
C. R.
,
Woodson
,
J. C.
,
Conrad
,
C. D.
,
Bachstetter
,
A. D.
, et al
(
2006
).
Influence of predator stress on the consolidation versus retrieval of long-term spatial memory and hippocampal spinogenesis
.
Hippocampus
,
16
,
571
576
.
Dorey
,
R.
,
Pierard
,
C.
,
Chauveau
,
F.
,
David
,
V.
, &
Beracochea
,
D.
(
2012
).
Stress-induced memory retrieval impairments: Different time-course involvement of corticosterone and glucocorticoid receptors in dorsal and ventral hippocampus
.
Neuropsychopharmacology
,
37
,
2870
2880
.
Dunsmoor
,
J. E.
,
Murty
,
V. P.
,
Davachi
,
L.
, &
Phelps
,
E. A.
(
2015
).
Emotional learning selectively and retroactively strengthens memories for related events
.
Nature
,
520
,
345
348
.
Elliott
,
A. E.
, &
Packard
,
M. G.
(
2008
).
Intra-amygdala anxiogenic drug infusion prior to retrieval biases rats towards the use of habit memory
.
Neurobiology of Learning and Memory
,
90
,
616
623
.
Foerde
,
K.
,
Knowlton
,
B. J.
, &
Poldrack
,
R. A.
(
2006
).
Modulation of competing memory systems by distraction
.
Proceedings of the National Academy of Sciences, U.S.A.
,
103
,
11778
11783
.
Gagnon
,
S. A.
, &
Wagner
,
A. D.
(
2016
).
Acute stress and episodic memory retrieval: Neurobiological mechanisms and behavioral consequences
.
Annals of the New York Academy of Sciences
,
369
,
55
75
.
Goldfarb
,
E. V.
,
Chun
,
M. M.
, &
Phelps
,
E. A.
(
2016
).
Memory-guided attention: Independent contributions of the hippocampus and striatum
.
Neuron
,
89
,
317
324
.
Goldfarb
,
E. V.
,
Frobose
,
M. I.
,
Cools
,
R.
, &
Phelps
,
E. A.
(
2017
).
Stress and cognitive flexibility: Cortisol increases are associated with enhanced updating but impaired switching
.
Journal of Cognitive Neuroscience
,
29
,
14
24
.
Goldfarb
,
E. V.
, &
Phelps
,
E. A.
(
2017
).
Stress and the trade-off between hippocampal and striatal memory
.
Current Opinion in Behavioral Sciences
,
14
,
47
53
.
Goldfarb
,
E. V.
,
Shields
,
G. S.
,
Daw
,
N. D.
,
Slavich
,
G. M.
, &
Phelps
,
E. A.
(
2017
).
Low lifetime stress exposure is associated with reduced stimulus–response memory
.
Learning and Memory
,
24
,
162
168
.
Goode
,
T. D.
,
Leong
,
K.-C.
,
Goodman
,
J.
,
Maren
,
S.
, &
Packard
,
M. G.
(
2016
).
Enhancement of striatum-dependent memory by conditioned fear is mediated by beta-adrenergic receptors in the basolateral amygdala
.
Neurobiology of Stress
,
3
,
74
82
.
Guenzel
,
F. M.
,
Wolf
,
O. T.
, &
Schwabe
,
L.
(
2013
).
Stress disrupts response memory retrieval
.
Psychoneuroendocrinology
,
38
,
1460
1465
.
Guenzel
,
F. M.
,
Wolf
,
O. T.
, &
Schwabe
,
L.
(
2014
).
Glucocorticoids boost stimulus–response memory formation in humans
.
Psychoneuroendocrinology
,
45
,
21
30
.
Hawley
,
W. R.
,
Grissom
,
E. M.
,
Patel
,
J. M.
,
Hodges
,
K. S.
, &
Dohanich
,
G. P.
(
2013
).
Reactivation of an aversive memory modulates learning strategy preference in male rats
.
Stress
,
16
,
73
86
.
Jiang
,
Y.
,
Song
,
J. H.
, &
Rigas
,
A.
(
2005
).
High-capacity spatial contextual memory
.
Psychonomic Bulletin & Review
,
12
,
524
529
.
Kim
,
J. J.
,
Lee
,
H. J.
,
Han
,
J. S.
, &
Packard
,
M. G.
(
2001
).
Amygdala is critical for stress-induced modulation of hippocampal long-term potentiation and learning
.
Journal of Neuroscience
,
21
,
5222
5228
.
Kuhlmann
,
S.
, &
Wolf
,
O. T.
(
2006
).
