Effects of Acute Stress on Rigid Learning,Flexible Learning,and Value-Based Decision-Making in Spatial Navigation

Abstract

The current study investigated how stress affects value-based decision-making during spatial navigation and different types of learning underlying decisions. Eighty-two adult participants (42 females) first learned to find object locations in a virtual environment from a fixed starting location (rigid learning) and then to find the same objects from unpredictable starting locations (flexible learning). Participants then decided whether to reach goal objects from the fixed or unpredictable starting location. We found that stress impairs rigid learning in females, and it does not impair, and even improves, flexible learning when performance with rigid learning is controlled for. Critically, examining how earlier learning influences subsequent decision-making using computational models, we found that stress reduces memory integration, making participants more likely to focus on recent memory and less likely to integrate information from other sources. Collectively, our results show how stress impacts different memory systems and the communication between memory and decision-making.

Keywords

psychological stress learning episodic memory decision making spatial perception open data

Stress is ubiquitous in our environment, and we often learn and make decisions on the basis of our knowledge under stress. Decades of research have revealed that acute stress affects learning and decision-making (for reviews, see Vogel & Schwabe, 2016; Porcelli & Delgado, 2017). Most research investigating how acute stress affects learning focuses on how stress affects memory (Buchanan et al., 2006; Elzinga et al., 2005; Gagnon et al., 2019; Gagnon & Wagner, 2016; Guenzel et al., 2013; Kuhlmann et al., 2005; Schwabe et al., 2007, 2008; Smeets et al., 2007; Zoladz et al., 2011) and whether stress shifts the balance between habitual or rigid behaviors and cognitive map-like, flexible behavior (Brown et al., 2020; Brunyé et al., 2016; Park et al., 2017; Raio et al., 2020; Schwabe et al., 2007; Schwabe & Wolf, 2011; van Gerven et al., 2016). One implication from these literatures is that when we are faced with a decision between different strategies to meet a goal, to the extent that we derive values for the different strategies from memory (He, Liu, Beveridge, et al., 2022; He, Liu, Eschapasse, et al., 2022), stress may fundamentally change how our strategy decision reflects past experiences. Our primary goal in the present experiment was to test how stress affects different types of learning and how memory is used in subsequent decision-making.

The distinction between memory from rigid learning and flexible learning has received substantial interest recently (Collins, 2019; Eckstein et al., 2021; Lockwood & Klein-Flügge, 2021), and in navigation, literature reveals that people fall along a continuum in how much they adhere to previously attempted routes or leverage spatial knowledge from those routes to navigate flexible or even novel paths to a goal (Anggraini et al., 2018; He, Liu, Eschapasse, et al., 2022). Acute stress at retrieval can impair model-based behavior (Brown et al., 2020), but stress induced shortly before or during learning can enhance long-term memory (Schwabe et al., 2008; Smeets et al., 2007; Zoladz et al., 2011). Recently, Guenzel et al. (2014) applied stress to participants before they learned different forms of spatial memory. Results showed that during retrieval a week later, local-landmark-guided navigation performance was impaired in males, whereas route retrieval was impaired in females. These results suggest that stress may differentially impair different types of learning and that such effects may be moderated by gender. Yet beyond this work, remarkably limited research has explored the effects of stress on route-oriented (comparatively rigid) versus map-like (comparatively flexible) navigational learning or its implications for decision-making when one can use both forms of memory.

In the real world, it is common to learn specific routes in a new environment, which come to inform relational knowledge of the space—for example, when moving to a new home, we often build allocentric knowledge as we learn goal-directed routes between our home and local destinations. To our knowledge, no studies have investigated how stress affects flexible learning when it builds on the outcome of earlier route-based learning. The present study addressed such progressive learning under stress, in addition to stress’s impact on decision-making; our findings may also have implications for serial relationships between rigid and flexible elaborative learning that often occur in other settings.

Literature on decision-making suggests that stress changes people’s decision biases in deterministic versus probabilistic scenarios (Lighthall et al., 2009, 2012; Preston et al., 2007; van den Bos et al., 2009) and can decrease people’s sensitivity to the outcomes and rewards of their actions (Berghorst et al., 2013; Bogdan & Pizzagalli, 2006; Kumar et al., 2014). Relevant to navigational decision-making, declarative memory mechanisms support simulation or planning over routes (e.g., Bellmund et al., 2016; Brown et al., 2016), and when stressed at retrieval, representations of environment details for both upcoming novel and familiar routes are reduced or disrupted (Brown et al., 2020). However, despite the integral role of memory in decision-making (Behrens et al., 2007; Bornstein et al., 2017; Duncan & Shohamy, 2016; Hertwig & Erev, 2009; Hertwig et al., 2004; Murty et al., 2016), very few studies have directly investigated how stress modulates the way memory is used in decision-making. For the present study, these literatures suggest that psychological stress may reduce the extent of memories for different strategies that are accessed and integrated when we derive the value of different navigational options for decision-making.

Statement of Relevance

We often learn and make decisions in stressful situations. How acute stress affects learning and decision-making has been studied for decades, but few studies have investigated whether stress differentially affects rigid learning (i.e., repeating learned behaviors) and flexible learning (i.e., learning the structure of the task), and even fewer studies have investigated how stress affects flexible learning when it builds on the outcome of more rigid route-based learning. Furthermore, few studies have investigated how stress affects the way declarative memory from such learning experience informs decision-making about the costs or benefits of different choices. Taking advantage of virtual reality and computational models, we found that acute stress differently impacts more rigid and more flexible learning, and it modulates the communication between episodic memory and how we make decisions about navigation options. Together, our results paint a comprehensive picture of how learning, memory, and decision-making are impacted by acute stress in sequence.

To address the gaps in the literature, we asked participants in our study to perform a value-based decision-making navigation task. Value-based decision-making refers to the decisions that are not based on an objectively correct answer but instead depend on an individual’s idiosyncratic goals and preferences (Busemeyer et al., 2019). Here, participants experienced virtual environments from comparatively rigid route learning and then wayfinding (encouraging more flexible knowledge) while we manipulated concurrent psychological stress. Critically, they then performed value-based decision-making between these two forms of navigation, informed by potential costs and their episodic memories with the strategies.

“Value” in value-based decision-making is intimately tied to memory. We recently developed a computational model to capture the extent of memory integration in value-based decision-making (He, Liu, Beveridge, et al., 2022). This model (see Method section) captures each individual’s extent of memory integration in decision-making (e.g., are decisions based on rewards or costs of my most recent experiences, or do they account for more remote experience?). Furthermore, the model can differentiate between whether peoples’ decisions primarily draw on past experiences with the current goal or integrate information from different but related goals.

Given the evidence cited above that acute stress at retrieval impairs retrieval and the scope of route simulation (Brown et al., 2020; Quesada et al., 2012; Schönfeld et al., 2014; Schwabe & Wolf, 2014), we hypothesized that acute stress reduces the extent of memory integration in decision-making, leading people to rely on more recent episodic memory. In addition, given evidence that stress can impair the ability to generalize across past experiences when one is required to transfer knowledge to novel situations (Dandolo & Schwabe, 2016; Kluen et al., 2017; Schwabe et al., 2022), we hypothesized that stress makes people less likely to integrate memory from related goal-directed experiences. Furthermore, given some evidence that gender could increase risk-taking behaviors in males more than females (Lighthall et al., 2009, 2012; Preston et al., 2007; van den Bos et al., 2009), we also explored whether stress and gender might modulate risk-taking behaviors, which has been rarely investigated before in navigation (Maxim & Brown, 2023).

In sum, from the above literature, we hypothesized that stress would (a) impair comparatively rigid route learning in females but not in males (Guenzel et al., 2014) and (b) impair the subsequent flexible learning, regardless of gender. On the other hand, given the evidence cited above that concurrent stress can enhance hippocampal-dependent encoding, a clear alternative outcome was also tested: that flexible relational learning that builds on prior route knowledge (He, Liu, Beveridge, et al., 2022) may instead be enhanced. Critically, given that memory retrieval can be impaired by concurrent stress, we hypothesized that stress would reduce the extent of memory integration in decision-making. Finally, we explored whether stress increased risk-taking behaviors in our paradigm.

