Running Head: STRESS AND LEARNING An Acute Stressor Enhances Striatal-based Implicit Learning

Stress has widespread effects on cognitive processes such as memory and learning. Studies show that in response to stress, animals and humans often switch from hippocampal-based spatial learning strategies to striatal-based stimulus-response learning strategies. While there is evidence that performance on striatal-based tasks is not impaired by stress, and in fact, may be optimal under stress compared to other strategies, no studies have examined how stress affects implicit learning, specifically tasks that rely on the striatum. In this study, we used the Triplets Learning Task (TLT) to examine the effects of stress on implicit learning. The TLT is an implicit sequence learning task in which participants respond more quickly and accurately to events that occur with high compared to low predictability without being aware that learning is taking place. In the TLT, young adults rely on the striatum for learning, which makes this an ideal task to investigate the effects of stress on implicit learning. To induce stress, we used the cold pressor test (CPT), which causes a physiological stress response similar to that of a mild daily stressor. Twenty-five young adults were randomly assigned to either the Control group (n=13) or the Stress group (n=12). There was a significant Triplet type (High vs. Low predictability) x Condition (Control vs. Stress) interaction (p = 0.025) for accuracy, such that participants in the Stress group learned more than Controls. These findings indicate that, in response to stress, striatal-based implicit sequence learning in young adults is enhanced. STRESS AND LEARNING 3 An Acute Stressor Enhances Striatal-based Implicit Learning Acute stress is known to influence cognitive processes such as memory and learning (Schwabe & Wolf, 2013). Physiologically, stress induces the release of a variety of hormones, including adrenaline and noradrenaline, and leads to the increased activation of the hypothalamus-pituitary-adrenal (HPA) axis, which triggers the release of glucocorticoids (Schwabe & Wolf, 2013). The release of neuropeptides, neurotransmitters, and hormones, such as glucocorticoids, mediate the stress response in both animals and humans (Schwabe & Wolf, 2013). Research shows that stress not only disrupts an organism’s physiological homeostasis, affecting various functions such as metabolism, immune response, and cardiovascular response, but also affects cognitive and affective processes (Pruessner et al., 2010; Dickerson & Kemeny, 2004). Because the hormones secreted in response to stress, particularly glucocorticoids, can easily cross the blood-brain barrier and access the brain, areas with high volumes of glucocorticoid receptors are highly susceptible to the effects of stress (Schwabe & Wolf, 2013). The prefrontal cortex and the hippocampus are two regions of the brain that are particularly sensitive to stress hormones, as they contain a high concentration of glucocorticoid receptors (Schwabe & Wolf, 2013; Pruessner et al., 2010). Learning and memory, cognitive processes supported by the frontal lobe and the hippocampus, are affected by stress due to the impact of stress and glucocorticoids on their underlying regions (Lupien, Maheu, Tu, Fiocco, & Schramek, 2007; Dickerson & Kemeny, 2004). An extensive literature suggests that tasks involving the prefrontal cortex and hippocampus are generally impaired by acute stress. Studies exploring the effects of acute stress on working memory show that when stressed, people show reduced activation in the dorsolateral STRESS AND LEARNING 4 prefrontal cortex (DLPFC) as well as impaired performance on working memory tasks (Qin, Hermans, van Marle, Luo & Fernandez, 2009; Schoofs, Wolf, & Smeets, 2009; Schoofs, Preub, & Wolf, 2008). For example, Schoofs et al. (2009) found that stress impaired working memory performance on O-Span and backward digit span tasks. Similarly, studies show that stress impairs performance on tasks involving the hippocampus (for reviews see Lupien et al., 2007; Wolf, 2006). Kuhlmann, Piel and Wolf (2005) found that in a declarative memory task, participants exposed to a psychosocial stressor performed significantly worse than a control group on free recall of a list of words learned 24 hours earlier, suggesting that stress impaired memory retrieval. Spatial learning, which also involves the hippocampus, has been found to be impaired by stress, as well (e.g. Schwabe, Oitzl, Philippsen, et al., 2007). Although there is evidence that stress negatively affects some forms of learning and memory, the relationship between stress and cognition is complex. The effects of stress on cognition seem to depend on a variety of factors, such as the type of information being learned or the cognitive system mediating performance (Ell, Cosley, & McCoy, 2011). Research shows that in certain doses and conditions, stress can actually improve learning and memory (Lupien et al., 2007; Wolf, 2006). For example, while stress generally impairs hippocampal-based memory and learning, it may prove helpful for memory and recall when the stressor is tied to the learned information; stress enhances consolidation and recall of both emotionally arousing (Kulmann & Wolf, 2006; Lupien et al., 2007; Wolf, 2006) and context-congruent words (i.e. when the to-beremembered material matches the stressor in time and place) (Smeets, Giesbrecht, Jelicic, & Merckelbach, 2007). Furthermore, striatal-based learning seems to be unaffected, or in some cases, enhanced by acute stress (e.g. Lighthall, Gorlick, Schoeke, Frank, & Mather, 2012; Mather & Lighthall, 2012; Schwabe & Wolf, 2013; Schwabe & Wolf, 2012). STRESS AND LEARNING 5 In response to stress, striatal-based learning strategies may be more favorable than hippocampal-based strategies. Behaviors that rely on the striatum can occur quickly and automatically, making the use of this system evolutionarily adaptive in stressful situations (Arnsten, 2009; Schwabe & Wolf, 2013). While certain types of learning and memory that rely on the hippocampus and prefrontal cortex require top-down processing, striatal-based learning and memory processes may be less cognitively demanding, and consequently, may be more efficient in response to stress (Schwabe & Wolf, 2013). The lower cognitive load associated with striatal-based processes may allow more cognitive resources to be