Arousal and cortisol interact in modulating memory consolidation in healthy young men
.
Behavioral Neuroscience
,
120
,
217
223
.
Leong
,
K. C.
,
Goodman
,
J.
, &
Packard
,
M. G.
(
2012
).
Buspirone blocks the enhancing effect of the anxiogenic drug RS 79948-197 on consolidation of habit memory
.
Behavioural Brain Research
,
234
,
299
302
.
Leong
,
K. C.
,
Goodman
,
J.
, &
Packard
,
M. G.
(
2015
).
Post-training re-exposure to fear conditioned stimuli enhances memory consolidation and biases rats toward the use of dorsolateral striatum-dependent response learning
.
Behavioural Brain Research
,
291
,
195
200
.
Leong
,
K. C.
, &
Packard
,
M. G.
(
2014
).
Exposure to predator odor influences the relative use of multiple memory systems: Role of basolateral amygdala
.
Neurobiology of Learning and Memory
,
109
,
56
61
.
Lovallo
,
W.
(
1975
).
The cold pressor test and autonomic function: A review and integration
.
Psychophysiology
,
12
,
268
282
.
Lupien
,
S. J.
,
Maheu
,
F.
,
Tu
,
M.
,
Fiocco
,
A.
, &
Schramek
,
T. E.
(
2007
).
The effects of stress and stress hormones on human cognition: Implications for the field of brain and cognition
.
Brain and Cognition
,
65
,
209
237
.
Mabry
,
T. R.
,
Gold
,
P. E.
, &
McCarty
,
R.
(
1995
).
Age-related changes in plasma catecholamine responses to acute swim stress
.
Neurobiology of Learning and Memory
,
63
,
260
268
.
Mather
,
M.
(
2007
).
Emotional arousal and memory binding: An object-based framework
.
Perspectives on Psychological Science
,
2
,
33
52
.
McCullough
,
A. M.
,
Ritchey
,
M.
,
Ranganath
,
C.
, &
Yonelinas
,
A. P.
(
2015
).
Differential effects of stress-induced cortisol responses on recollection and familiarity-based recognition memory
.
Neurobiology of Learning and Memory
,
123
,
1
10
.
McCullough
,
A. M.
, &
Yonelinas
,
A. P.
(
2013
).
Cold-pressor stress after learning enhances familiarity-based recognition memory in men
.
Neurobiology of Learning and Memory
,
106
,
11
17
.
McEwen
,
B. S.
, &
Sapolsky
,
R. M.
(
1995
).
Stress and cognitive function
.
Current Opinion in Neurobiology
,
5
,
205
216
.
McGaugh
,
J. L.
(
2004
).
The amygdala modulates the consolidation of memories of emotionally arousing experiences
.
Annual Review of Neuroscience
,
27
,
1
28
.
Mednick
,
S. C.
,
Makovski
,
T.
,
Cai
,
D. J.
, &
Jiang
,
Y. V.
(
2009
).
Sleep and rest facilitate implicit memory in a visual search task
.
Vision Research
,
49
,
2557
2565
.
Okuda
,
S.
,
Roozendaal
,
B.
, &
McGaugh
,
J. L.
(
2004
).
Glucocorticoid effects on object recognition memory require training-associated emotional arousal
.
Proceedings of the National Academy of Sciences, U.S.A.
,
101
,
853
858
.
Packard
,
M. G.
(
2009
).
Anxiety, cognition, and habit: A multiple memory systems perspective
.
Brain Research
,
1293
,
121
128
.
Packard
,
M. G.
, &
Gabriele
,
A.
(
2009
).
Peripheral anxiogenic drug injections differentially affect cognitive and habit memory: Role of basolateral amygdala
.
Neuroscience
,
164
,
457
462
.
Packard
,
M. G.
, &
McGaugh
,
J. L.
(
1996
).
Inactivation of hippocampus or caudate nucleus with lidocaine differentially affects expression of place and response learning
.
Neurobiology of Learning and Memory
,
65
,
65
72
.
Packard
,
M. G.
, &
Wingard
,
J. C.
(
2004
).
Amygdala and "emotional" modulation of the relative use of multiple memory systems
.
Neurobiology of Learning and Memory
,
82
,
243
252
.
Park
,
C. R.
,
Zoladz
,
P. R.
,
Conrad
,
C. D.
,
Fleshner
,
M.
, &
Diamond
,
D. M.
(
2008
).
Acute predator stress impairs the consolidation and retrieval of hippocampus-dependent memory in male and female rats
.
Learning & Memory
,
15
,
271
280
.