Open Practices Statement

The study reported in this article was not preregistered. The data and analysis code (in JASP) have been made publicly accessible via OSF and can be accessed at https://osf.io/tb9c6. The materials are widely available on the Internet.

Method

Participants

Eighty-five college students participated in this experiment for monetary compensation. Participants spent between 80 and 120 min completing the experiment. Three participants’ data were excluded because motion sickness made them unable to finish the experiment. As a result, 82 participants (42 females) were included in the data analysis, and the male-to-female ratios in the control and stress conditions were both 20:21. We based our sample size on that used by Guenzel et al. (2014) because their experimental design resembled ours the most. In particular, the effect size of stress on repeating a learned route was Cohen’s d = 0.66, which would require a sample size of 39 participants in each condition to reach a power of 80%. All participants (ages = 8–23 years) gave written and informed consent. The research was approved by the institutional review board (approval code: H18233). All procedures were performed in accordance with the institutional guidelines.

Materials and procedure

After participants signed the consent form, we reminded them that this experiment was a stress study and briefly explained the stress setup should they be assigned to that group. They confirmed verbally that they were willing to participate if assigned to the stress group. We did not tell participants their group assignment at this time to maintain a similar level of stress across both the control and stress groups.

To assess psychological stress responses from our participants, we leveraged our acute stress state questionnaire developed in Gagnon et al. (2019) and used in Brown et al. (2020). This questionnaire is a simplification of the State-Trait Anxiety Inventory (Spielberger et al., 1983), which asks participants for five distinct affective ratings on the same scale from 1 (least strong) to 4 (most strong) for the immediately preceding trials—“I feel happy,” “I feel safe,” “I feel anxious,” “I feel stressed,” and “I feel distracted.” Participants are introduced to the five ratings and asked to identify their affective state as best they can using the different categories. The benefit of this questionnaire is that it provides a momentary assessment of participants’ immediate state, which encourages them to think about how stressed they are psychologically (e.g., not to conflate feelings of distraction with stress). We collected this measure at the start of the experiment (before the first fixed-phase navigation trials, described below) and again at the end of the tasks. This provided an individual baseline measurement for their subjective stress levels, and the difference between these two measurements served as a psychological measurement of their stress response to the task. The first 24 participants, however, completed only the final questionnaire, so their uncorrected affective response differences are omitted from the results below.

We also collected data on the physiological impacts of stress: Salivary cortisol samples were collected (three or four per participant). Because these measures are indirect indicators of psychological stress (i.e., vary across diurnal cycle and in response to physical exertion, even consumption of coffee, and other hormonal factors that can differ between genders), we have typically found more robust associations between psychological stress indicators and behavioral outcomes; nevertheless, such salivary cortisol samples were collected to provide support that our measured stress responses were associated with hypothalamic-pituitary-adrenal axis activity. After participants provided consent, we collected the first salivary cortisol measurement (Time 1). Specifically, participants saturated a small cotton swab (Salimetrics cortisol duplicate testing) orally for a minute. Participants then completed questionnaires and cognitive tests that are not reported here, which took approximately 30 to 45 min. At this point, an additional salivary cortisol measurement (Time 2) was collected, after which each participant was randomly assigned to either the stress or the control group and asked to provide a second consent reflecting their willingness to continue after they knew their group assignment. Given the noise inherent in salivary cortisol, we therefore averaged Time 1 and Time 2 data to create a single pretreatment baseline measure to compare with the posttreatment cortisol level.

Participants assigned to the stress condition underwent preparation for the stress manipulation following the baseline stress measures and prior to the navigation task phases so we could manipulate psychological stress. This procedure follows those established by Gagnon et al. (2019) and Brown et al. (2020). We used a threat of shock to induce acute stress. Specifically, two electrodes from a Biopac (Goleta, CA) stimulator system (STM100C with constant current unit) were attached to the participant’s left ankle, which had been exfoliated to remove any oil that could affect the flow of current. The researcher worked with the participant to identify a shock level that corresponded with a 7 out of 10 pain level, with 10 being the highest level of pain. The participants were told that the shock would begin at 0.01 mA current, and the researcher would increase the current until the participant reached their 7 out of 10. In this way, each participant’s shock level was individualized and recorded. Six participants reached the maximum safe current from the stimulator system (50 mA). Participants were instructed that the shock would be delivered randomly during the navigation tasks and was not dependent on their actions or performance. The shock was coded to be pseudorandom, so that a participant would receive no more than six shocks over the course of the experiment to avoid habituation. The shock delivery began at the start of the fixed navigation phase (described below) and continued until the participant completed the rest of the navigation tasks (described below). Thirty minutes after stress-manipulation onset, or from Time 2 for control participants, a third cortisol measurement (Time 3) was collected. For a small subset of participants who took an additional 30 min after Time 3 to complete the navigational task, a fourth cortisol measurement (Time 4) was collected. Because the current study built on our previous study, the experimental procedure of the spatial navigation task (described below) was almost identical to the task used in that study (He, Liu, Beveridge, et al., 2022), except for electrical stimulation in the stress condition.

Navigation in a virtual environment

Before the navigation tasks of our experiment, all participants were familiarized with the control scheme and task objectives in a small practice virtual environment (4 rooms × 4 rooms presented with the Unity game engine; https://unity.com). Participants then performed a series of wayfinding and value-based decision-making tasks in a larger desktop virtual environment (Fig. 1), as described in detail in our other recent study (He, Liu, Beveridge, et al., 2022). In brief, participants navigated in a virtual environment consisting of 36 rooms (with walls between them) and two distal landmarks to provide global directional orientation. There were three experiment task phases, across which the participant’s goal was always to find three target objects (“T1” to “T3” in Fig. 1) repeatedly. In the fixed phase, participants always started in a fixed starting location (“S” in Fig. 1), and after finding a target object, they would be teleported back to the same starting location (finding one object per trial and the order was always fixed: apple, banana, watermelon, and then repeat). This fixed starting location enabled participants to follow previously learned routes to find the targets (akin to navigating from work to home), thus reflecting a form of comparatively rigid learning (note that this was not assumed in our study to equate to stimulus–response learning—indeed, sequencing routes, at least in early learning, are well known to tap hippocampal memory mechanisms and the associative striatum before such routes gradually become habitualized; Goodroe et al., 2018). In our study, participants could vary in the extent to which they acquired allocentric knowledge from these experiences, and this fact was reflected in our subquestions examining how stress affects performance in the next phase while factoring in this knowledge from the fixed phase, as well as being reflected in our approach to individualizing the reward costs in the final decision-making phase to relative performance in these initial phases. In the subsequent random phase, everything was the same as in the fixed phase except that the starting location and the object to be found (e.g., banana, apple, watermelon) were randomly selected. Here, random starting locations encouraged participants to learn the environment’s configuration holistically and more explicitly than in the fixed phase in order to achieve good wayfinding efficiency—thus reflecting a form of more flexible learning.

Fig. 1.

Virtual environment and procedure of the experiment (top row) and task flow of the decision-making phase (bottom row). In the room grid at the top right, “S” indicates the starting location in the fixed phase, and T1 to T3 indicate Targets 1 to 3. Electrical stimulation (yellow flash icons) was delivered in all three phases in the stress condition.

