Threat-induced modulation of hippocampal and striatal memory systems during navigation of a virtual environment

Abstract

The brain is composed of multiple memory systems that mediate distinct types of navigation. The hippocampus is important for encoding and retrieving allocentric spatial cognitive maps, while the dorsal striatum mediates procedural memories based on stimulus-response (S-R) associations. These memory systems are differentially affected by emotional arousal. In particular, rodent studies show that stress typically impairs hippocampal spatial memory while it spares or sometimes enhances striatal S-R memory. The influence of emotional arousal on these separate navigational memory systems has received less attention in human subjects. We investigated the effect of dynamic changes in anticipatory anxiety on hippocampal spatial and dorsal striatal S-R memory systems while participants attempted to solve a virtual eight-arm radial maze. In Experiment 1, participants completed a hippocampus-dependent spatial version of the eight-arm radial maze that required allocentric spatial memory to successfully navigate the environment. In Experiment 2, participants completed a dorsal striatal S-R version of the maze where no allocentric spatial cues were present, requiring the use of S-R navigation. Anticipatory anxiety was modulated via threat of receiving an unpleasant electrical shock to the wrist during memory retrieval. Results showed that threat of shock was associated with more errors and increased use of non-spatial navigational strategies in the hippocampal spatial task, but did not influence memory performance in the striatal S-R task. Findings indicate a dissociation regarding the influence of anticipatory anxiety on memory systems that has implications for understanding how fear and anxiety contribute to memory-related symptoms in human psychopathologies.

1. Introduction

Extensive evidence from maze learning studies has pointed to the existence of at least two interacting memory systems important for navigation, one principally mediated by the hippocampus and the other by the dorsal striatum (i.e. caudate nucleus and putamen; Packard, Hirsh, & White, 1989; Packard & McGaugh, 1996; Iaria, Petrides, Dagher, Pike, & Bohbot, 2003; for review, see Packard & Goodman, 2013). The hippocampus is responsible for encoding and retrieval of mental representations of space, which may be used for allocentric navigation (i.e. using distal cues in the environment to navigate toward a goal), whereas the dorsal striatum mediates procedural and egocentric stimulus-response (S-R) strategies, such as a series of motor behaviors in response to a stimulus (O’Keefe & Nadel, 1978; White & McDonald, 2002; Squire, 2004). It has been suggested that differential effects of emotional arousal on these memory systems (Packard & Goodman, 2012; Schwabe, 2013) may be useful for understanding how stress and anxiety alter the function of neural systems underlying symptoms of psychopathology, such as addiction or posttraumatic stress disorder (White, 1996; Goodman et al., 2012, 2014; Goodman and Packard, 2015a, 2016). However, the dynamic influence of emotional arousal on these separate navigational memory systems is poorly understood, especially in humans. In the present study we dynamically manipulated (i.e., trial-by-trial) phasic increases in anticipatory anxiety in human volunteers during navigation in a hippocampal spatial task or a striatal S-R task.

Previous studies employing both laboratory animals and human subjects have shown that high levels of tonic stress impair encoding and retrieval of hippocampal spatial memory, but facilitate encoding and retrieval of striatal S-R memory (Wingard & Packard, 2008; Leong et al., 2012, 2015; Vogel & Schwabe, 2018). Moreover, in learning situations where either a spatial or S-R strategy could be employed to solve a task, stress promotes the use of a striatal S-R strategy over a hippocampal spatial strategy (Packard & Wingard, 2004; Schwabe et al., 2007; Schwabe, Dalm, Schächinger, & Oitzl, 2008; Schwabe, Haddad, & Schachinger, 2008; Taylor et al., 2014; Goldfarb, Shields, Daw, Slavich, & Phelps, 2017; for reviews, see Schwabe, 2013; Goldfarb & Phelps, 2017). These differential effects of stress on separate memory systems has important clinical implications. For example, as striatal S-R memory is believed to contribute some of the harmful maladaptive habits observed in various psychiatric disorders (Schwabe, 2013; Packard, Goodman, & Ressler, 2018), the observation that stress induces a shift from a flexible ‘cognitive’ system to a habit-based memory system may serve as an experimental model of increased negative habitual activities (e.g., addiction) following stressful life events (e.g., Bruns & Geist, 1984; de Silva & Marks, 1999).

The procedures that have been used to examine the emotional modulation of memory systems in human participants, such as the cold pressor test and drug administration, lead to sustained (tonic) increases in subjective and autonomic arousal with the potential to influence memory up to an hour following the advent of the manipulation (Schwabe et al., 2007; Schwabe & Wolf, 2009; Schwabe, Tegenthoff, Höffken, & Wolf, 2010). While these procedures are useful for modeling the function of memory systems following a stressful event, they do not always produce an emotional state consistent with the worry, unease, and uncertainty typically associated with dynamic modulation of anticipatory anxiety. In addition, the sustained increase in emotional arousal following typical stress-induction protocols is not conducive to within-subjects designs, and confounds the ability to quickly manipulate the level of emotional arousal from trial to trial. One technique that has been employed to elicit transient increases in anticipatory anxiety is the “threat of shock” protocol (Schmitz & Grillon, 2012). In this protocol, a cue signaling an unpredictable shock (i.e., a shock that may occur at any time) effectively leads to higher subjective and physiological levels of anticipatory anxiety, as well as phasic impairments in a variety of cognitive processes, including memory (for review, see Robinson, Vytal, Cornwell, & Grillon, 2013).