allocated toward coping with the stressor (Qin et al., 2009; Schwabe & Wolfe, 2013). Furthermore, during stressful events, there may be a shift toward striatal-based learning to avoid hesitation and delays in carrying out potentially life-saving behaviors, making striatal-based learning more adaptive in the presence of a stressor (Qin et al., 2009; Schwabe & Wolf, 2013). In dual-solution tasks in which either a hippocampal-based strategy or a striatal-based strategy can be used, both animals and humans often switch from hippocampal-based learning strategies to striatal-based learning strategies in response to stress (Schwabe et al., 2007; Schwabe, Schachinger, de Kloet, & Oitzl, 2010a; Schwabe, Schachinger, de Kloet, & Oitzl, 2010b; Schwabe & Wolf, 2013; Schwabe & Wolf, 2012). Schwabe et al. (2007) found that in a task that typically elicits a spatial learning strategy (hippocampal-based), participants exposed to a psychosocial stressor often switched to a stimulus-response strategy (striatal-based) to complete the task. Moreover, data show that under stress, spatial learning is actually impaired, while stimulus-response learning is not (for review see Packard, 2009; Schwabe, Schachinger, de Kloet, & Oitzl, 2010b; Schwabe & Wolf, 2013). In a navigation task in which hippocampaland striatal-based learning strategies were dissociable, Schwabe et al. (2010b) found that stress STRESS AND LEARNING 6 impaired hippocampus-dependent learning but did not impair striatal-based learning in mice. Similarly, Schwabe et al. (2007) found that in humans, stress facilitated striatal-based learning strategies in a spatial learning task that allowed differentiation between hippocampal-based and striatal-based strategies, such that participants who underwent psychosocial stress used stimulusresponse learning strategies (striatal-based) more than spatial learning strategies (hippocampalbased) to complete the task. Consistent with this, a recent fMRI study showed that classification learning, which may be controlled by either hippocampal or striatal memory systems, correlated with hippocampal activation in a control condition, but with striatal activation after stress (Schwabe & Wolf, 2012). Not only does stress have an effect on learning strategies in dual-solution tasks, but it also affects learning in typically striatal-based tasks. For example, Lighthall et al. (2012) found that stress enhances positive reinforcement learning. Using a probabilistic learning task in which participants learn to choose symbols with a higher probability of positive outcomes and to avoid selecting symbols with a higher probability of negative outcomes, they found that stress enhanced learning of associations with positive but not negative outcomes. Thus, under stress, striatal-based learning seems to be either unaffected or enhanced. The striatum is typically recruited for more habitual, procedural types of learning and memory, known as implicit learning (Lupien et al., 2007; Rieckmann & Backman, 2009). Implicit learning occurs when regularities in the environment are acquired without intent to learn or conscious awareness that learning has occurred (Reber, 1989). Implicated in a wide range of skills, implicit learning is essential to learning predictable patterns from environmental input, allowing us to adapt to new people, places, and technologies. Reber (1989) argues that implicit learning is fundamental to human and animal survival, allowing for learning of the world in a STRESS AND LEARNING 7 very basic way. One important type of implicit learning is learning that occurs through incremental stimulus-response binding based on the frequency of stimulus occurrence (Reber, 1989; Rieckmann & Backman, 2009). That is, one can automatically learn the likelihood that a particular event will occur following a certain cue. This type of learning is critical, as it allows individuals to understand the probability that an ecologically important event will occur without ever being consciously aware of this knowledge (Reber, 1989). Studies using various implicit learning tasks have shown that the automatic nature of this type of probabilistic sequence learning enables faster and/or more accurate responses to predictable rather than unpredictable events (Howard Jr. & Howard, 1997; Howard et al., 2004; Howard Jr., Howard, Dennis, & Kelly, 2008). In this study, we explored the effect of stress on implicit sequence learning. While various studies have explored the effects of stress on more declarative, or explicit types of learning, to our knowledge, no study has examined how stress affects implicit learning, specifically probabilistic implicit sequence learning that relies on the striatum. To examine the effects of stress on this type of implicit learning, we used the Triplets Learning Task (TLT; Howard Jr. et al., 2008). The TLT is a probabilistic serial reaction time task derived from the widely used Serial Reaction Time task (SRT; Nissen & Bullemer, 1987) to measure implicit learning (Howard Jr. et al., 2008). Learning in the TLT has been shown to correlate with striatal activation in young adults, making it an ideal task to look at the effects of stress on implicit sequence learning (Simon, Vaidya, Howard Jr., & Howard, 2012). In the TLT, participants are exposed to a probabilistic pattern, and as learning progresses they become faster and more accurate to highly predictable events compared to events that occur with low predictability (Howard Jr. et al., 2008; Simon, Howard Jr., & Howard, 2011). A sensitive measure of implicit STRESS AND LEARNING 8 sequence learning, the TLT provides a continuous, performance-based measure of learning with reduced motor sequencing (Howard Jr. et al., 2008). Furthermore, a recognition measure is used to probe explicit awareness of the learned pattern in the TLT. This measure determines whether or not participants recognize individual triplets as those that occurred with high compared to low predictability; if participants cannot distinguish between these triplets, it can be said that their learning was implicit (Howard Jr. et al., 2008). To induce stress, we used the cold pressor test (CPT), which causes a physiological stress response similar to that of a mild daily stressor (Lovallo, 1975). In the CPT, participants are asked to submerge their hand, up to the wrist, into ice water for as long as possible, up to three minutes. The CPT reliably activates the HPA axis