Payne
,
J. D.
,
Jackson
,
E. D.
,
Ryan
,
L.
,
Hoscheidt
,
S.
,
Jacobs
,
J. W.
, &
Nadel
,
L.
(
2006
).
The impact of stress on neutral and emotional aspects of episodic memory
.
Memory
,
14
,
1
16
.
Perez
,
C. S.
,
Serafin
,
N.
,
Prado-Alcala
,
R. A.
,
Roozendaal
,
B.
, &
Quirarte
,
G. L.
(
2017
).
Glucocorticoid administration into the dorsolateral but not dorsomedial striatum accelerates the shift from a spatial toward procedural memory
.
Neurobiology of Learning and Memory
,
141
,
124
133
.
Plessow
,
F.
,
Schade
,
S.
,
Kirschbaum
,
C.
, &
Fischer
,
R.
(
2012
).
Better not to deal with two tasks at the same time when stressed? Acute psychosocial stress reduces task shielding in dual-task performance
.
Cognitive, Affective & Behavioral Neuroscience
,
12
,
557
570
.
Quirarte
,
G. L.
,
de la Teja
,
I. S.
,
Casillas
,
M.
,
Serafin
,
N.
,
Prado-Alcala
,
R. A.
, &
Roozendaal
,
B.
(
2009
).
Corticosterone infused into the dorsal striatum selectively enhances memory consolidation of cued water-maze training
.
Learning & Memory
,
16
,
586
589
.
Raio
,
C. M.
,
Brignoni-Perez
,
E.
,
Goldman
,
R.
, &
Phelps
,
E. A.
(
2014
).
Acute stress impairs the retrieval of extinction memory in humans
.
Neurobiology of Learning and Memory
,
112
,
212
221
.
Roozendaal
,
B.
(
2002
).
Stress and memory: Opposing effects of glucocorticoids on memory consolidation and memory retrieval
.
Neurobiology of Learning and Memory
,
78
,
578
595
.
Roozendaal
,
B.
, &
McGaugh
,
J. L.
(
2011
).
Memory modulation
.
Behavioral Neuroscience
,
125
,
797
824
.
Roozendaal
,
B.
,
Okuda
,
S.
,
Van der Zee
,
E. A.
, &
McGaugh
,
J. L.
(
2006
).
Glucocorticoid enhancement of memory requires arousal-induced noradrenergic activation in the basolateral amygdala
.
Proceedings of the National Academy of Sciences, U.S.A.
,
103
,
6741
6746
.
Sandi
,
C.
,
Loscertales
,
M.
, &
Guaza
,
C.
(
1997
).
Experience-dependent facilitating effect of corticosterone on spatial memory formation in the water maze
.
European Journal of Neuroscience
,
9
,
637
642
.
Schoofs
,
D.
,
Preuss
,
D.
, &
Wolf
,
O. T.
(
2008
).
Psychosocial stress induces working memory impairments in an n-back paradigm
.
Psychoneuroendocrinology
,
33
,
643
653
.
Schwabe
,
L.
,
Bohringer
,
A.
,
Chatterjee
,
M.
, &
Schachinger
,
H.
(
2008
).
Effects of pre-learning stress on memory for neutral, positive and negative words: Different roles of cortisol and autonomic arousal
.
Neurobiology of Learning and Memory
,
90
,
44
53
.
Schwabe
,
L.
,
Oitzl
,
M. S.
,
Philippsen
,
C.
,
Richter
,
S.
,
Bohringer
,
A.
,
Wippich
,
W.
, et al
(
2007
).
Stress modulates the use of spatial versus stimulus–response learning strategies in humans
.
Learning & Memory
,
14
,
109
116
.
Schwabe
,
L.
,
Oitzl
,
M. S.
,
Richter
,
S.
, &
Schachinger
,
H.
(
2009
).
Modulation of spatial and stimulus–response learning strategies by exogenous cortisol in healthy young women
.
Psychoneuroendocrinology
,
34
,
358
366
.
Schwabe
,
L.
,
Schachinger
,
H.
,
de Kloet
,
E. R.
, &
Oitzl
,
M. S.
(
2010a
).
Corticosteroids operate as a switch between memory systems
.
Journal of Cognitive Neuroscience
,
22
,
1362
1372
.
Schwabe
,
L.
,
Schachinger
,
H.
,
de Kloet
,
E. R.
, &
Oitzl
,
M. S.
(
2010b
).
Stress impairs spatial but not early stimulus–response learning
.
Behavioural Brain Research
,
213
,
50
55
.
Schwabe
,
L.
,
Tegenthoff
,
M.