Both the fixed and random phases were limited to six trials each. The number of trials was intentionally kept small so that, on the one hand, participants got adequate exposure to each form of navigation in order to make decisions about them at the beginning of the subsequent decision-making phase (described more below and in detail in He, Liu, Beveridge, et al., 2022), but, on the other hand, participants’ behavior would not be characterized by stimulus–response mapping, which could dominate behavioral expression in the decision-making phase (described below); overlearning in the fixed and random phases could also limit the amount of behavioral changes and variability in the decision-making phase, limiting our ability to apply our computational model (He, Liu, Beveridge, et al., 2022) and quantify episodic memory integration and generalization in our decision-making task (and, critically for this article, how it is affected by stress). Instead, in our design, knowledge of the environment would continue to improve in the subsequent decision-making phase (and thus this improving knowledge could continue to influence participants’ decisions across trials). Because the fixed phase is comparatively more rigid and provides explicit encouragement to orient from a familiar location toward a familiar route, unlike the random phase, the decision in the decision-making phase was between a familiar orientation and route versus a random one, not between a habit and a declarative memory strategy.

Our design in the navigation phases (fixed and random) shared some similarities with the design by Guenzel et al. (2014), as stated in the introduction, but there were also several differences. First, stress was concurrent with learning in the current study, whereas stress was applied before learning in Guenzel et al. (2014). As laid out in the introduction, this is the expected result in potential enhancing affects of stress on encoding for declarative memory details of the task. Second, retrieval was contemporaneous with learning in the current study, whereas retrieval was 1 week after learning in Guenzel et al. (2014).

Third, the current project included flexible learning (random starting location), and the outcome of such learning can build on the outcome of the learning phase, which encourages relatively more rigid learning of specific routes to goals (fixed starting location). We were motivated by our recent study applying reinforcement-learning models to a large sample using the same overall paradigm reported here. In that study, reinforcement modeling provided formal evidence that people who learned routes better in our fixed-route learning phase exhibited more model-based behavior (defined by their fixed-phase experiences) when transitioning to the random phase, whereas people who struggled to learn routes initially failed to exhibit efficient model-based behavior later on (He, Liu, Eschapasse, et al., 2022). Because of this, although this was a secondary aim of the study, we drew on this clear evidence that flexible learning builds on more rigid route learning in our task to explore how stress affects subsequent flexible learning with and without controlling for the performance of the rigid learning.

Fourth, in our study, the fixed and random phases shared the same environment structure, objects, and goals. This difference was a critical feature for our primary aim of testing predictions about memory integration from prior learning under stress, and we explain the reasons for this difference below when describing the decision-making phase.

The first two differences will extend our understanding of the timing-dependent nature of the stress effects, and the third difference will reveal the chain effects of stress on learning that happen progressively (from route-based to flexible), which is often the case, from learning relationships between a new home and nearby destinations to learning sequentially in educational settings.

Decision-making phase

The decision-making phase was the critical task period for our primary study goals. In the decision-making phase, participants started with $20, and they were instructed that they would keep the remainder of this money (the amount depended on their decision-making) after completing 30 trials in this phase. In each decision-making trial, they were shown which target object (familiar from the preceding fixed and random phases) they needed to find on the next trial, and then they were presented with a choice of whether to find this object from the fixed starting location (a strategy and challenge now familiar from their experience with the fixed phase)—the fixed option—or to find this object from a random starting location, as in the random phase—the random option. Critically, different penalties (how much money was deducted for every second they spent in finding the object; Fig. 1) were associated with each option, such that the participant needed to consider the expected value of one choice versus the other in order to maximize their earnings. The penalty in the random option was always less than that in the fixed option, and as described in He, Liu, Beveridge, et al. (2022), the specific penalty ratio for the two options was defined by their relative performance in the previous fixed and random phases (e.g., an individual who struggled more with random would be presented with a relatively lighter monetary penalty for that option). This participant-derived penalty was static across decision-making trials and ensured that, mathematically, the relative value of each choice was similar across participants. All of this was designed to encourage participants to value each option carefully before choosing. After each decision, participants then navigated to the goal from the start position generated by their chosen strategy for that trial, before then proceeding to the next decision-making and navigation trial.

As mentioned above, the fixed and random phases shared environmental structure and goals in our design in service of this decision-making phase. This was done because, first, without shared goals between the fixed and random phases, it would not be possible for participants to have priors in the key decision-making phase between the two strategies experienced during the fixed and random phases—for a given goal, they would have no directly contrasting expected values of choosing strategy A versus B on the basis of their episodic memory of their attempts to reach this goal. Moreover, a lack of shared goals between the fixed and random phases would prevent us from calibrating the relative penalties between the two strategies in the decision-making phase to have experimental control over the relative value—indeed, people did vary in their degree of allocentric knowledge and path efficiency acquired in the fixed phase (something we quantified formally in analyses in our recent reinforcement-learning-centered study; He, Liu, Eschapasse, et al., 2022), which made it critical that we had these shared goals and thus gave participants the ability to calibrate the relative value of the options to their own learning (otherwise, for some individuals, the relative value may have dramatically favored one strategy over another in the decision-making phase, limiting the extent to which participants would benefit from deliberation).

Memory-Integration Models

The present study built on a computational model and related psychological findings recently developed and published by our group (He, Liu, Beveridge, et al., 2022). In this section, we briefly detail key features of this modeling approach. A rational decision-maker should choose the option on each decision-making trial that incurs the least loss of money. This expected value (cost) was the product of the penalty rate and expected wayfinding time for that trial. Because the penalty rate was given directly, we needed to infer only how participants estimated the wayfinding time in the next trial for both options to determine which option participants would choose. To this end, as also described in He, Liu, Beveridge, et al. (2022), we applied a one-dimensional Gaussian filter (https://docs.scipy.org/doc/scipy/reference/generated/scipy.ndimage.gaussian_filter1d.html) with sigma (the extent of temporal memory integration of previous performance) as a free parameter to each participant’s wayfinding time, separately for the fixed and random options. By using a Gaussian filter, we assumed that participants’ prediction of their future performance could be influenced by the recency of their episodic memory in a temporally graded manner (i.e., the more recent the memory, the higher the influence on prediction). The free parameter sigma reflected this temporal scope and extent of memory integration (i.e., the smaller the sigma, the heavier the influence of recent memory in predicting the future). Sigma was estimated using maximum likelihood estimation for each participant.

To illustrate how our computation models were applied in the current study, imagine that one participant spent 6 s, 5 s, 4 s, 3 s, 2 s, and 1 s (from remote to recent) to find an apple from a fixed starting location and spent 8 s, 5 s, 3 s, 2 s, 4 s, and 2 s (from remote to recent) to find it from a random starting location. If this participant’s sigma were estimated to be 1, then they would expect to spend 1.43 s to find the apple from the fixed starting location and 2.59 s to find it from a random starting location in the next trial. If the penalty were 1 cent per second in the fixed option and 0.5 cents per second in the random option, then the participant should choose the random option because the expected loss of the fixed option (−1 × 1.43 = −$1.43) is greater than for the random option (−0.5 × 2.59 = −$1.29). Note that the time spent in finding each object in each option (fixed or random), including the wayfinding time in the initial fixed and random phases, was continuously updated throughout the experiment to feed into the Gaussian filter—that is, the model continually generated new predictions for each decision trial, reflecting the participant’s evolving history with the task.

The above model features address the temporal extent of memory integration. Our computational models can also reveal whether participants’ decisions reflect generalization of their episodic memory across related sources. For example, when asked to find the apple, one could base their decision on the performance with the fixed and random experiences specific to apple, or they could base it on the performance with all targets (apple, banana, and watermelon). We named the former model of their decision-making “target specific” and the latter “target common.” Computationally, in the target-specific model, the Gaussian filter was applied only to the performance specific to the target object in the next trial, but it was applied to the performance of all target objects regardless of what the next target object was in the target-common model. We used the Bayesian information criterion (BIC) to determine which model fitted the data better for each participant.

Results

We used JASP (JASP Team, 2021) for statistical analyses and Matplotlib (Hunter, 2007), Seaborn, and Raincloud (Allen et al., 2019) for data visualization.