and elicits a stress response (McRae et al., 2006; Schoofs, Wolf, & Smeets, 2009; Lighthall et al., 2012). A low-risk technique, the CPT has been used in a wide range of medical research and across a wide range of populations, including children and older adults (e.g. Mather, Gorlick, & Lighthall, 2009; Lighthall et al., 2012; Von Baeyer, Piira, Chambers, Trapanotto, & Zelter, 2006). This is the first study to examine the effect of stress on implicit sequence learning using a task that relies on the striatum. In line with research showing that striatal-based processes are unaffected or even enhanced by stress, we predicted that implicit sequence learning, as measured by the TLT, would either be unaffected or improved by mild stress induced by the CPT. Method Participants Participants were 25 Georgetown University students (8 male; 17 female, M age = 19.08 ± 1.35), recruited through the Georgetown Research Volunteers Program or word of mouth. STRESS AND LEARNING 9 They received either course credit or a $25 gift certificate to Amazon.com for participating. The Georgetown University Institutional Review Board approved all procedures. Study Design This was a 2 (Group: Stress vs. Control) x 4 (Epoch) x 2 (Triplet type: High vs. Low) design, where Group varied between-subjects, and Epoch and Triplet type varied within-subjects. Participants were randomly assigned to either the Stress (5 male; 7 female, M age = 18.75 ± 1.14) or Control group (3 male; 10 female, M age = 19.38 ± 1.50). Cold Pressor Task The cold pressor task (CPT) was used to induce stress in participants assigned to the Stress group. In the CPT, participants were asked to place their right hand, up to the wrist, into a bucket of water ranging from 0.6°-1.7° C for as long as possible, up to three minutes. Participants were explicitly told that if at any point they felt they needed to remove their hand, they could do so, otherwise, the experimenter would notify them when to remove their hand. The CPT is a reliable and valid manipulation that has been used in a variety of psychological studies to effectively mimic the physiological response to mild stress (Schwabe & Wolfe, 2012; Lighthall et al., 2009; Schoofs et al., 2009, etc.). Subjects in the Control group were treated identically in all ways, except that the water was room temperature (~24° C). The apparatus used for the CPT was a plastic container filled with approximately 48 ounces of water. For the Stress group, ice cubes were added to the water to cool it to between ~0° -2°C. A ThermoWorks (Lindon) thermometer was used to attain precise measurements of water temperature in both groups, and an iPhone 4s was used to record how long participants had their hands submerged in the water. A Tetra Whisper10 air pump (Blacksburg) was used to keep STRESS AND LEARNING 10 the water continuously circulating, preventing warm pockets from forming around the subjects’ hand. Triplets Learning Task The Triplets Learning Task (TLT) is a probabilistic serial reaction time task used to measure implicit learning (Howard Jr. et al., 2008; Figure 1). Participants view four open circles in a horizontal line across the computer screen. On each trial, participants observe a series of cues consisting of two red cues, followed by a green target to which they respond. Each red cue appears for 120 ms with 150 ms between cues, and the green cue remains on the screen until a correct response is made. In the current experiment, unbeknownst to participants, the first cue predicted the location of the target, but in a probabilistic manner, such that the target was more predictable in some triplets (high predictability) than in others (low predictability). Participants encountered 16 different high predictability (HP) triplets, presented on 80% of the trials, and 48 different low predictability (LP) triplets, presented on 20% of the trials (Howard Jr. et al., 2008). For example, a participant might have been assigned the pattern of 1r2r3r4r, where 1, 2, 3 and 4 represent the possible locations of cues and targets in the four open circles from left to right, and “r” denotes a random occurrence in any of the four possible locations (i.e. 1r2 means 112,122,132, or 142). In this case, triplets such as 1r2, 2r3, 3r4 and 4r1 would occur with high predictability, such that participants learn that the location of the first cue will predict the location of the target. However, on 20% of the trials, the participant would encounter low predictability triplets, such that the first cue will predict the location of the target in any of the remaining 3 locations. For example, in pattern 1r2r3r4r, if cue 1 is in location 1, a target in any location other than 2 would be a low predictability triplet. Learning is revealed through participants responding faster and STRESS AND LEARNING 11 more accurately to triplets that occur with high predictability than to triplets with low predictability (Howard Jr. et al., 2008; Simon et al., 2011). Overall, there are six possible triplet patterns for the TLT; each individual received one of these six patterns. In this experiment, all six patterns occurred equally often in both groups. There were 50 triplet trials per block, 5 blocks per epoch, and 2 epochs per session. Between blocks, participants were allowed to take a short break, with a longer break occurring between sessions. Recognition Test To probe explicit awareness of the pattern seen during the TLT, participants were given a recognition test in which they were presented with the 64 possible triplets they saw during the TLT, one time each in a randomly determined order, using the same event timing that they saw during the task. In this case, they did not respond to the target position as they had during the TLT, but instead were asked to rate the frequency with which they thought each triplet had occurred during the TLT by pressing “1” if they thought a given triplet occurred “more often,” and “2” if they thought that it occurred “less often.” In previous studies, results from the recognition test have shown that people are not explicitly aware of the pattern learned in the TLT (Howard Jr. et al., 2008; Simon et al., 2011; Simon et al., 2012), confirming that the learning is implicit. Procedure Upon arrival, participants filled out a consent form, a biographical questionnaire, and a health-screening questionnaire. Next, participants’ left index fingers were fitted with a Nature Spirit Wearable