,
Hoffken
,
O.
, &
Wolf
,
O. T.
(
2013
).
Mineralocorticoid receptor blockade prevents stress-induced modulation of multiple memory systems in the human brain
.
Biological Psychiatry
,
74
,
801
808
.
Schwabe
,
L.
, &
Wolf
,
O. T.
(
2012
).
Stress modulates the engagement of multiple memory systems in classification learning
.
Journal of Neuroscience
,
32
,
11042
11049
.
Schwabe
,
L.
, &
Wolf
,
O. T.
(
2014
).
Timing matters: Temporal dynamics of stress effects on memory retrieval
.
Cognitive, Affective & Behavioral Neuroscience
,
14
,
1041
1048
.
Segal
,
S. K.
,
Simon
,
R.
,
McFarlin
,
S.
,
Alkire
,
M.
,
Desai
,
A.
, &
Cahill
,
L. F.
(
2014
).
Glucocorticoids interact with noradrenergic activation at encoding to enhance long-term memory for emotional material in women
.
Neuroscience
,
277
,
267
272
.
Smeets
,
T.
(
2011
).
Acute stress impairs memory retrieval independent of time of day
.
Psychoneuroendocrinology
,
36
,
495
501
.
Smeets
,
T.
,
Otgaar
,
H.
,
Candel
,
I.
, &
Wolf
,
O. T.
(
2008
).
True or false? Memory is differentially affected by stress-induced cortisol elevations and sympathetic activity at consolidation and retrieval
.
Psychoneuroendocrinology
,
33
,
1378
1386
.
Smith
,
A. M.
,
Floerke
,
V. A.
, &
Thomas
,
A. K.
(
2016
).
Retrieval practice protects memory against acute stress
.
Science
,
354
,
1046
1048
.
Smith
,
K. S.
, &
Graybiel
,
A. M.
(
2016
).
Habit formation
.
Dialogues in Clinical Neuroscience
,
18
,
33
43
.
Thoma
,
M. V.
,
Kirschbaum
,
C.
,
Wolf
,
J. M.
, &
Rohleder
,
N.
(
2012
).
Acute stress responses in salivary alpha-amylase predict increases of plasma norepinephrine
.
Biological Psychology
,
91
,
342
348
.
Tollenaar
,
M. S.
,
Elzinga
,
B. M.
,
Spinhoven
,
P.
, &
Everaerd
,
W.
(
2009
).
Immediate and prolonged effects of cortisol, but not propranolol, on memory retrieval in healthy young men
.
Neurobiology of Learning and Memory
,
91
,
23
31
.
Trammell
,
J. P.
, &
Clore
,
G. L.
(
2014
).
Does stress enhance or impair memory consolidation?
Cognition & Emotion
,
28
,
361
374
.
Tseng
,
Y. C.
, &
Lleras
,
A.
(
2013
).
Rewarding context accelerates implicit guidance in visual search
.
Attention, Perception, & Psychophysics
,
75
,
287
298
.
van Stegeren
,
A. H.
,
Wolf
,
O. T.
, &
Kindt
,
M.
(
2008
).
Salivary alpha amylase and cortisol responses to different stress tasks: Impact of sex
.
International Journal of Psychophysiology
,
69
,
33
40
.
Vanelzakker
,
M. B.
,
Zoladz
,
P. R.
,
Thompson
,
V. M.
,
Park
,
C. R.
,
Halonen
,
J. D.
,
Spencer
,
R. L.
, et al
(
2011
).
Influence of pre-training predator stress on the expression of c-fos mRNA in the hippocampus, amygdala, and striatum following long-term spatial memory retrieval
.
Frontiers in Behavioral Neuroscience
,
5
,
30
.
Vogel
,
S.
,
Klumpers
,
F.
,
Schroder
,
T. N.
,
Oplaat
,
K. T.
,
Krugers
,
H. J.
,
Oitzl
,
M. S.
, et al
(
2016
).
Stress induces a shift towards striatum-dependent stimulus–response learning via the mineralocorticoid receptor
.
Neuropsychopharmacology
,
42
,
1262
1271
.
Wingard
,
J. C.
, &
Packard
,
M. G.
(
2008
).
The amygdala and emotional modulation of competition between cognitive and habit memory
.
Behavioural Brain Research
,
193
,
126
131
.
Zellin
,
M.
,
von Muhlenen
,
A.
,
Muller
,
H. J.
, &
Conci
,
M.
(
2014
).
Long-term adaptation to change in implicit contextual learning
.
Psychonomic Bulletin & Review
,
21
,
1073
1079
.