Manipulation check of acute psychological stress induction

We conducted a repeated measures analysis of variance (ANOVA) with stress condition (control vs. stress) and gender (male vs. female) as independent variables for explaining psychological stress ratings relative to baseline. We found a significant main effect of condition on stress levels before and after treatment, F(1, 54) = 4.81, p = .033, but no main effect of gender or interaction (ps > .24; Fig. 2a). We note that although participants showed significant psychological stress induction, they were not significantly more distracted, F(1, 54) = 1.69, p = .199, which could have provided a different account for the decision-making phase. The increase in stress was also reflected in divergent stress versus positive affect: The stress group exhibited significantly increased stress relative to positive affect from our questionnaire over the course of the experiment, F(1, 26) = 4.79, p = .038, with no interaction with gender, F(1, 26) = 2.80, p = .11. The control group showed no change in relative stress versus positive affect over the course of the experiment, F(1, 28) = 0.09, p = .77, with no interaction with gender, F(1, 28) = 0.09, p = .77. This comparison was done because certain levels of stress may be perceived as invigorating and in the right mindset can be associated with a positive mood and improved cognitive outcomes (Crum et al., 2017). Collectively, these data indicate that participants of both genders felt more stressed by the end the experiment in the stress condition than in the control condition.

Fig. 2.

Changes of stress level between condition and gender from baseline (“Pre”) to poststudy tasks (“Post”). Mean psychological stress level (a) is shown as a function of time point and condition, separately for males and females. Psychological stress levels increased across genders in the stress group. Mean cortisol level (b) is also shown as a function of time point and condition, separately for males and females. In males, baseline-corrected posttask cortisol was significantly higher in the stress group than in the control group, indicating that cortisol levels from individuals’ idiosyncratic baselines entering the study were comparatively more level following the stress manipulation, reducing the diurnal trend that was reflected significantly more strongly in the control group. Error bars represent standard error of the mean.

We also collected salivary cortisol measurements to complement the psychological stress ratings. The majority of the participants finished the experiment in less than 1 hr, and thus only 24 participants had cortisol measurement at Time 4. Therefore we excluded Time 4 from analysis. We conducted a repeated measures ANOVA with stress condition and gender as between-subjects variables and time (cortisol difference relative to baseline) as a within-subjects variable (Fig. 2b). There was a main effect of time, F(1, 75) = 12.95, p < .001, and a marginal three-way interaction indicating that gender influenced the divergence in cortisol values from baseline, F(1, 75) = 3.71, p = .058 (Time × Gender: p = .16, Time × Condition: p = .74). Unpacking this marginal interaction, we found that males in the stress condition had significantly greater baseline-corrected cortisol levels at Time 3 than males in the control condition, t(38) = 2.24, p = .03. This was the same pattern we observed in Gagnon et al. (2019), demonstrating that the stress manipulation significantly counteracts an effect of diurnal cycle/time on cortisol. By contrast, we did not detect elevated cortisol levels in females in the stress group relative to those in the control group, t(40) = −1.61, p = .12, although we reiterate that baseline-corrected psychological stress did not differ between female and male participants in the stress group (indeed, it was qualitatively highest).

The null effect from baseline in female cortisol levels was anticipated given that a female’s menstrual cycle can cause hormonal fluctuations that impact cortisol levels and cortisol changes to psychological stressors (although cortisol levels remain relevant to cognition). A review of the literature (Kudielka & Kirschbaum, 2005) found that women in the luteal phase have similar salivary cortisol responses to acute stressors as males do, but women in the follicular phase or those taking oral contraceptives have significantly reduced salivary cortisol responses to acute stressors. Because of this, we collected data a priori on menstrual phase and hormonal contraceptives use. In our study, only 26% (11 of 42) of females reported being in the luteal phase and not being on oral contraceptives. Purely qualitatively, when we compared females in the luteal menstrual phase (eight in the control group and three in the stress group), the salivary cortisol differences followed a similar pattern to that of males.

Recently two studies, Guenzel et al. (2014) and van Gerven et al. (2016), found equivalent cortisol changes in men and women following their stress manipulations. However, we do not view our results as incompatible with theirs. van Gerven et al. (2016) did not find a significant effect of cortisol in their stress manipulation regardless of gender and do not provide a breakdown of the cortisol affects according to gender (they do note that only eight of their 116 participants were on hormonal contraceptives, but they do not report on menstrual cycle stages of their female participants)—this leaves the question open as to whether they did or did not observe the same outcome as we did. Guenzel et al. (2014), on the other hand, restricted their recruitment to females who were not on oral contraceptives and who were not on their menses (although they did not assess the menstrual phase more granularly). They cite similar considerations as we do regarding contraceptive use and menstrual cycle and, given their exclusionary criteria, effectively report data from a female participant subgroup similar to ours.

In our study, males’ baseline-corrected cortisol levels correlated with their increase in subjective psychological stress from baseline (r = .39, p = .05) and with the increase in stress from baseline relative to subjective positive affect levels (r = .53, p = .007). By contrast, females’ increase in psychological stress did not track the response in cortisol to stress manipulation (as expected given the above considerations; r = −.06, p = .76; pre-post stress relative to positive affect: r = −.07, p = .73).

Different stress hormone levels at baseline could influence cognition as stress is manipulated (or not) in the subsequent tasks. We therefore controlled for preexperiment cortisol-level differences (as a covariate) in the following cognitive analyses. The statistical results without controlling for the baseline hormone level can be found in the Supplemental Material available online. Here, we simply note that although controlling for baseline cortisol broadly strengthened our outcomes, the patterns of results with and without controlling for baseline cortisol levels were qualitatively the same, and a formal interaction test revealed that controlling for basal cortisol did not change the relationships between gender and our dependent measures, which we report below.

Acute stress affects learning differently in more rigid and more flexible conditions

Prior to the decision-making phase, participants were forced to sample both navigational strategies, providing some spatial memory and a basis for decision-making on the first trial of the decision-making phase. Given that learning precedes the decision-making phase, we first asked whether acute stress affects learning differently under the rigid condition (routes from a fixed, predictable position) and subsequent flexible condition—and if so, how? Here, we used excessive distance as a measure of learning performance, which has been widely used in the spatial navigation literature (He et al., 2021; He, McNamara, Bodenheimer, & Klippel, 2019; He, McNamara, & Brown, 2019; Newman et al., 2007) and has been defined as follows: (actual traversed distance – optimal distance)/optimal distance. An excessive distance of 0 indicated perfect wayfinding performance (actual traversed distance = optimal distance), and an index of 1 indicated that the actual traversed distance was 100% longer than the optimal distance.

First, we ran a two-way ANOVA with baseline cortisol level as a covariate in the fixed phase. The main effect of condition was not significant, F(1, 75) = 2.31, p = .13, η² = .024, but the main effect of gender was significant, F(1, 75) = 13.87, p < .001, η² = .14. Importantly, the interaction between condition and gender was significant, F(1, 75) = 7.04, p = .01, η² = .07 (Fig. 3a). A post hoc test showed that stress did not affect male’s performance in the fixed phase, t(75) = 0.78, p = .44, Cohen’s d = 0.25, but impaired females’ performance significantly, t(78) = −2.97, p = .004, Cohen’s d = −0.94. We considered the possibility that this performance difference between conditions for females could be due to females who were randomly assigned to the stress condition coincidentally having worse general spatial ability than females who were randomly assigned to the control condition. To test this possibility, we compared the self-reported sense of direction, which has been shown to be correlated with spatial ability in both real-world and virtual environments (Hegarty et al., 2002; Pazzaglia & De Beni, 2001), between these participants: We found no significant difference, t(40) = −0.79, p = .43, Cohen’s d = −0.18.

Fig. 3.

Wayfinding performance in the (a) fixed and (b) random phases as a function of gender and condition. The whiskers from the box-and-whisker plots are drawn within the 1.5 interquartile range value. Each dot represents one participant. The kernel density of the violin plots is set to 1.