Pulse Oximeter Model: CMS-50F heart rate monitor (Columbia). After recording baseline heart rate for two minutes, participants were read the instructions for the CPT, and were asked to submerge the right hand into the water. After participants removed their hand STRESS AND LEARNING 12 from the water, they rated how painful the experience of submerging their hand was on a scale of 0-10, with 0 being “not painful at all” and 10 being “the worst pain you have ever felt.” Participants then completed the Positive and Negative Affective Scale (PANAS), a self-report measure of the positive and negative affect that a person is experiencing at a given time (Watson, Clark, & Tellegen, 1988). Together, these measures were used to indicate the subjective painfulness of the CPT and to capture any differences in affect possibly relating to the CPT. Following the CPT, pain rating, and PANAS, participants were read instructions for the TLT. Participants were told that four circles would be presented in a horizontal line across the computer screen, and that they would fill in sequentially red, red, and then green. They were asked to respond to the green “target” by pressing the key on the stimulus-response (SR) box— “7, 8, 9” and “-”— which corresponded to locations 1, 2, 3 and 4, respectively. Participants were given feedback to focus either more on accuracy or more on speed. Unbeknownst to them, this feedback guided them to maintain a 92% accuracy rate. Participants are guided toward this 92% accuracy level because attempting to maintain 100% accuracy would cause them to respond more slowly, and would impair our ability to detect learning in the TLT. There were short breaks between blocks and longer breaks between sessions. Importantly, all participants used their right hand both to the complete the CPT and to respond to the TLT stimuli, despite hand dominance. This was done because the heart rate monitor was on the left hand throughout the experiment in all participants. There was approximately 5 minutes between the CPT and TLT during which time participants could dry and warm their right hand. After completing the TLT, which took approximately 40 minutes, participants were given the recognition test. The experimental session concluded with Backward Digit Span, a standard STRESS AND LEARNING 13 test of working memory. Before leaving, all participants were debriefed and either paid or awarded class credit for their participation. Results Stress Manipulation Group differences in subjective experience of the CPT, including length of time with hand submerged and pain ratings, allowed quantification of the success of the stress manipulation. There was a significant difference in the length of time the two groups had their hands submerged, t(23) = 4.54, p < 0.001 (Stress group: M = 96.23 seconds, SD = 19.26; Control group: M = 180 seconds, SD = 0.00). There was also a significant difference in subjective pain ratings, t(23) = -10.24, p < 0.001, such that participants in the Stress group found the experience of submerging their hand into the water to be more painful (M = 4.75, SD = 2.21) than Controls (M = 0.31, SD = 0.48). There were no significant gender differences on either the length of time the hand was submerged or on subjective pain ratings (p > 0.05). Implicit Learning To measure implicit learning in the TLT, we compared performance on highpredictability triplets to performance on low-predictability triplets, using measures of mean of median reaction time (MMRT) on correct trials and mean accuracy. Because people often show pre-existing response tendencies to repetitions (e.g. 111, 222) and trills (e.g. 121, 232), responding significantly more slowly to these trials even prior to training, and because these two trial types occur with low predictability in all patterns (given the nature of the regularity used) and thus are not counterbalanced, these triplet types were excluded from the analyses reported below (Howard Jr. et al., 2008; Howard et al., 2004). There were ten blocks in each session, but data were grouped and analyzed by epoch, such that there were five blocks per epoch, and four epochs. Thus, in order to obtain mean of STRESS AND LEARNING 14 median RTs, median RTs were determined for correct responses for each Triplet type (HP and LP) for each participant on each block, and these medians were then averaged across five blocks to obtain a single mean RT for each individual for each Triplet type for each of the four epochs. Similarly, mean accuracy was determined for each participant for each Triplet type for each of the four epochs. Reaction time Figure 2a shows the mean of median reaction time (MMRT) to HP versus LP triplets, split by group, across epochs, and Figure 2b shows the MMRT difference scores (MMRT to LP – MMRT to HP), split by group across epochs. A 2 (Group: Stress and Control) x 4 (Epoch: 1-4) x 2 (Triplet type: HP and LP) mixeddesign ANOVA was conducted for MMRT. There was a significant main effect of Epoch, F(3,69) = 21.68, p < 0.001, such that participants responded more quickly over time (E1: M = 392.03, SD = 40.85; E2:M = 375.73, SD = 51.45; E3: M = 360.06, SD = 46.57; E4: M = 359.32, SD = 43.33). There was also a significant main effect of Triplet type, F(1,23) = 37.63, p < 0.001), such that participants responded faster to HP triplets (M = 363.65, SD = 42.12) than to LP triplets (M = 379.88, SD = 45.66), signaling that they did in fact learn about triplet probabilities. There were no significant interactions of Group x Triplet type, Epoch x Triplet type, or of Group x Epoch x Triplet type, p > 0.05, indicating that by the MMRT measure the groups showed similar learning. Accuracy The same 2 x 4 x 2 ANOVA was run for accuracy; these data are shown in Figures 3a and 3b. The analysis showed a significant main effect of Epoch (F(3,69)=6.98, p= 0.004), such that accuracy changed over time (E1: M = 0.94, SD = 0.03; E2: M = 0.93, SD = 0.03; E3: M = 0.93, SD = 0.03; E4: M = 0. 