The present study adapted the threat of shock protocol to examine the influence of dynamic phasic anticipatory anxiety on separate navigational memory systems. Two virtual eight-arm radial maze tasks were employed, one involving spatial memory (Experiment 1) and the other involving S-R memory (Experiment 2). As previous evidence in rodents has indicated that emotional arousal has the potential to influence memory systems during retrieval (e.g., Roozendaal, Griffith, Buranday, de Quervain, & McGaugh, 2003; Elliott & Packard, 2008), the present study specifically examined the influence of anticipatory anxiety during the retrieval phase in each navigation task. It was predicted based on prior research (Packard & Goodman, 2012) that anticipatory anxiety would impair hippocampal spatial memory retrieval and enhance dorsal striatal S-R memory retrieval, while also increasing the use of dorsal striatal navigational strategies in situations where either a spatial or S-R strategy could be employed.

2. Method

2.1. Participants

Sixty-two healthy adult participants (mean age = 23, SD = 6.327; 32 women and 30 men) each took part in one of the two experiments described below. A similar number of men and women were recruited to examine potential sex differences (e.g., Guenzel, Wolf, & Schwabe, 2014a). Exclusionary criteria included a history of psychiatric, neurological, or major medical illnesses or current use of psychoactive medication. The study was approved by the Institutional Review Board at the University of Texas at Austin.

2.2. Electrical shock procedures

A 50 ms electrical shock was delivered to the right wrist via electrodes connected to a constant voltage BIOPAC STM200 stimulator (BIOPAC Systems Inc.; Santa Barbara, CA). The intensity (i.e. voltage) of the shock was calibrated in a stepwise manner, starting from a low barely perceptible setting and ending at a level that the participant rated as being “highly annoying and unpleasant, but not painful,” based on the protocols of prior threat conditioning studies (e.g., LaBar, Gatenby, Gore, LeDoux, & Phelps, 1998; Dunsmoor, Kragel, Martin, & LaBar, 2014).

2.3. Skin conductance

Skin conductance was recorded throughout the task using a BIOPAC MP160 system (BIOPAC Systems Inc.; Santa Barbara, CA) with pre-gelled electrodes attached to the hypothenar eminence of the palmar surface of the left hand. Skin conductance level (SCL) for each phase of the study was calculated using an in-house program written in Python (Python Software Foundation) and were transformed using z-scores to normalize the distribution.

2.4. Behavioral procedures

Following attachment of electrodes and shock calibration, participants completed either a spatial version (Experiment 1) or S-R version (Experiment 2) of the eight-arm radial maze task. The radial maze task was created using commercially available software (Unity; Unity Technologies SF; San Francisco, CA). The maze consisted of eight arms radiating from a center platform, and a treasure chest was located at the end of each arm. Participants navigated the maze using a game controller (Logitech F310; Newark, CA). The radial maze task was created based on the 4/8 radial maze employed extensively by Bohbot and colleagues (Iaria et al., 2003; Bohbot et al., 2004, 2007; Banner, Bhat, Etchamendy, Joober, & Bohbot, 2011; Bohbot et al., 2012).

The general procedure for the eight-arm radial maze is depicted in Fig. 1. The behavioral procedures common to Experiments 1 and 2 were as follows. There were 12 trials each divided into two separate phases (Phase 1: Encoding and Phase 2: Retrieval). Participants began each phase in the center of the maze. Phase 1 was an encoding phase where subjects had to collect four out of eight coins located at the ends of four out of eight radial arms. The other four arms were blocked with a translucent wall, blocking entry into the arm. The pattern of blocked arms was different for each trial, and there were never more than two blocked arms in a row (i.e., adjacent to one another) for a given trial. All participants experienced the same patterns of blocked arms in the same order. Participants were informed that each treasure chest contained one coin and that their objective during Phase 1 was to collect the four coins from the treasure chests of the four unblocked “open” arms. Upon colliding with a coin, the coin disappeared, and a three-note jingle was played over headphones to indicate the coin was successfully collected. Participants were explicitly instructed that they would need to remember where they had retrieved the coins, because during Phase 2 they would need to locate the additional 4 coins from the previously blocked arms. Participants were also informed that returning to a previously entered arm in Phase 2 would be considered an error. After all the coins were collected in Phase 1, the screen faded to black. There was no threat of shock during Phase 1.

Fig. 1.

The eight-arm radial maze and threat of shock protocol. A. Participants completed 12 trials of an eight-arm radial maze. Each trial was divided into two separate phases, Phase 1 and Phase 2. Participants began each phase in the center of the maze. For Phase 1, four of the arms were blocked with a wall, preventing entry. Participants were instructed to collect four coins located in the treasure chests of the four open arms. In Phase 2, all of the arms were open (i.e., no walls), and participants were instructed to retrieve the remaining coins from the four arms that were previously blocked. For the spatial radial maze (Experiment 1), various spatial cues surrounded the maze, as pictured above. For the S-R radial maze (Experiment 2), no extra-maze spatial cues were present. B. An instruction was presented for 10s prior to each phase. The instruction indicated the objective of the following phase: “Get Coins from Open Arms” for Phase 1, and “Find Remaining Coins” for Phase 2. During the instruction, participants were also informed whether the following phase would be SAFE or THREAT. Phase 1 was always SAFE, whereas Phase 2 could be SAFE or THREAT.