We then ran a two-way ANOVA (Stress Condition × Gender) with baseline cortisol level as a covariate in the random phase. The main effect of gender was significant, F(1, 75) = 10.37, p = .002, η² = .12—specifically, males performed better than females—but the main effect of condition and the interaction were not (ps > .12; Fig. 3b). Note that performance in the stress group was numerically better than in the control group. To further test that stress did not disrupt learning in the random phase, we ran a signed Bayesian independent-samples t test with the alternate hypothesis that learning outcome was better in the control condition than in the stress condition. The Bayes factor (BF₁₀) value was 0.11, suggesting moderate to strong evidence favoring the null hypothesis (Lee & Wagenmakers, 2014).

In the real world, we may often acquire map-like or route-invariant knowledge of a new environment after targeted learning of specific routes—for example, when moving to a new home, we often learn important trajectories between that location and destinations such as the local grocery store. Our paradigm mirrored this. Because similarity of spatial layout between environments heavily influenced transfer of learning (Newman et al., 2007) and our fixed and random phases occurred in the same environment, we reasoned that learning in the random phase was affected by learning in the fixed phase. As a result, it was possible that there was no detrimental effect of stress in the random phase for females (in contrast to the fixed phase) because female participants in the stress condition traversed more of the environment in the fixed phase and hence had more learning. As a result, the additional learning would carry over to the random phase and give the female participants in the stress condition an advantage, which offset the detrimental effects of stress. If this were the case, we should have observed a negative correlation between fixed-phase performance and random-phase performance (worse performance in the fixed phase leading to better performance in the random phase). However, this was not the case: Among female participants, the correlation between fixed- and random-phase performance was .54 in the control group and .53 in the stress group (for males, the correlations were .32 and .54, respectively.).

To investigate how stress affected learning in the random phase when the learning in the fixed phase was controlled for, we applied an ANOVA to random-phase performance with fixed-phase performance and baseline cortisol level as covariates. Strikingly, this demonstrated a significant main effect of condition, F(1, 74) = 6.67, p = .012, η² = .06, but no interaction between condition and gender, F(1, 74) = 2.21, p = .14, η² = .002, indicating that stress enhanced learning in the random phase in both genders when learning in the fixed phase was controlled for. These results further underscore that the lack of a detrimental effect of stress in the random phase, particularly in females who showed a detriment in the preceding phase, was not due to the additional learning in the fixed phase.

Acute stress reduces memory integration in decision-making

We next asked the key question: How does acute stress affect the way memory is used in decision-making? Following our previous model-fitting procedure (He, Liu, Beveridge, et al., 2022), we first fitted two models, namely the target-specific and target-common models, for each participant and compared which model had a better fit. We then extracted the estimated parameter sigma, which reflected the extent of memory integration, from the better performing model for each participant. We then ran an ANOVA model with sigma as the dependent variable (the larger the size of sigma, the larger the extent of memory integration or more weights are given to remote events), condition and gender as independent variables, and baseline cortisol level as a covariate.

The main effect of condition was significant, F(1, 75) = 14.10, p < .001, η² = .15, but not the main effect of gender, F(1, 75) = 0.35, p = .55, η² = .004, or the interaction, F(1, 75) = 0.25, p = .62, η² = .003 (Fig. 4a). These results showed that stress significantly reduced the extent of memory integration, regardless of gender, suggesting that participants focus more on recent episodic memory in decision-making. Because the extent of memory integration may be limited by individual working memory capacity (He et al., 2021), these results could suggest that participants who were randomly assigned to the stress condition coincidentally had lower working memory capacity than participants who were randomly assigned to the control condition. To test this possibility, we compared the working memory capacity (operation span; Draheim et al., 2018) between participants from the stress and control conditions, and we found no significant difference, t(77) = 0.28, p = .78, Cohen’s d = 0.063.

Fig. 4.

Effect of stress on (a) the extent of memory integration during decision-making and (b) decision time, separately for each gender and condition. The whiskers from the box-and-whisker plots are drawn within the 1.5 interquartile range value. Each dot represents one participant. The kernel density of the violin plots is set to 1.

We also found that decision time, which was the temporal interval between when the option was presented and a decision was made in the decision-making phase, was significantly reduced in the stress condition—main effect of condition: F(1, 75) = 7.93, p = .006, η² = .09—regardless of gender—interaction between condition and gender: F(1, 75) = 1.47, p = .23, η² = .02 (Fig. 4b). We also replicated our previous finding (He, Liu, Beveridge, et al., 2022) that sigma was positively correlated with decision time, r(82) = .36, p < .001. Combining the sigma and decision-time results together, we reasoned that the effect of stress on the extent of memory integration in decisions could be mediated by decision time. For example, stress decreased decision time, which did not give participants enough time to reflect on their past experience. To test this possibility, we ran a mediation analysis with condition as the predictor, decision time as the mediator, and sigma as the outcome.

We found that the direct effect (i.e., the effect of stress on sigma absent the decision time) was significant (z = −2.48, p = .013), the indirect effect (i.e., the effect of stress on sigma through decision time) was marginally significant (z = −1.92, p = .055), and the total effect (the combination of direct and indirect effects) was significant (z = −3.25, p = .001). If decision time were a complete mediator, then we should have observed a significant indirect effect with a nonsignificant direct effect. Therefore, the significant direct effect showed that although the effect of stress on sigma may have been partially modulated by decision time, the estimated causal relationship between stress and sigma could not be simply or solely accounted for by it.

Because stress reduced the temporal extent of memory integration, we further asked whether stress also reduced the generalization of memory integration. That is, did stress make people less likely to integrate memory from other sources to form decision-making? A comparison of the participants choosing the target-common versus target-specific model between the control and the stress conditions could answer this question. We reiterate that the target-common model suggests that people integrate memory from all targets regardless of the goal target, whereas the target-specific model suggests that people primarily use memory with respect to the goal target. If stress made people less likely to integrate memory from different sources, then we should observe that more participants’ decisions were better fitted by the target-specific model in the stress condition than in the control condition. To this end, we computed the percentage of participants whose decisions were better fitted by the target-specific model for the stress and conditions separately and conducted a χ² test of independence. Indeed, we found a higher percentage of participants whose decisions were better fitted by the target-specific model in the stress condition than in the control condition (stress: 29.2% vs. control: 9.7%), χ²(1, N = 82) = 4.97, p = .026.

In sum, these results suggest that stress reduced memory integration in decision-making, biasing peoples’ decisions more to outcomes of recent memory and making people less likely to integrate memory from different sources.

Acute stress does not appear to have affected risk-taking behavior in the current paradigm

Last, we explored whether stress changed participants’ risk-taking behaviors and whether this change was modulated by gender. Because the random option in the decision-making phase appeared as a higher-risk, higher-reward option compared with the fixed option (although note that the underlying penalties were calibrated to learning in the fixed and random phases to ensure that a rational decision-maker could compute the relative risk or reward for the two options for a given goal), we used the overall percentage of a participant choosing the random option in the decision-making task as an index for risk-taking behavior. Here, none of the main effects (condition or gender) or the interaction were significant (ps > .33; Fig. 5). However, we considered the possibility that the lack of significant difference may carry over from participants’ individual differences in the performance of the fixed and random phases (the ratio between these two performances) and the random-option penalty. We therefore included these two factors as covariates in an analysis of covariance, but we still found no significant main effect of condition, F(1, 75) = 0.62, p = .435, η² = .008, or interaction, F(1, 75) = 0.55, p = .46, η² = .007. Although we did not find evidence in this exploratory analysis for stress modulating an indicator of riskiness in our task, future studies could revisit this question by, for example, altering the calibration of the relative costs of the two strategies to participants’ abilities and thus rendering the expected costs harder for participants to compute or perhaps more imbalanced according to the individual’s ability with the two options.

Fig. 5.

Effect of stress on the percentage of participants who chose the risky option (the random option) in decision-making, separately for each gender and condition. The whiskers from the box-and-whisker plots are drawn within the 1.5 interquartile range value. Each dot represents one participant. The kernel density of the violin plots is set to 1.