92, SD = 0.03), converging to the 92% accuracy to which the end of block feedback was pushing participants. There was also a significant main STRESS AND LEARNING 15 effect of Triplet type, F(1,23) = 10.18, p = 0.004), such that participants responded more accurately to HP triplets (M = 0.93, SD = 0.02) than to LP triplets (M = 0.92, SD = 0.03). Most important, there was a significant Group x Triplet type interaction, F(1,23) = 6.06, p = 0.022. Post hoc analyses show that the HP versus LP difference was greater for the Stress group (M difference score = 0.02, SD = 0.02) than for the Control group (M difference score = 0.003, SD = 0.012; t(23) = -2.20, p = 0.038), suggesting that the Stress group learned more about the predictable pattern than the Control group according to the accuracy measure. The Epoch x Triplet type and Group x Epoch x Triplet type interactions were not significant, p > 0.05. Notably, there was no main effect of Group (p > 0.05) in the MMRT or the accuracy analyses, suggesting that the CPT only affected TLT learning, and not overall speed or accuracy. Individual differences in learning Because the above analyses suggest that learning, as measured via accuracy, is sensitive to the effects of stress from the CPT, we examined the extent to which this pattern can be seen for individuals. We calculated individual learning scores by subtracting overall mean accuracy to LP triplets from overall mean accuracy to HP triplets, obtaining a mean accuracy difference score for each subject. To examine whether, on average, more participants in the Stress group had better learning, we first did a median split (median= 0.006) on the mean accuracy difference scores. We defined individuals with a difference score above this median as “high learners,” and individuals below this median as “low learners” (Figure 4). We then did a Chi square test to determine if there were more “high learners” in the Stress group than in the Control group. There was a significant Chi square, such that more people in the Stress group were high learners than people in the Control group, Χ = 4.89, p = 0.027, indicating that individuals in the Stress group had greater overall learning. STRESS AND LEARNING 16 Recognition To determine whether TLT learning was implicit, we assessed ratings in the recognition task (Figure 5). A Group (Stress, Control) x Triplet type (HP, LP, repetitions and trills) repeated measures ANOVA was conducted on mean triplet ratings, and revealed a significant main effect of Triplet type, F(3, 69) = 3.83, p = 0.013, showing that ratings to the four triplet types were different. The ANOVA also revealed a marginally significant Group x Triplet type interaction, F(3,69) = 2.404, p = 0.075, showing that the two groups rated the four triplet types differently. Because our measure of learning was based on participant responses to HP versus LP triplets, however, we were concerned about explicit awareness of HP and LP triplets. Therefore, we excluded repetitions and trills, which were used primarily as a test of whether participants understood the recognition rating task, from our recognition analyses, as well. A Group x Triplet type (HP, LP) repeated measures ANOVA was conducted, and revealed a significant main effect of Triplet type, F(1, 23) = 3.97, p = 0.058, showing that ratings to HP and LP triplet types were different; participants rated HP triplets as having occurred more often. Because these analyses indicated awareness (rating that high predictability triplets had occurred significantly more often), we ran Chi square analyses for each individual to identify anyone who may have had explicit awareness about the pattern. This analysis showed that one person in the Stress group rated HP triplets as having occurred significantly more frequently than LP triplets. We ran the Group (Stress, Control) x Triplet type (HP, LP) ANOVA on recognition ratings after excluding this individual, and found that the Triplet type main effect was no longer significant (p > 0.05). This suggests that this individual’s ratings were driving the main effect of Triplet type reported above. To further explore whether this person’s “awareness” affected performance on TLT learning, we correlated the Chi square values of each individual with mean overall accuracy TLT difference scores. We ran this analysis both with and without the “aware” STRESS AND LEARNING 17 participant, and found that there was no significant correlation in either analysis. This suggests that awareness, as determined through the recognition measure, was not related to learning in the TLT in this study (Figures 6a and 6b). To ensure that the “aware” individual did not affect overall learning seen in the TLT, we reran all analyses reported above, excluding this individual. In all MMRT and accuracy measures, we found the same pattern of results as those reported above. Therefore, we suggest that any awareness seen in the recognition measure was not indicative of actual explicit awareness of the pattern, and thus we included the individual in our general analyses. Additional measures We compared participants in the two groups on the additional measures using unpaired t-tests. There were no significant group differences in positive or negative affect using the PANAS (p > 0.05), nor in working memory as measured by Backward Digit Span (p > 0.05), indicating that the stress manipulation did not influence affect or working memory. Discussion The present study examined the effects of acute stress on implicit sequence learning using the TLT. We found that stress enhanced implicit sequence learning in young adults. Young adults who underwent mild stress, induced by the CPT, responded more accurately to triplets that occurred with high predictability than to those that occurred with low predictability, compared to the non-stressed Control group. Although learning was also seen via the MMRT measure, such that participants responded faster to HP triplets than to LP triplets, there were no significant differences between groups. The lack of a difference may be explained by a floor effect, as reaction time in the Control group was fast, making it difficult for the Stress group to respond faster. Importantly, there were no group differences in overall reaction time or accuracy STRESS AND LEARNING 18 but rather there were group differences in learning as seen through the accuracy Triplet type effect, such that stress affected learning and not general performance. To our knowledge, this is the first study to examine how stress affects implicit sequence learning using a specifically striatal-dependent task. Previous research shows that stress negatively affects certain types of learning and memory, such as working memory, declarative memory, and spatial learning; however, there is also evidence that in some situations, stress does not affect, or may even enhance learning and memory (Lupien et