Phase 2 was a memory test phase that followed Phase 1. In Phase 2 all of the arms were open, and participants were instructed to retrieve the four remaining coins from the four arms that were previously blocked in Phase 1. To successfully navigate to the four remaining coins, the participant had to rely on their memory of Phase 1 to determine which arms still contained coins. A coin only became visible to the participant when they approached a treasure chest containing a coin. Upon colliding with a coin in Phase 2, the coin disappeared, and the jingle was played. If the participant, approached a treasure chest from which a coin was already retrieved, a “wrong” buzzer was played. Entering an arm from which a coin was previously retrieved in Phase 1 was recorded as a “between-phase re-entry” (Fig. 2A). If during Phase 2 the participant entered the same arm more than once, this was recorded as a “within-phase re-entry” (Fig. 2B). In addition, the use of a serial strategy was also detected using an in-house formula written in C# (Microsoft; Seattle, WA). A serial strategy (Fig. 2C) was recorded if the participant entered at least four successive arms in a row (for instance, the North, Northeast, East, and SouthEast arms) during Phase 2. Upon collecting the four coins, a black screen appeared and specified the objective for the next phase, (i.e., “Phase 1: Collect four coins” or “Phase 2: Find remaining coins”). After this 10 s inter-phase interval, the next phase was started.

Fig. 2.

Between-phase re-entries, within-phase re-entries, and serial strategies. A. A between-phase re-entry was recorded during Phase 2 when the participant entered an arm they had already entered during Phase 1. B. A within-phase re-entry was recorded when the participant entered the same arm more than once during the same phase. C. A serial strategy was recorded when the participant entered at least four successive arms in a row. The figure depicts an example of a serial strategy where the participant entered eight successive arms in a row.

Prior to the beginning of the task, participants were informed that during each phase a sign would appear in the upper-right hand corner of the screen and that this sign indicated the likelihood of the participant receiving an electrical shock to the wrist during that phase (see Fig. 1B). A “SAFE” sign with green lettering indicated that they definitely would not receive a shock while they navigated through the environment. In contrast, a “THREAT” sign with red lettering meant they could receive a shock at any time while they navigated the environment. The SAFE sign was accompanied with a translucent green border around the screen and the THREAT sign with a similar red border. In addition, the sign was also displayed prior to the start of each phase in the center of the screen (i.e. during the 10 s inter-phase interval) and was accompanied by a brief description: “No Shock” for SAFE and “Shock At Any Time” for THREAT. Importantly, Phase 1 was always SAFE. Phase 2 was either “SAFE” or “THREAT.” Thus, this procedure was developed to determine the influence of threat of shock during Phase 2 (i.e., the memory retrieval phases), as opposed to Phase 1 (i.e., the memory encoding phases).

During Phase 2, there were five SAFE trials and seven THREAT trials. A shock was only given on two of the seven THREAT trials (10 s into Phase 2 of Trial 2, and 15 s into Phase 2 of Trial 8), and these two THREAT trials with shock were excluded from analysis. Thus, the number of trials included in the analysis were balanced between THREAT and SAFE. The rationale for including two shocked THREAT trials was to reinforce the impression that the threat of shock was legitimate, but to ensure that the shock itself did not confound interpretation of the results during Phase 2. The sequence of SAFE and THREAT trials was STTSSTSTTSTT for half the participants, and TTSTTSTTSTSS for the other half. Following the maze task, participants were presented with a list of emotions (Amusement, Anticipation, Awe, Fear, Relief, Sexual Desire, Joy, Anxiety, Calmness, and Boredom) and were asked to rate how accurately each emotion described their experience during the SAFE and THREAT trials, using a 7-point Likert scale. These ratings helped determine participants’ subjective emotional response to the procedure and verify the ecological validity of the threat task on inducing feelings of threat and negative emotional arousal.

For Experiment 1, participants completed a spatial radial maze task (N = 30; 16 women, 14 men). For this task, participants started each phase facing a randomized direction. Extra-maze spatial cues (i.e., mountains, trees, a boulder, etc.) surrounded the maze and could be used to navigate to arms containing coins or avoiding arms from which coins were already retrieved. For Experiment 2, a separate group of participants completed an S-R radial maze task (N = 32; 16 women, 16 men). For the S-R radial maze, participants started each phase facing the same arm (i.e., the North arm). There were no cues surrounding the maze. Thus, participants could learn the locations of remaining coins by memorizing a pattern of open or closed arms from the starting point. Participants in Experiment 2 were also informed that, if they returned to the center of the maze and stayed there for 3 s, a white line would appear on the floor pointing to the North arm. As soon as they left the center of the maze, the line would disappear. This line was included in the experiment to help subjects reorient themselves to the starting point, but could not be used as a reliable spatial cue because it disappeared when participants navigated the maze. For Experiments 1 and 2, before the start of the task, participants were given a practice trial in the virtual radial maze with the same rules as described above (i.e., collect coins from open arms in Phase 1, and find remaining coins in Phase 2), but without THREAT or SAFE conditions. This allowed participants to familiarize themselves with the rules and the game controls before the start of the experiment.