Discussion

To our knowledge, this is the first study to collectively investigate how acute psychological stress affects (a) the way memory from previous learning experiences is used in value-based decision-making in spatial navigation and (b) learning under more rigid and predictable circumstances versus flexible learning performance. We found that acute stress impairs rigid learning in females, but it does not impair flexible learning. Using a computational approach, we found that acute stress reduced memory integration in decision-making in both genders, associated with participants (a) giving more weight to recent episodic memory and (b) being less likely to include memory from other sources to inform decision-making. Last, we found no evidence that acute stress changed risk-taking behavior in decision-making in the scenario presented by our paradigm.

The major research question of the current study was how stress affects translation of episodic memory into value-based decision-making. Mirroring neural evidence for a restricted scope of prospective thought/memory retrieval in Brown et al. (2020), our results revealed that stress reduced both (a) the temporal extent and (b) generalization of memory integration in value-based decision-making, regardless of gender. Our findings make three important contributions to the literature regarding stress and decision-making. First, these outcomes suggest that people’s decisions under stress can become more reactionary: overweighting recent experience, when rewards are better optimized in our task by integrating over one’s broader history with a strategy. But it also renders values of strategies less generalized from experiences with other goals, which can help optimize decisions for a particular location (He, Liu, Beveridge, et al., 2022). Second, using a naturalistic spatial navigation paradigm is rare in the decision-making literature and provides important ecological validity. Finally, our study reveals effects of concurrent stressors on value-based decision-making, when stress is often induced in adjoining periods (e.g., before, using the Trier Social Stress Test).

Because acute stress induced at retrieval impairs memory (Gagnon et al., 2019; Quesada et al., 2012; Schönfeld et al., 2014; Schwabe & Wolf, 2014), this may explain the reduction of the extent of memory integration that we observed by limiting access to a broad range of episodes for on-line integration into a decision. Another possibility is that stress changes our tendency from relying on remote episodic memory to recent ones (i.e., to the extent that stress disrupts cognitive control systems, the effect may emerge from disrupted strategic engagement of executive resources for memory search and behavior selection). It is also unclear why people are more likely to use information from a single source for decision-making under stress. Stress can impair the ability to generalize across past memory when people are required to transfer memory to novel situations (Dandolo & Schwabe, 2016; Kluen et al., 2017; Schwabe et al., 2022). Another possibility is that, as with temporal scope, stress changes our strategy for integrating information from different sources. Last, because encoding and retrieval processes are interleaved by the nature of such navigational experiences, it is important to consider that both the reduced integration extent and source generalization outcomes could be altered if stress changes integrative processes at the time of encoding each episode (e.g., integrative encoding mechanisms; Shohamy & Wagner, 2008). Our findings are thus highly generative for future studies examining such possibilities, and it will also be of great interest to test generalization in groups other than our sample composition (e.g., older adults).

An interesting aspect of our design is that participants were given initial memories with the two navigational strategies in series prior to the decision-making phase. This enabled us to examine how stress affects performance on sequential acquisition of the route knowledge and more model-based knowledge that the participants were later required to choose between during memory-guided decision-making. As summarized above, we show that stress impairs route learning in more rigid circumstances (at least in females) but does not impair flexible learning. This finding is in line with one of the key findings in Guenzel et al. (2014). However, one key difference in the experimental procedure between our two studies is the timing of stress application and the temporal interval between learning and retrieval. In this sense, we extend Guenzel et al.’s (2014) finding by showing that concurrent acute stress impairs more rigid learning in females. This is important because when people learn new routes over repetitions, encoding and retrieval mechanisms may often co-occur, interacting to (a) guide behavior with what we remember from the prior experience and (b) update, elaborate, or correct the route memory from the current experience as it unfolds. Although stress can facilitate declarative memory encoding and rigid learning in some contexts, our findings contribute evidence to the literature that stress timing and duration, along with gender, may factor into disruption of route acquisition (Schwabe et al., 2010).

In turn, psychological stress did not impair learning in the random phase, and we show that this was very unlikely the result of stress adaptation or additional exploration in the fixed phase. Instead, one explanation is that although stress could shift learning or retrieval strategies from flexible to familiar or habitual (Brown et al., 2020; Park et al., 2017; Schwabe et al., 2007; Schwabe & Wolf, 2011), it does not necessarily impair flexible learning processes themselves. A recent study showed that acute stress led to faster extraction of probability-based regularities and did not impair extraction of serial-order-based regularities (Tóth-Fáber et al., 2021), suggesting that acute stress may even enhance flexible learning under some circumstances. Indeed, we found that stress enhanced flexible learning when learning achieved from the preceding fixed-route phase was controlled for. Understanding how stress affects different types of flexible learning has important and widespread translational and theoretical implications in many settings. Given our findings, an interesting question for future studies will be to what extent this stress-related enhancement depends on having engaged in fixed-phase learning. Given that our prior work provides evidence that random-phase learning can build on route knowledge acquired in the fixed phase (He, Liu, Eschapasse, et al., 2022), we hypothesize that stress may be differentially beneficial for flexible learning when it builds on a schema from prior learning (as in our study) than when flexible learning must be engaged in de novo.

In summary, using a naturalistic spatial navigation paradigm, we found that concurrent acute stress impairs rigid learning (at least in females), and it does not impair flexible learning. Acute stress reduces memory integration, biasing people on recent episodic memory and making them more likely to rely on information from a single source in value-based decision making. Together, our results paint a comprehensive picture of how learning, memory, and decision-making are impacted by acute stress in sequence.

Supplemental Material

sj-docx-1-pss-10.1177_09567976231155870 – Supplemental material for Effects of Acute Stress on Rigid Learning, Flexible Learning, and Value-Based Decision-Making in Spatial Navigation

Supplemental material, sj-docx-1-pss-10.1177_09567976231155870 for Effects of Acute Stress on Rigid Learning, Flexible Learning, and Value-Based Decision-Making in Spatial Navigation by Qiliang He, Elizabeth H. Beveridge, Vanesa Vargas, Ashley Salen and Thackery I. Brown in Psychological Science

Footnotes

Acknowledgements

Q. He and E. H. Beveridge contributed equally to this research.

Transparency

Action Editor: Daniela Schiller

Editor: Patricia J. Bauer

Author Contributions

Qiliang He: Conceptualization; Formal analysis; Methodology; Project administration; Supervision; Visualization; Writing – original draft; Writing – review & editing.

Elizabeth H. Beveridge: Conceptualization; Data curation; Investigation; Writing – review & editing.

Vanesa Vargas: Data curation; Investigation

Ashley Salen: Data curation; Investigation.

Thackery I. Brown: Conceptualization; Funding acquisition; Supervision; Writing – review & editing.