al., 2007; Wolf, 2006). The present study supports the idea that stress enhances implicit sequence learning in young adults. Our findings are in line with previous research showing that in response to stress, animals and humans may favor striatal-based processes (Schwabe et al., 2007; Schwabe et al., 2010a; Schwabe & Wolf, 2013; Schwabe & Wolf, 2012). In the TLT, learning has been shown to positively correlate with striatal activation in young adults (Simon et al., 2012). Specifically, in a functional neuroimaging study examining the neural underpinnings of implicit sequence learning in the TLT in younger and older adults, Simon et al. (2012) found that increased striatal activation was associated with greater implicit learning in young adults (Simon et al., 2012), suggesting that the striatum may be the most efficient system for implicit sequence learning in young adults. Similarly, Ashby, Turner, and Horvitz (2010) reported that the striatum is critical for sequence learning tasks. Thus, our findings expand on previous research, suggesting that stress enhances striatal efficiency and the associated learning. It is important to note, however, that implicit learning involves an interaction between the striatum and the hippocampus. Although research shows that the striatum is more suitable for learning probabilistic regularities than the hippocampus (Ashby et al., 2010; Hartley & Burgess, 2005), implicit sequence learning relies on both striatal and extrastriatal brain regions STRESS AND LEARNING 19 (Rieckmann & Backman, 2009; Poldrack & Packard, 2003). Various studies have shown hippocampal involvement in early implicit sequence learning (e.g. Schendan, Searl, Melrose, & Stern, 2003; Rieckmann, Fischer & Backman, 2010; Bennett, Madden, Vaidya, Howard Jr,, & Howard, 2011; Simon et al., 2012). For example, Schendan et al. (2003) reported both hippocampal and striatal involvement in a motor-based serial reaction time task, with greater hippocampal activation in early learning than in later learning. Similarly, Simon et al. (2012) found both hippocampal and striatal activation in early learning in young adults, with later learning positively correlating only to the amount of striatal activation. Implications for aging The striatum and hippocampus may interact differently in implicit sequence learning in older adults than in younger adults, such that the hippocampus remains active throughout learning in older adults. Therefore, this study should be expanded to explore how stress affects implicit sequence learning in older adults differently than in young adults. Research shows that in older adults, the hippocampus is involved in both early and late learning in tasks where use of the striatum may be more efficient in younger adults (Rieckmann et al., 2010; Dennis & Cabeza, 2010; Simon et al., 2012). While the relationship between the striatum and hippocampus may be competitive in young adults (Rieckmann et al., 2010; Foerde, Knowlton, & Poldrack, 2006; Poldrack & Packard, 2003; Poldrack et al., 2001), it may be cooperative during implicit sequence learning in older adults (Dennis & Cabeza, 2010). Rieckmann et al. (2010) showed that increases in striatal activation and decreases in hippocampal activation were beneficial to performance on a serial reaction time task in young adults, whereas older adults showed continued hippocampal involvement in later implicit learning. Simon et al. (2012) also showed that late in training, individual differences in implicit STRESS AND LEARNING 20 sequence learning were related to individual differences in striatal activation in young adults and hippocampal activation in older adults. Older adults have pronounced declines in the striatum (Bennett et al., 2011; Raz et al., 2005; Gunning-Dixon, Head, McQuain, Acker, & Raz, 1998), which is associated with changes in how the striatum is used in certain tasks compared to younger adults (Maddox, Pacheco, Reeves, Zhu, & Schnyer, 2010; Simon et al., 2012). Older adults do not learn as much or as quickly on the TLT, and recruit brain regions differently with training than young adults (Simon et al., 2012). The hippocampus is involved throughout learning in the TLT in older adults (Simon et al., 2012), and may be the most efficient system for them in this type of learning (Dennis & Cabeza, 2010). Because the hippocampus is particularly sensitive to stress, however, stress may tax the hippocampus such that it may no longer be the most efficient for implicit learning in older adults, and may no longer serve to compensate for impaired striatal function. Therefore, we predict that stress will impair TLT learning in older adults. Limitations It is important to note some limitations to this study. First, no physiological measures of stress were used to ensure that the CPT did in fact stress participants. Although heart rate was taken throughout the experiment, it was used as an “at home” measure to explore whether the CPT caused physiological changes in participants. To date, we are unsure how to quantify this data but intend to do so in the future. Also, in future studies, cortisol should be measured to ensure that the CPT is inducing stress. Stress induces the release of corticosteroids, modulating cognitive performance (Romer, Schulz, Richter, Lass-Hennemann, & Schachinger, 2011; Lupien et al., 2005; Lupien & McEwen, 1997). Increased levels of cortisol have been seen in previous studies using the CPT, indicating the effectiveness of the CPT (e.g. Lighthall et al., 2012; Schwabe & Wolf, 2012), so even though we did not measure cortisol directly in this study, STRESS AND LEARNING 21 we feel confident that the stress manipulation had similar effects to those seen in previous studies. Furthermore, research shows that peak cortisol levels occur approximately 20 minutes after a stressor (Schwabe, Haddad, & Schachinger, 2008), which corresponds to about Epoch 2 in the present experiment. Notably, it was in Epoch 2 that significant differences emerged in accuracy between the Stress group and Control group. Similarly, there was a trend toward significant differences in MMRT difference scores between the two groups during Epoch 2, as well. While group differences in performance in the present study indicate that the CPT did induce stress, it is important to verify this finding with a physiological measure. A second limitation in this study is that because of variations in cortisol levels, it is