2.5. Reported strategy use

In Experiment 1, after completing the spatial radial maze task, participants were prompted to report their strategy for finding the coins. If a participant reported using a strategy that involved counting or remembering a pattern of blocked or open arms from an extra-maze spatial cue, they were categorized as having used a combined “spatial-response” strategy. If the participant reported using extra-maze spatial cues and made no mention of counting or memorizing patterns, they were categorized as having used a “spatial” strategy. If the participant did not write any response to the prompt or if they made no mention of memorizing patterns or using extra-maze spatial cues, they were excluded from analyses of strategy use. The criteria for identifying spatial and response strategy users were based on extensive prior research employing the virtual eight-arm radial maze (e.g., Banner et al., 2011; Bohbot, Del Balso, Conrad, Konishi, & Leyton, 2013; West et al., 2015). Each response was categorized by a research assistant blind to the experimental conditions and all other data from each participant.

2.6. Statistics

Dependent measures for maze performance included average number of between-phase re-entries and within-phase re-entries, as well as path length, latency, and speed (path length/latency). In addition, the number of trials in which the use of a serial strategy was detected (i.e., visiting at least four arms in a row) was also included as a dependent measure, and SCL was included as a measure of physiological arousal. All analyses were conducted using Phase 2 data and by comparing the dependent measures across SAFE and THREAT (without shock) conditions for each participant. The 2 THREAT trials with shock were not included in the analyses, nor was data from Phase 1. Paired-samples t-tests (two-tailed) were employed to examine differences in the dependent measures across SAFE and THREAT conditions. In addition, a 2 × 2 repeated measures ANOVA (strategy × trial type) was also employed to determine the influence of strategy use (spatial vs. spatial-response) on errors during SAFE and THREAT conditions. A 2 × 10 (trial type × emotion) repeated-measures ANOVA was used to investigate differences in the ratings of different emotions across the SAFE and THREAT conditions. Statistical analyses were performed using GraphPad Prism 7 (GraphPad; San Diego, CA). The results from each statistical test were considered significant at p < 0.05.

3. Results

3.1. Experiment 1: Spatial radial maze task

3.1.1. Maze performance

Experiment 1 was designed to determine the influence of anticipatory anxiety on memory in a hippocampus-based spatial radial maze task. The average number of errors (between-phase re-entries and within-phase re-entries) during SAFE (“no shock”) and THREAT (“shock at any time”) trials is depicted in Fig. 3A–B. A paired-samples t-test (two-tailed) revealed that participants demonstrated a greater number of between-phase re-entries during THREAT trials (M = 5.767) relative to SAFE trials (M = 4.033), t(29) = 2.688, p = 0.0118, d = 0.491. In addition, participants also made more within-phase re-entries during THREAT trials (M = 3.733) compared to SAFE trials (M = 2.067), t(29) = 2.114, p = 0.0432, d = 0.386. These findings suggest that the threat of shock impaired memory retrieval in the spatial radial maze.

Fig. 3.

The influence of SAFE and THREAT on maze performance variables in the spatial radial maze task (Experiment 1). A, B. THREAT was associated with a greater mean number of between-phase re-entries and within-phase re-entries, compared to SAFE, suggesting a memory impairment under THREAT. C. There was a greater mean number of trials in which a serial strategy was detected under THREAT versus SAFE conditions. D–F. There was no significant difference in path length, latency, or speed across SAFE and THREAT conditions. a.u. = arbitrary units. *p < 0.05.

The percentage of SAFE trials and THREAT trials in which a serial strategy was detected is depicted in Fig. 3C. Paired samples t-test (two-tailed) revealed a significantly greater percentage of THREAT trials (M = 19.34%) versus SAFE trials (M = 9.31%) where a serial strategy was detected, t(29) = 3.746, p < 0.001, d = 0.684. Taken together with findings reported above that THREAT trials were associated with more errors (i.e., re-entries), the present results suggest that when participants could not remember the locations of the coins they would revert to a strategy where they visited each successive arm (i.e., a serial strategy) until finding the remaining coins.

Regarding other measures of maze performance, paired samples t-tests (two-tailed) revealed no significant differences in latency (seconds; THREAT M = 66.81, SAFE M = 61.27; t(29) = 1.397, p = 0.1730, d = 0.255), path length (THREAT M = 863.61, SAFE M = 809.70; t(29) = 1.377, p = 0.1790, d = 0.252), or speed (THREAT M = 13.966, SAFE M = 14.249; t(29) = 1.236, p = 0.2265, d = 0.223) between THREAT and SAFE trials (Fig. 3D–F). Importantly, as there was no difference in speed between SAFE and THREAT trials, the increase in errors under THREAT may not be readily attributed to a speed-accuracy tradeoff.