ORCID iD

Qiliang He

Supplemental Material

Additional supporting information can be found at

References

Allen

Poggiali

Whitaker

Marshall

T. R.

Kievit

R. A.

(2019). Raincloud plots: A multi-platform tool for robust data visualization. Wellcome Open Research, 4, Article 63. https://doi.org/10.12688/wellcomeopenres.15191.1

Anggraini

Glasauer

Wunderlich

(2018). Neural signatures of reinforcement learning correlate with strategy adoption during spatial navigation. Scientific Reports, 8(1), Article 10110. https://doi.org/10.1038/s41598-018-28241-z

Behrens

T. E. J.

Woolrich

M. W.

Walton

M. E.

Rushworth

M. F. S.

(2007). Learning the value of information in an uncertain world. Nature Neuroscience, 10(9), Article 9. https://doi.org/10.1038/nn1954

Bellmund

J. L.

Deuker

Navarro Schröder

Doeller

C. F.

(2016). Grid-cell representations in mental simulation. eLife, 5, Article e17089. https://doi.org/10.7554/eLife.17089

Berghorst

Bogdan

Frank

Pizzagalli

(2013). Acute stress selectively reduces reward sensitivity. Frontiers in Human Neuroscience, 7, Article 133. https://doi.org/10.3389/fnhum.2013.00133

Bogdan

Pizzagalli

D. A.

(2006). Acute stress reduces reward responsiveness: Implications for depression. Biological Psychiatry, 60(10), 1147–1154. https://doi.org/10.1016/j.biopsych.2006.03.037

Bornstein

A. M.

Khaw

M. W.

Shohamy

Daw

N. D.

(2017). Reminders of past choices bias decisions for reward in humans. Nature Communications, 8, Article 15958. https://doi.org/10.1038/ncomms15958

Brown

T. I.

Carr

V. A.

LaRocque

K. F.

Favila

S. E.

Gordon

A. M.

Bowles

Bailenson

J. N.

Wagner

A. D.

(2016). Prospective representation of navigational goals in the human hippocampus. Science, 352(6291), 1323–1326. https://doi.org/10.1126/science.aaf0784

Brown

T. I.

Gagnon

S. A.

Wagner

A. D.

(2020). Stress disrupts human hippocampal-prefrontal function during prospective spatial navigation and hinders flexible behavior. Current Biology, 30(10), 1821–1833. https://doi.org/10.1016/j.cub.2020.03.006

10.

Brunyé

T. T.

Wood

M. D.

Houck

L. A.

Taylor

H. A.

(2016). The path more travelled: Time pressure increases reliance on familiar route-based strategies during navigation. The Quarterly Journal of Experimental Psychology, 70(8), 1439–1452. https://doi.org/10.1080/17470218.2016.1187637

11.

Buchanan

T. W.

Tranel

Adolphs

(2006). Impaired memory retrieval correlates with individual differences in cortisol response but not autonomic response. Learning & Memory, 13(3), 382–387. https://doi.org/10.1101/lm.206306

12.

Busemeyer

J. R.

Gluth

Rieskamp

Turner

B. M.

(2019). Cognitive and neural bases of multi-attribute, multi-alternative, value-based decisions. Trends in Cognitive Sciences, 23(3), 251–263. https://doi.org/10.1016/j.tics.2018.12.003

13.

Collins

A. G. E.

(2019). Reinforcement learning: Bringing together computation and cognition. Current Opinion in Behavioral Sciences, 29, 63–68. https://doi.org/10.1016/j.cobeha.2019.04.011

14.

Crum

A. J.

Akinola

Martin

Fath

(2017). The role of stress mindset in shaping cognitive, emotional, and physiological responses to challenging and threatening stress. Anxiety, Stress, & Coping, 30(4), 379–395. https://doi.org/10.1080/10615806.2016.1275585

15.

Dandolo

L. C.

Schwabe

(2016). Stress-induced cortisol hampers memory generalization. Learning & Memory, 23(12), 679–683. https://doi.org/10.1101/lm.042929.116

16.

Draheim

Harrison

T. L.

Embretson

S. E.

Engle

R. W.

(2018). What item response theory can tell us about the complex span tasks. Psychological Assessment, 30(1), 116–129. https://doi.org/10.1037/pas0000444

17.

Duncan

K. D.

Shohamy

(2016). Memory states influence value-based decisions. Journal of Experimental Psychology: General, 145(11), 1420–1426. https://doi.org/10.1037/xge0000231

18.

Eckstein

M. K.

Wilbrecht

Collins

A. G.

(2021). What do reinforcement learning models measure? Interpreting model parameters in cognition and neuroscience. Current Opinion in Behavioral Sciences, 41, 128–137. https://doi.org/10.1016/j.cobeha.2021.06.004

19.

Elzinga

B. M.

Bakker

Bremner

J. D.

(2005). Stress-induced cortisol elevations are associated with impaired delayed, but not immediate recall. Psychiatry Research, 134(3), 211–223. https://doi.org/10.1016/j.psychres.2004.11.007

20.

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, 1369(1), 55–75. https://doi.org/10.1111/nyas.12996

21.

Gagnon

S. A.

Waskom

M. L.

Brown

T. I.

Wagner

A. D.

(2019). Stress impairs episodic retrieval by disrupting hippocampal and cortical mechanisms of remembering. Cerebral Cortex, 29(7), 2947–2964. https://doi.org/10.1093/cercor/bhy162

22.

Goodroe

S. C.

Starnes

Brown

T. I.

(2018). The complex nature of hippocampal-striatal interactions in spatial navigation. Frontiers in Human Neuroscience, 12, Article 250. https://doi.org/10.3389/fnhum.2018.00250

23.

Guenzel

F. M.

Wolf

O. T.

Schwabe

(2013). Stress disrupts response memory retrieval. Psychoneuroendocrinology, 38(8), 1460–1465. https://doi.org/10.1016/j.psyneuen.2012.12.010

24.

Guenzel

F. M.

Wolf

O. T.

Schwabe

(2014). Sex differences in stress effects on response and spatial memory formation. Neurobiology of Learning and Memory, 109, 46–55. https://doi.org/10.1016/j.nlm.2013.11.020

25.

Han

A. T.

Churaman

T. A.

Brown

T. I.

(2021). The role of working memory capacity in spatial learning depends on spatial information integration difficulty in the environment. Journal of Experimental Psychology: General, 150(4), 666–685. https://doi.org/10.1037/xge0000972

26.

Liu

J. L.

Beveridge

E. H.

Eschapasse

Vargas

Brown

T. I.

(2022). Episodic memory integration shapes value-based decision-making in spatial navigation. Journal of Experimental Psychology: Learning, Memory, and Cognition, 48(8), 1098–1109. https://doi.org/10.1037/xlm0001133

27.

Liu

J. L.

Eschapasse

Beveridge

E. H.

Brown

T. I.

(2022). A comparison of reinforcement learning models of human spatial navigation. Scientific Reports, 12, Article 13923. https://doi.org/10.1038/s41598-022-18245-1

28.

McNamara

T. P.

Bodenheimer

Klippel

(2019). Acquisition and transfer of spatial knowledge during wayfinding. Journal of Experimental Psychology: Learning, Memory, and Cognition, 45(8), 1364–1386. https://doi.org/10.1037/xlm0000654

29.

McNamara

T. P.

Brown

T. I.

(2019). Manipulating the visibility of barriers to improve spatial navigation efficiency and cognitive mapping. Scientific Reports, 9, Article 11567. https://doi.org/10.1038/s41598-019-48098-0

30.

Hegarty

Richardson

A. E.

Montello

D. R.

Lovelace

Subbiah

(2002). Development of a self-report measure of environmental spatial ability. Intelligence, 30(5), 425–447. https://doi.org/10.1016/S0160-2896(02)00116-2

31.

Hertwig

Barron

Weber

E. U.

Erev

(2004). Decisions from experience and the effect of rare events in risky choice. Psychological Science, 15(8), 534–539. https://doi.org/10.1111/j.0956-7976.2004.00715.x

32.

Hertwig

Erev

(2009). The description–experience gap in risky choice. Trends in Cognitive Sciences, 13(12), 517–523. https://doi.org/10.1016/j.tics.2009.09.004

33.

Hunter

J. D.

(2007). Matplotlib: A 2D graphics environment. Computing in Science Engineering, 9(3), 90–95. https://doi.org/10.1109/MCSE.2007.55

34.

JASP Team. (2021). JASP [Computer software]. https://jasp-stats.org/

35.

Kluen

L. M.

Agorastos

Wiedemann

Schwabe

(2017). Noradrenergic stimulation impairs memory generalization in women. Journal of Cognitive Neuroscience, 29(7), 1279–1291. https://doi.org/10.1162/jocn_a_01118

36.

Kudielka

B. M.

Kirschbaum

(2005). Sex differences in HPA axis responses to stress: A review. Biological Psychology, 69(1), 113–132. https://doi.org/10.1016/j.biopsycho.2004.11.009

37.