important to control for possible confounds such as time of day, exercise, and recency of food intake when administering the CPT and subsequent learning task. These considerations were not taken in the present study. A third limitation is that the range of water temperature used in the CPT was very small and particularly cold. While previous studies used water ranging from 0°-5° C (Lighthall et al., 2012) or 0°-2° C (Schwabe et al., 2012), the water in the current study ranged from 0.6°-1.7° C. Because of the cold water temperatures, participants in the stress condition were generally unable to maintain hand submersion for the entire 3 minutes, possibly weakening the effect of the CPT. Mitchell, MacDonald and Brodie (2004), found that variations in water temperature as small as 2° C may result in significant differences in pain experiences in both men and women. Another possible concern is that the extreme cold could have affected the response times of the Stress group because they had their right hand submerged in the ice water and then subsequently used it to respond to the TLT using the SR box. However, this is unlikely because we found no significant group differences in reaction time, indicating that the water temperature did not affect STRESS AND LEARNING 22 general responding. Finally, in the present study there was no verbal measure of explicit awareness in the TLT. Although previous studies show that participants are generally unaware of the learned regularities in the TLT and are unable to report learning strategies used in the task (Howard Jr. et al., 2008; Simon et al., 2011), this study is limited in that there is only one measure of explicit awareness and not the additional verbal report. Despite these limitations, however, this study provides the first evidence that mild stress improves implicit sequence learning on a striatal-based implicit learning task. Future studies should replicate and extend these findings to further understand the complex relationship between stress and cognition. STRESS AND LEARNING23 ReferencesArnsten, A. F. T. (2009). Stress signaling pathways that impair prefrontal cortex structure andfunction. Nature, 10, 410-422.Ashby, F. G., Turner, B. O., & Horvitz, J. C. (2010). Cortical and basal ganglia contributions tohabit learning and automaticity. Trends in Cognitive Sciences, 14, 208-215. Bennett, I. J., Madden, D. J., Vaidya, C. J., Howard Jr., J. H., & Howard, D. V. (2011). Whitematter integrity correlates of implicit sequence learning in healthy aging. Neurobiologyof Aging, 32, 2317.1-2317.12.Dennis , N. A., & Cabeza , R. (2010). Age-related dedifferentiation of learning systems: AnfMRI study of implicit and explicit learning. Neurobiology of Aging, 32, 2318.e17–2318.e30.Dickerson, S. S., & Kemeny, M. E. (2004). Acute stressors and cortisol responses: A theoreticalintegration and synthesis of laboratory research. Psychological Bulletin, 130, 355-391.Ell, S. W., Cosley, B., & McCoy, S. K. (2011). When bad stress goes good: increased threatreactivity predicts improved category learning performance. Psychonomic Bulletin &Review, 18, 96-102. Foerde, K., Knowlton, B. J., & Poldrack, R. A. (2006). Modulation of competing memorysystems by distraction. PNAS, 103, 11778-11783. Gunning-Dixon, F. M., Head, D., McQuain, J., Acker, J. D., & Raz, N. (1998). Differential agingof the human striatum: A prospective MR imaging study. American Journal ofNeuroradiology, 19, 1501–1507. Hartley, T., & Burgess, N. (2005). Complementary memory systems: Competition, cooperationand compensation. Trends in Neuroscience, 28, 169-170. STRESS AND LEARNING24 Howard, D. V., Howard Jr., J. H., Japikse, K., DiYanni, C., Thompson, A., & Somberg, R.(2004). Implicit sequence learning: Effects of level of structure, adult age, and extendedpractice. Psychology and Aging, 19, 79-92.Howard Jr., J. H., & Howard, D. V. (1997). Age differences in implicit learning of higher orderdependencies in serial patterns. Psychology and Aging, 12, 634–656. Howard Jr., J. H., Howard, D. V., Dennis, N. A., & Kelly, A. J. (2008). Implicit learning ofpredictive relationships in three-element visual sequences by young and old adults.Journal of Experimental Psychology: Learning, Memory, and Cognition, 34, 1139-1157.Kuhlmann, S., Piel, M., & Wolf, O. T. (2005). Impaired memory retrieval after psychosocialstress in healthy young men. The Journal of Neuroscience, 25, 2977-2982.Kuhlmann, S., & Wolf, O. T. (2006). Arousal and cortisol interact in modulating memoryconsolidation in healthy young men. Behavioral Neuroscience, 120, 217-223.Lighthall, N. R., Gorlick, M. A., Schoeke, A., Frank, M. J., & Mather, M. (2012). Stressmodulates reinforcement learning in younger and older adults. Psychology and Aging, 28,35-46.Lovallo, W. (1975). The cold pressor test and autonomic function: a review and integration.Psychophysiology, 12, 268-282.Lupien, S. J., Fiocco, A., Wan, N., Maheu, F., Lord, C., Schramek, T., & Tu, M. T. (2005). Stress hormones and human memory function across the lifespan.Psychoneuroendocrinology, 30, 225-242. 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. STRESS AND LEARNING25 Lupien, S. J., & McEwen, B. S. (1997). The acute effects of corticosteroids on cognition:Integration of animal and human model studies. Brain Research Reviews, 24, 1-27.Maddox, W. T., Pacheco, J., Reeves, M., Zhu, B., & Schnyer, D. M. (2010). Rule-based andinformation-integration category learning in normal aging. Neuropsychologia, 48, 2998-3008. Mather, M., Gorlick, M. A., & Lighthall, N. R. (2009). To brake or accelerate when the lightturns yellow?: Stress reduces older adults' risk taking in a driving game. PsychologicalScience, 20, 174-176.Mather, M., & Lighthall, N. R. (2012). Risk and reward are processed differently in decisionsmade under stress. Current Directions in Psychological Science, 21, 36-41.McRae, A. L., Saladin, M. E., Brady, K. T., Upadhyaya, H., Back, S. E., & Timmerman, M. A.(2006). Stress reactivity: biological and subjective responses to the cold pressor and triersocial stressors. Human Psychopharmacology, 21, 377-385.Mitchell, L. A., MacDonald, R. A. R., & Brodie, E. E. (2004). Temperature and the cold pressortest. The Journal of Pain, 5, 233-238.Nissen, M. J. & Bullemer, P. (1987). Attentional requirements of learning: Evidence fromperformance measures. Cognitive Psychology. 