3.1.2. Reported strategy use

After the task, participants were prompted to report the strategy they used to remember the locations of the coins. This information was important for determining whether reported strategy use was a factor modulating the effect of anticipatory anxiety on memory. Most participants reported using a combined spatial-response strategy (n = 17; e.g., “I located the arm with the bales of hay on the left side and labeled it 1, then going clockwise I labeled the rest of the arms 2–8. I counted which numbers were the open arms then recited those numbers and went to the opposite arms when I went back for phase two”). Fewer participants employed a spatial strategy (n = 9; e.g., “I would use environmental cues, such as the tree or rock, during phase 1 to try and remember where the coins were and then I would try and pick all the places I didn’t remember during phase 2”). However, a chi-square analysis revealed that the number of spatial strategy users vs. spatial-response strategy users did not differ significantly (χ²(1, 26) = 2.462, p = 0.1167, v = 0.116). The remaining participants (n = 4) did not report a strategy, their description was unclear, or their responses could not be categorized as either “spatial-response” or “spatial” (e.g., “I was just going from location to location without really thinking. I tried each location once”).

To determine whether strategy use influenced the effect of threat on memory, a 2 × 2 repeated measures ANOVA (trial type [THREAT or SAFE] as a within-subjects factor and strategy [spatial or spatial-response] as a between-subjects factor) was computed on the number of between-phase re-entries (Fig. 4A). The analysis revealed a main effect of trial type (F(1, 24) = 7.719, p = 0.0104, η_p² = 0.242), but no main effect of strategy (F(1, 24) = 2.415, p = 0.133, η_p²= 0.092) and no trial type × strategy interaction (F(1, 24) = 0.2295, p = 0.6363, η_p² = 0.009) for between-phase re-entries. The same analysis computed on number of within-phase re-entries (Fig. 4B) indicated a main effect of trial type (F(1, 24) = 13.51, p = 0.0012, η_p² = 0.502) and a main effect of strategy (F(1, 24) = 4.574, p = 0.0429, η_p² = 0.160), but no trial type × strategy interaction (F(1, 24) = 1.456, p = 0.2394, η_p² = 0.057). Taken together, anticipatory anxiety remained effective in impairing memory across those using a spatial or spatial-response strategy, however the spatial-response strategy was associated with fewer within-phase re-entries overall. Importantly, it should be noted that the present experiment may not be appropriately powered to examine whether the influence of anticipatory anxiety depended on strategy use. It would be beneficial for future research to include more participants in order to increase power for these analyses, which would potentially reveal additional effects.

Fig. 4.

Spatial and spatial-response strategies in the spatial radial maze task (Experiment 1). A. Between-phase re-entries were examined across spatial and spatial-response strategy users. There was a main effect of trial type (p < 0.05), but no effect of strategy (p > 0.05) or interaction between strategy and trial type (p > 0.05), suggesting that THREAT was associated with more between-phase re-entries regardless of strategy. B. Within-phase re-entries were examined across spatial and spatial-response strategy users. There was a main effect of trial type (p < 0.05) as well as a main effect of strategy (p < 0.05), but there was no interaction between strategy and trial type (p > 0.05).

3.1.3. Skin conductance level

Skin conductance level (SCL) was used to assess tonic levels of emotional arousal across SAFE and THREAT conditions in the spatial radial maze, and is depicted in Fig. 5A. SCL values were transformed into z-scores, and paired-samples t-tests (two-tailed) were run to examine potential differences in SCL between SAFE and THREAT conditions. During the 10 s inter-phase intervals before Phase 2, participants were informed whether the imminent phase would be SAFE or THREAT. Results indicated that SCL was significantly higher during the THREAT versus SAFE inter-phase intervals, t(29) = 3.354, p = 0.0022, d = 0.613. In addition, as depicted in Fig. 5B, SCL was also higher during the THREAT vs. SAFE runs (i.e., during memory retrieval in Phase 2), t(29) = 4.118, p < 0.001, d = 0.751. The findings suggest that the THREAT warning increased physiological arousal in the spatial radial maze.

Fig. 5.

Physiological and subjective emotional responses to SAFE versus THREAT conditions during the spatial radial maze task (Experiment 1). A. A THREAT warning during the inter-phase interval preceding Phase 2 was associated with greater mean skin conductance level (SCL) than a SAFE warning. B. During the Phase 2 runs (i.e., when a participant collected the remaining coins), a THREAT sign was associated with greater mean SCL than a SAFE sign. C. Participants were presented with a list of emotions and asked to rate how accurately each emotion described their experience during SAFE and THREAT trials. Anticipation, fear, and anxiety were rated higher for THREAT compared to SAFE trials. Relief, joy, calmness, and boredom were rated higher for SAFE compared to THREAT trials. *p < 0.05.