Kuhlmann

Piel

Wolf

O. T.

(2005). Impaired memory retrieval after psychosocial stress in healthy young men. The Journal of Neuroscience, 25(11), 2977–2982. https://doi.org/10.1523/JNEUROSCI.5139-04.2005

38.

Kumar

Berghorst

L. H.

Nickerson

L. D.

Dutra

S. J.

Goer

F. K.

Greve

D. N.

Pizzagalli

D. A.

(2014). Differential effects of acute stress on anticipatory and consummatory phases of reward processing. Neuroscience, 266, 1–12. https://doi.org/10.1016/j.neuroscience.2014.01.058

39.

Lee

M. D.

Wagenmakers

E.-J.

(2014). Bayesian cognitive modeling: A practical course. Cambridge University Press. https://doi.org/10.1017/CBO9781139087759

40.

Lighthall

N. R.

Mather

Gorlick

M. A.

(2009). Acute stress increases sex differences in risk seeking in the Balloon Analogue Risk Task. PLOS ONE, 4(7), Article e6002. https://doi.org/10.1371/journal.pone.0006002

41.

Lighthall

N. R.

Sakaki

Vasunilashorn

Nga

Somayajula

Chen

E. Y.

Samii

Mather

(2012). Gender differences in reward-related decision processing under stress. Social Cognitive and Affective Neuroscience, 7(4), 476–484. https://doi.org/10.1093/scan/nsr026

42.

Lockwood

P. L.

Klein-Flügge

M. C.

(2021). Computational modelling of social cognition and behaviour—A reinforcement learning primer. Social Cognitive and Affective Neuroscience, 16(8), 761–771. https://doi.org/10.1093/scan/nsaa040

43.

Maxim

Brown

T. I.

(2023). Toward an understanding of cognitive mapping ability through manipulations and measurement of schemas and stress. Topics in Cognitive Science, 15(1), 75–101. https://doi.org/10.1111/tops.12576

44.

Murty

V. P.

FeldmanHall

Hunter

L. E.

Phelps

E. A.

Davachi

(2016). Episodic memories predict adaptive value-based decision-making. Journal of Experimental Psychology: General, 145(5), 548–558. https://doi.org/10.1037/xge0000158

45.

Newman

E. L.

Caplan

J. B.

Kirschen

M. P.

Korolev

I. O.

Sekuler

Kahana

M. J.

(2007). Learning your way around town: How virtual taxicab drivers learn to use both layout and landmark information. Cognition, 104(2), 231–253. https://doi.org/10.1016/j.cognition.2006.05.013

46.

Park

Lee

Chey

(2017). Stress enhances model-free reinforcement learning only after negative outcome. PLOS ONE, 12(7), Article e0180588. https://doi.org/10.1371/journal.pone.0180588

47.

Pazzaglia

De Beni

(2001). Strategies of processing spatial information in survey and landmark-centred individuals. European Journal of Cognitive Psychology, 13(4), 493–508. https://doi.org/10.1080/09541440125778

48.

Porcelli

A. J.

Delgado

M. R.

(2017). Stress and decision making: Effects on valuation, learning, and risk-taking. Current Opinion in Behavioral Sciences, 14, 33–39. https://doi.org/10.1016/j.cobeha.2016.11.015

49.

Preston

S. D.

Buchanan

T. W.

Stansfield

R. B.

Bechara

(2007). Effects of anticipatory stress on decision making in a gambling task. Behavioral Neuroscience, 121(2), 257–263. https://doi.org/10.1037/0735-7044.121.2.257

50.

Quesada

A. A.

Wiemers

U. S.

Schoofs

Wolf

O. T.

(2012). Psychosocial stress exposure impairs memory retrieval in children. Psychoneuroendocrinology, 37(1), 125–136. https://doi.org/10.1016/j.psyneuen.2011.05.013

51.

Raio

C. M.

Konova

A. B.

Otto

A. R.

(2020). Trait impulsivity and acute stress interact to influence choice and decision speed during multi-stage decision-making. Scientific Reports, 10, Article 7754. https://doi.org/10.1038/s41598-020-64540-0

52.

Schönfeld

Ackermann

Schwabe

(2014). Remembering under stress: Different roles of autonomic arousal and glucocorticoids in memory retrieval. Psychoneuroendocrinology, 39, 249–256. https://doi.org/10.1016/j.psyneuen.2013.09.020

53.

Schwabe

Bohringer

Chatterjee

Schachinger

(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(1), 44–53. https://doi.org/10.1016/j.nlm.2008.02.002

54.

Schwabe

Hermans

E. J.

Joëls

Roozendaal

(2022). Mechanisms of memory under stress. Neuron, 110(9), 1450–1467. https://doi.org/10.1016/j.neuron.2022.02.020

55.

Schwabe

Oitzl

M. S.

Philippsen

Richter

Bohringer

Wippich

Schachinger

(2007). Stress modulates the use of spatial versus stimulus-response learning strategies in humans. Learning & Memory, 14(1–2), 109–116. https://doi.org/10.1101/lm.435807

56.

Schwabe

Wolf

O. T.

(2011). Stress increases behavioral resistance to extinction. Psychoneuroendocrinology, 36(9), 1287–1293. https://doi.org/10.1016/j.psyneuen.2011.02.002

57.

Schwabe

Wolf

O. T.

(2014). Timing matters: Temporal dynamics of stress effects on memory retrieval. Cognitive, Affective & Behavioral Neuroscience, 14(3), 1041–1048. https://doi.org/10.3758/s13415-014-0256-0

58.

Schwabe

Wolf

O. T.

Oitzl

M. S.

(2010). Memory formation under stress: Quantity and quality. Neuroscience & Biobehavioral Reviews, 34(4), 584–591. https://doi.org/10.1016/j.neubiorev.2009.11.015

59.

Shohamy

Wagner

A. D.

(2008). Integrating memories in the human brain: Hippocampal-midbrain encoding of overlapping events. Neuron, 60(2), 378–389. https://doi.org/10.1016/j.neuron.2008.09.023

60.

Smeets

Giesbrecht

Jelicic

Merckelbach

(2007). Context-dependent enhancement of declarative memory performance following acute psychosocial stress. Biological Psychology, 76(1–2), 116–123. https://doi.org/10.1016/j.biopsycho.2007.07.001

61.

Spielberger

C. D.

Gorsuch

R. L.

Lushene

Vagg

P.R.

Jacobs

G. A.

(1983). Manual for the state-trait anxiety inventory. Consulting Psychologists Press.

62.

Tóth-Fáber

Janacsek

Szőllősi

Á.

Kéri

Nemeth

(2021). Regularity detection under stress: Faster extraction of probability-based regularities. PLOS ONE, 16(6), Article e0253123. https://doi.org/10.1371/journal.pone.0253123

63.

van den Bos

Harteveld

Stoop

. (2009). Stress and decision-making in humans: Performance is related to cortisol reactivity, albeit differently in men and women. Psychoneuroendocrinology, 34(10), 1449–1458. https://doi.org/10.1016/j.psyneuen.2009.04.016

64.

van Gerven

D. J. H.

Ferguson

Skelton

R. W

. (2016). Acute stress switches spatial navigation strategy from egocentric to allocentric in a virtual Morris water maze. Neurobiology of Learning and Memory, 132, 29–39. https://doi.org/10.1016/j.nlm.2016.05.003

65.

Vogel

Schwabe

(2016). Learning and memory under stress: Implications for the classroom. NPJ Science of Learning, 1, Article 16011. https://doi.org/10.1038/npjscilearn.2016.11

66.

Zoladz

P. R.

Clark

Warnecke

Smith

Tabar

Talbot

J. N.

(2011). Pre-learning stress differentially affects long-term memory for emotional words, depending on temporal proximity to the learning experience. Physiology & Behavior, 103(5), 467–476. https://doi.org/10.1016/j.physbeh.2011.01.016

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.01 MB