19, 1–32.Packard, M. G. (2009). Anxiety, cognition, and habit: A multiple memory systems perspective.Brain Research, 121-128.Poldrack, R. A., Clark, J., Pare-Blagoev, E. J., Shohamy, D., Moyano, J. C., Myers, C., & Gluck,M. A. (2001). Interactive memory systems in the human brain. Nature, 414, 546-550. STRESS AND LEARNING26 Poldrack, R. A., & Packard, M. G. (2003). Competition among multiple memory systems:converging evidence from animal and human brain studies. Neuropsychologia, 41, 245-251.Pruessner, J. C., Dedovic, K., Pruessner, M., Lord, C., Buss, C., Collins, L...& Lupien, S. J.(2010). Stress regulation in the central nervous system: evidence from structural andfunctional neuroimaging studies in human populations. Psychoneuroendocrinology, 35,179-191.Qin, S., Hermans, E. J., van Marle, H. J. F., Luo, J., & Fernandez, G. (2009). Acutepsychological stress reduces working memory-related activity in the dorsolateralprefrontal cortex. Biological Psychiatry, 66, 25-32.Raz, N., Lindenberger, U., Rodrigue, K. M., Kennedy, K. M., Head, D., Williamson, A.,Williamson, A., Dahle, C., Gerstorf, D. & Acker, J. D. (2005). Regional brain changes inaging healthy adults: General trends, individual differences and modifiers. CerebralCortex, 15, 1676–1689.Reber, A. S. (1989). Implicit learning and tacit knowledge. Journal of ExperimentalPsychology, 118, 219-235. Rieckmann, A., & Backman, L. (2009). Implicit learning in aging: Extant patterns and newdirections. Neuropsychological Review, 19, 490-503. Rieckmann, A., Fischer, H., & Backman, L. (2010). Activation in striatum and medial temporallobe during sequence learning in younger and older adults: Relations to performance.NeuroImage, 50, 1303-1312. Romer, S., Schulz, A., Richter, S., Lass-Henneman, J., & Schachinger, H. (2011). Oral cortisolimpairs implicit sequence learning. Psychopharmacology, 215, 33-40. STRESS AND LEARNING27 Schendan, H. E., Searl, M. M., Melrose, R. J., & Stern, C. E. (2003). An fMRI study of the roleof the medial temporal lobe in implicit and explicit sequence learning. Neuron, 37, 1013-1025.Schoofs, D., Preub, D., & Wolf, O. T. (2008). Psychosocial stress induces working memoryimpairments in an n-back paradigm. Neuropsychoendocrinology, 33, 643-653. Schoofs, D., Wolf, O. T., & Smeets, T. (2009). Cold pressor stress impairs performance onworking memory tasks requiring executive functions in healthy young men. BehavioralNeuroscience, 123, 1066-1075.Schwabe, L., Haddad, L., & Schachinger, H. (2008). HPA axis activation by a socially evaluatedcold-pressor test. Psychoneuroendocrinology, 33, 890-895.Schwabe, L., Oitzl, M. S., Philippsen, C., Richter, S., Bohringer, A., Wippich, W., &Schachinger, H. (2007). Stress modulates the use of spatial versus stimulus-responselearning strategies in humans. Learning and Memory, 14, 109-116.Schwabe, L., Schachinger, H., de Kloet, E. R., & Oitzl, M. S. (2010a). Corticosteroids operate asa switch between memory systems. Journal of Cognitive Neuroscience, 22, 13621372.Schwabe, L., Schachinger, H., de Kloet, E. R., & Oitzl, M. S. (2010b). Stress impairs spatial butnot early stimulus–response learning. Behavioral Brain Research, 213, 50-55.Schwabe, L., & Wolf, O. T. (2013). Stress and multiple memory systems: from ‘thinking’ to‘doing’. Trends in Cognitive Sciences, 17, 60-68.Schwabe, L., & Wolf, O. T. (2012). Stress modulates the engagement of multiple memorysystems in classification learning. The Journal of Neuroscience, 32, 11042-11049. STRESS AND LEARNING28 Simon, J. R., Howard Jr., J. H., & Howard, D. V. (2011). Age differences in implicit learning ofprobabilistic unstructured sequences. Journal of Gerontology: PsychologicalSciences, 66B, 32-38.Simon, J. R., Vaidya, C. J., Howard Jr., J. H., & Howard, D. V. (2012). The effects of aging onthe neural basis of implicit associative learning in a probabilistic triplets learning task.Journal of Cognitive Neuroscience, 24, 451-463.Smeets, T., Giesbrecht, T., Jelicic, M., & Merckelbach, H. (2007). Context-dependentenhancement of declarative memory performance following acute psychosocial stress.Biological Psychology, 76, 116-123.Von Baeyer, C. L., Piira, T., Chambers, C. T., Trapanotto, M., & Zeltzer, L. K.(2005). Guidelines for the cold pressor task as an experimental pain stimulus for use withchildren. The Journal of Pain, 6, 218-227.Watson, D., Clark, L. A., & Tellegen, A. (1988). Development and validation of brief measuresof positive and negative affect: The PANAS scales. Journal of Personality and SocialPsychology, 54, 1063-1070.Wolf, O. T. (2006). Effects of stress hormones on the structure and function of the human brain.Expert Review of Endocrinology & Metabolism, 1, 623-632. STRESS AND LEARNING29 Figure 1. Triplets learning task. Sample of one triplet in the TLT. STRESS AND LEARNING30 Figure 2a. Mean of median reaction times for correct trials (MMRT) over epochs for highpredictability (HP) and low predictability (LP) triplets by condition. Error bars represent thestandard error of the mean. Also 2b. Difference scores of mean of median reaction times for correct trials (i.e. MMRT for LPtriplets-MMRT for HP triplets) over epochs by condition. Error bars represent the standard errorof the mean. STRESS AND LEARNING31 Figure 3a. Mean accuracy over epochs for high predictability (HP) and low predictability (LP)triplets by condition. Error bars represent the standard error of the mean. Also 3b. Difference scores of mean accuracy (i.e. mean accuracy for HP triplets-mean accuracyfor LP triplets) over epochs by condition. Error bars represent the standard error of the mean. STRESS AND LEARNING32 Figure 4. Individual learning scores for overall mean accuracy by condition, where each barrepresents an individual. The red line indicates the median value of overall mean accuracydifference scores (i.e., accuracy learning measure) for all participants. STRESS AND LEARNING33 Figure 5. Mean recognition ratings of “Occurred More Often,” by Triplet type and Condition.Error bars represent the standard error of the mean. STRESS AND LEARNING34 Figure 6a. Correlation between Chi square values for each individual with mean overallaccuracy difference score, excluding “aware” participant. Also 6b. Correlation between Chi square values for each individual with mean overall accuracydifference score, including “aware” participant (circled).-10123456789 ChiSquare -.02 -.01 0 .01 .02 .03 .04 .05 .06 .07ControlStress Overall Mean Accuracy Difference Scores-.50.511.522.533.5 ChiSquare -.02 -.01 0 .01 .02 .03 .04 .05 .06 .07j