3.1.4. Subjective emotional responses to threat

After the spatial radial maze task, participants were presented with a list of emotions and were asked to rate how accurately each emotion described their experience during the SAFE and THREAT conditions (Fig. 5C). Two participants did not respond to the questions and were excluded from the analysis. A 2 × 10 repeated-measures ANOVA (Trial Type X Emotion) indicated a significant main effect of emotion (F(9, 270) = 14.97, p < 0.001, η_p² = 0.403), no main effect of trial type (F (1, 270) = 0.0708, p = 0.7903, η_p² < 0.001), and a significant trial type X emotion interaction (F(9, 270) = 42.81, p < 0.001, η_p² = 0.588). Bonferroni’s post-hoc comparisons indicated that participants reported significantly higher accuracy for Anticipation, Fear, and Anxiety during THREAT (M = 5.500, 4.107, and 5.000, respectively) relative to SAFE conditions (M = 2.000, 1.393, and 1.464, respectively; p < 0.001 for all comparisons; d = 1.732, 1.343, and 1.749, respectively). In contrast, participants reported greater Relief, Joy, Calmness, and Boredom during SAFE (M = 4.786, 2.929, 5.321, and 3.786, respectively) relative to THREAT conditions (M = 2.143, 1.643, 1.929, and 2.036, respectively; p < 0.001, p = 0.0087, p < 0.001, and p < 0.001, respectively; d = 1.308, 0.636, 1.679, and 0.866, respectively). Ratings for Amusement, Awe, and Sexual Desire did not differ significantly between SAFE and THREAT trials.

3.2. Experiment 2: Stimulus-response (S-R) radial maze task

3.2.1. Maze performance

The average number of errors (between-phase re-entries and within-phase re-entries) during SAFE and THREAT trials in the S-R radial maze task is depicted in Fig. 6A–B. A paired-samples t-test (two-tailed) revealed no difference in between-phase re-entries during THREAT trials (M = 3.5313) versus SAFE trials (M = 2.813), t(31) = 1.306, p = 0.2012, d = 0.231. In addition, there was no difference in within-phase re-entries during THREAT trials (M = 2.033) compared to SAFE trials (M = 1.219), t(31) = 1.642, p = 0.1107. d = 0.291. These findings suggest that the threat of shock did not impair memory retrieval in the S-R radial maze.

Fig. 6.

The influence of SAFE and THREAT on maze performance variables in the S-R radial maze task (Experiment 2). A, B. Between-phase re-entries and within-phase re-entries did not differ between SAFE and THREAT conditions. C. The use of a serial strategy did not differ between THREAT versus SAFE conditions. D–F. There was no significant difference in path length or latency across SAFE and THREAT conditions. However, participants moved faster on average during THREAT. a.u. = arbitrary units. *p < 0.05.

The percentage of trials in which a serial strategy was detected in the S-R radial maze task is depicted in Fig. 6C. A paired samples t-test (two-tailed) indicated no difference in percentage of THREAT trials (M = 17.5%) and SAFE trials (M = 15.62%) in which a serial strategy was detected, t(31) = 0.3853, p = 0.7026, d = 0.068. Regarding other maze performance variables, paired-samples t-tests (two-tailed) revealed no significant differences in latency (seconds; THREAT M = 55.25, SAFE M = 55.83; t(31) = 1.988, p = 0.8437, d = 0.035; Fig. 6D) or path length (THREAT M = 826.22, SAFE M = 783.28; t(31) = 1.284, p = 0.2088, d = 0.227; Fig. 6E). However, there was a significant difference in speed (path length [arbitrary units]/latency [seconds]); Fig. 6F) with participants moving faster on average during THREAT trials (M = 16.29) versus SAFE trials (SAFE M = 15.27; t(31) = 1.236, p = 0.0087, d = 0.496).

3.2.2. Skin conductance level

Skin conductance level across SAFE and THREAT conditions in the S-R radial maze task is depicted in Fig. 7A. Results indicated that SCL was significantly higher during the THREAT versus SAFE inter-phase intervals, t(31) = 3.549, p = 0.0013, d = 0.627. In addition, SCL was also higher during the THREAT vs. SAFE runs, t(31) = 3.956, p < 0.001, d = 0.699. Thus, as observed in the spatial radial maze task (Experiment 1), the findings from Experiment 2 confirm that the THREAT warning increased physiological arousal in the S-R radial maze.

Fig. 7.

Physiological and subjective emotional responses to SAFE versus THREAT conditions during the S-R radial maze task (Experiment 2). A. A THREAT warning during the inter-phase interval preceding Phase 2 was associated with greater mean skin conductance level (SCL) than a SAFE warning. B. During the Phase 2 runs (i.e., when a participant collected the remaining coins), a THREAT sign was associated with greater mean SCL than a SAFE sign. C. Participants were presented with a list of emotions and asked to rate how accurately each emotion described their experience during SAFE and THREAT conditions. Anticipation, fear, and anxiety were rated higher for THREAT compared to SAFE trials. Relief, joy, calmness, and boredom were rated higher for SAFE compared to THREAT trials. *p < 0.05.

3.2.3. Subjective emotional responses to threat

After the S-R radial maze task, participants were presented with a list of emotions and were asked to rate how accurately each emotion described their experience during the SAFE and THREAT conditions (Fig. 7B). A 2 × 10 repeated-measures ANOVA (Trial Type X Emotion) indicated a significant main effect of emotion (F(9, 310) = 14.26, p < 0.001, η_p² = 0.393), no main effect of trial type (F(1, 310) = 0.01105, p = 0.9163, η_p² < 0.001), and a significant trial type X emotion interaction (F(9, 310) = 31.04, p < 0.001, η_p² = 0.474). Bonferroni’s post-hoc comparisons indicated that participants reported significantly higher accuracy for Anticipation, Fear, and Anxiety during THREAT (M = 4.406, 4.125, and 4.563, respectively) relative to SAFE conditions (M = 2.344, 1.281, and 1.375, respectively; p < 0.001 for all comparisons; d = 0.970, 1.337, and 1.499, respectively). In contrast, participants reported greater Relief, Joy, Calmness, and Boredom during SAFE (M = 5.219, 3.094, 5.094, and 3.144, respectively) relative to THREAT conditions (M = 2.313, 1.938, 2.781, and 1.906, respectively; p < 0.001, p = 0.0229, p < 0.001, and p < 0.0174, respectively; d = 1.366, 0.544, 1.087, and 0.558, respectively). Ratings for Amusement, Awe, and Sexual Desire did not differ significantly between SAFE and THREAT trials.

4. Discussion

The present study was designed to investigate the influence of dynamic changes in anticipatory anxiety on distinct memory systems during navigation of a virtual environment. Threat of shock was associated with higher physiological arousal (i.e., skin conductance level) and subjective fear, anxiety, and anticipation, relative to safe “no shock” conditions. In a hippocampus-dependent spatial radial maze task (Experiment 1), a higher number of errors was observed when subjects anticipated an aversive shock to the wrist, suggesting an impairment in the retrieval of hippocampal spatial memory under anticipatory anxiety. In addition, in the spatial radial maze, there was a higher percentage of threat trials versus safe trials in which a dorsal striatum-based serial strategy (Rueda-Orozco et al., 2008) was detected. In the S-R radial maze, which did not contain extra-maze spatial cues (Experiment 2), there were no differences in errors or in the use of a serial strategy between safe and threat conditions. Taken together, the present findings indicate that a dynamic increase in anticipatory anxiety impairs spatial memory and induces a greater reliance on S-R procedural navigation (i.e., serial strategies). In contrast, anticipatory anxiety does not interfere with solving a navigational task that specifically relies on an S-R memory strategy.

Spatial memory and S-R memory in the virtual eight-arm radial maze involve the function of distinct neural systems. Previous research employing functional neuroimaging has shown that spatial memory in the radial maze is associated with greater volume and neural activation of the hippocampus, whereas S-R memory in the radial maze has been linked to greater volume and neural activation of the dorsal striatum (Banner et al., 2011; Bohbot, Iaria, & Petrides, 2004; Bohbot, Lerch, Thorndycraft, Iaria, & Zijdenbos, 2007; Horga et al., 2015; Iaria et al., 2003; Konishi & Bohbot, 2013). Thus, the influence of threat on memory and serial strategy use observed in the present experiments may be attributed to differences in how emotional arousal influences the function of hippocampal and dorsal striatal memory systems. That is, anxiety induced by threat of shock may be associated with impaired hippocampal spatial memory and increased use of a dorsal striatal navigational strategy. The present findings are largely consistent with prior evidence that high levels of emotional arousal impair hippocampal spatial memory and enhance the use of dorsal striatal S-R memory across a variety of memory tasks in lower animals and human subjects (for reviews, see Packard and Goodman, 2012, 2013; Schwabe, 2013; Goldfarb & Phelps, 2017; Goodman, McIntyre et al., 2017). The present study confirms that similar effects may also be observed using a “threat of shock” paradigm, which is a more specific procedure for inducing dynamic (i.e., trial-by-trial) changes in anticipatory anxiety in experimental situations (Robinson et al., 2013) compared to protocols for elicitation of stress (e.g., the cold pressor test; Schwabe & Schächinger, 2018).

Interestingly, when prompted to describe the strategy that was used to solve the spatial memory task, most participants reported using a combined spatial-response strategy, meaning that they would use the extra-maze spatial cues as reference points for memorizing a pattern of arms that still contain, or no longer contain, coins. There are different ways to design the virtual eight-arm radial maze and some investigators have run the task without varying the facing direction between phases, which has allowed participants to employ a response strategy without looking at spatial cues (i.e., in this case, the first arm they are facing becomes the reference point). Response strategies in the virtual eight-arm radial maze have been associated with increased dorsal striatal activity (Iaria et al., 2003) and, therefore, it is tempting to speculate that the combined spatial-response strategy observed in the present study may result from cooperation between hippocampal and striatal memory systems. Importantly, in the present study, using a pure spatial strategy or combined spatial-response strategy did not protect against the impairing effect of anticipatory anxiety on memory in the spatial radial maze task. However, participants who reported using a spatial- response strategy showed fewer within-phase re-entries in general across the SAFE and THREAT conditions, relative to those reporting a spatial strategy. Future research should investigate whether (1) the spatial-response strategy arises from cooperation between memory systems and (2) whether this approach represents a more ideal strategy for preventing errors in the spatial eight-arm radial maze task.

Whereas anxiety induced by threat of shock impaired spatial memory and enhanced the use of a serial S-R strategy in the spatial task, threat of shock was not associated with enhanced memory in a radial maze task that specifically required the use of an S-R strategy (Experiment 2). Considering extensive prior evidence that emotional arousal is associated with enhanced acquisition and retrieval of dorsal striatal S-R memory (Packard & Wingard, 2004; Leong, Goodman, & Packard, 2012; Guenzel, Wolf, & Schwabe, 2014; Leong, Goodman, & Packard, 2015; Vogel & Schwabe, 2018; but see also Guenzel, Wolf, & Schwabe, 2013), it was expected that threat of shock would produce a similar memory enhancement in the present S-R radial maze task. The reason we did not observe an S-R memory enhancement under threat may be attributed to the lack of competition between memory systems in the S-R radial maze. In learning situations where hippocampal and dorsal striatal memory processes compete for behavioral control, stress-induced impairment of hippocampal memory presumably allows S-R memories to be acquired and retrieved unencumbered by competitive interference from the hippocampus (Wingard & Packard, 2008; Sadowski, Jackson, Wieczorek, & Gold, 2009; Goodman & McIntyre, 2017). This hippocampal impairment is believed to be a critical mechanism for the enhancing effects of emotional arousal on dorsal striatal memory (Packard, 2009; Schwabe, 2013; Goldfarb & Phelps, 2017; Packard et al., 2018). In the present study, as no extra-maze spatial cues were available in the S-R task employed in Experiment 2, spatial memory mechanisms may not have strongly competed with the acquisition and retrieval of an S-R memory strategy. Thus, threat-induced impairment of spatial memory may not have benefitted S-R memory in this task.

Notwithstanding the lack of an emotional memory enhancement in the S-R radial maze, the present findings provide a clear dissociation regarding the influence of anxiety on different kinds of memory, in which threat of shock impaired memory in the spatial task but not the S-R task. Importantly, the two memory tasks employed in the present study involve similar non-mnemonic processes, such as motivational, motor, and attentional demands, but differ in terms of the type of memory mechanisms required by each task. Thus, differences in how threat of shock influenced performance variables across the spatial and S-R versions of the radial maze could not be readily attributed to a potential effect of threat on the non-mnemonic processes similar in both tasks.

Emotional modulation of memory systems depends on the function of the basolateral complex of the amygdala (BLA; McGaugh, 2004; Packard, 2009). In rodents, extensive prior evidence indicates that lesion or temporary inactivation of the BLA prevents stress and anxiety from impairing hippocampal spatial memory and enhancing dorsal striatal S-R habit memory (Kim, Lee, Han, & Packard, 2001; Packard & Gabriele, 2009; Leong & Packard, 2014; Goode, Leong, Goodman, Maren, & Packard, 2016). In human subjects, enhanced use of dorsal striatal S-R strategies under emotional arousal has also been associated with greater amygdala activity, as well as increased functional connectivity between the amygdala and dorsal striatum and decreased functional connectivity between the amygdala and hippocampus (Schwabe, Tegenthoff, Höffken, & Wolf, 2013). Stress hormones released during periods of high emotional arousal also play an important role in the emotional modulation of hippocampal and dorsal striatal memory systems (Quirarte et al., 2009; Schwabe et al., 2010; Schwabe, Höffken, Tegenthoff, & Wolf, 2011, 2013; Goodman, Leong, & Packard, 2015; Goode et al., 2016; Siller-Pérez, Serafín, Prado-Alcalá, Roozendaal, & Quirarte, 2017).

Previous studies employing a “threat of shock” paradigm indicate that threat engages activity of brain regions typically implicated in emotional arousal and defensive behaviors (Robinson et al., 2013). Specifically, several studies have reported increased coupling between the dorsomedial prefrontal cortex (dmPFC) and amygdala during threat of shock, relative to safe conditions (Robinson, Charney, Overstreet, Vytal, & Grillon, 2012; Vytal, Overstreet, Charney, Robinson, & Grillon, 2014; Robinson et al., 2016). Interestingly, threat of shock has also been associated with greater coupling between amygdala and dorsal striatum, as well as between dmPFC and dorsal striatum (Vytal et al., 2014), which may be part of the mechanism through which threats can lead to greater use of S-R habit memory. However, further research using neuroimaging methods will be required to determine the precise neural mechanisms underlying the effects of threat on memory systems in the present study.

The emotional modulation of memory systems observed in laboratory settings may be important for understanding the development of some psychopathologies. In particular, there is emerging evidence that dorsal striatal S-R habit memory may be associated with the formation and expression of habit-like behavioral symptoms across a range of psychiatric disorders, such as drug addiction and obsessive-compulsive disorder, among others (White, 1996; Everitt & Robbins, 2005; Schwabe, Dickinson, & Wolf, 2011; Goodman, Leong, & Packard, 2012; Berner & Marsh, 2014; Gillan & Robbins, 2014; Goodman, Marsh, Peterson, & Packard, 2014). Stressful life events and anxiety exacerbate S-R habitual symptoms in these disorders, increasing expression and relapse (e.g., Breese et al., 2005; Bruns & Geist, 1984; de Silva & Marks, 1999; McLaren & Crowe, 2003). Thus, studies examining the influence of emotional arousal on memory systems may prove useful in determining how anxiety leads to habit-like pathologies and may also lead to new therapeutic interventions based on procedures that have previously been shown to suppress S-R habit memory and block the emotional enhancement of habits (Schwabe et al., 2010; Schwabe, Höffken, et al., 2011, 2013; Leong et al., 2012; Goodman & Packard, 2015b; Goode et al., 2016; Goodman, Ressler, & Packard, 2016; Goodman, Ressler, & Packard, 2017).