The Tap‐to‐Safety Task: A Novel fMRI Paradigm Assessing Repetitive Threat‐Neutralization

ABSTRACT

Existing experimental threat‐related paradigms focus primarily on active or passive avoidance behavior, but do not model the common behavioral pattern of repetitive, effortful actions aimed at neutralizing perceived threats. Here, we describe and provide initial validation for the Tap‐to‐Safety (TTS) Task, a novel human paradigm designed to experimentally elicit repetitive threat‐neutralization behavior during functional magnetic resonance imaging (fMRI). Adult participants completed the TTS Task; one sample completed the task online and an additional sample completed the task in person, with a subsample completing fMRI. Task stimuli included a threat cue (CS+) paired with an aversive unconditioned stimulus (US), safety cues (CS−) never paired with the US, and safe generalization stimuli (GSs) varying in similarity to the CS+. During an extinction phase, the CS+ was no longer paired with the US. Trials included passive viewing trials, without a neutralization option; and choice trials, in which participants could tap a button repeatedly to gain protection from the US (i.e., repetitive threat‐neutralization) while reducing accumulation of reward points. Linear mixed‐effects models (LMEs) were used to assess behavioral and neural responses. For fMRI analyses in a subset of participants, a priori regions of interest (ROIs) were used with Bonferroni correction. In‐person results (n = 49) demonstrated increased threat expectancy, anxiety, and repetitive threat‐neutralization behavior that were higher to the threat cue than to safety cues (ps < 0.001, η _p ² > 0.42), and generalized across safe stimuli resembling the threat cue (ps < 0.001, η _p ² > 0.42). During extinction, risk and anxiety ratings gradually decreased (ps < 0.015, η _p ² > 0.01), whereas neutralization behavior persisted (p = 0.10). Greater trial‐wise neutralization predicted lower post‐neutralization threat expectancy and anxiety ratings (ps < 0.005, η _p ² > 0.11). Behavioral results were largely replicated in an online sample (n = 88). Analyses of fMRI data (n = 31) indicated that neural activity pre‐neutralization in anterior insula, dorsal anterior cingulate cortex (dACC), and dorsal striatum scaled with threat relevance of stimuli (ps < 0.001, η _p ² > 0.30) and with magnitude of neutralization (ps < 0.003, η _p ² > 0.06). These findings support the use of the TTS Task for quantifying the behavioral and neural mechanisms of repetitive threat‐neutralization. Results point to a key role of the salience network and dorsal striatum. Future research in clinical populations is warranted.

Key Points

• Repetitive threat‐neutralization is a commonly observed behavioral pattern with clinical relevance.

• The Tap‐to‐Safety (TTS) Task is a novel tool assessing repetitive threat‐neutralization.

• Anterior insula, dorsal anterior cingulate cortex (dACC), and dorsal striatum are implicated in repetitive threat‐neutralization.

In this study, we present a novel task to model repetitive threat‐neutralization behavior during neuroimaging. Results suggest that the dorsal anterior cingulate cortex (dACC), anterior insula, and dorsal striatum interact to drive neutralization behavior in the face of perceived threats.

1. Introduction

Threat‐related behaviors, such as avoidance, are known to exacerbate and maintain threat expectancies (van Dis et al. 2022; Engelhard et al. 2015; Wong et al. 2023; Xia et al. 2017), conferring substantial anxiety‐related functional impairment (American Psychiatric Association 2013). There is a wealth of lab‐based animal and human research on approach‐avoidance behavior (Wong et al. 2022), which has led to increased understanding of the neural systems underlying these behaviors (Aupperle and Paulus 2010; Kirlic et al. 2017), and informed published clinical guidelines for reducing avoidance behavior in treatment for anxiety‐related disorders (Craske et al. 2008). This large literature has primarily focused on active avoidance, in which the individual acts to move away from threat, and passive avoidance, in which the individual withholds an action to avoid threat (reviews: Krypotos et al. 2018; LeDoux and Daw 2018; Letkiewicz et al. 2023). In human avoidance paradigms, participants typically perform a single button press to approach or avoid (Krypotos et al. 2018), limiting the ability of these paradigms to model real‐world behaviors that are performed repeatedly to neutralize threat. This is a significant limitation, given that repetitive threat‐neutralization behaviors, such as repeated checking, cleaning, or reassurance‐seeking, are common across anxiety‐related, trauma‐related, and obsessive‐compulsive disorders (American Psychiatric Association 2013).

Here, we present a novel experimental paradigm designed to elicit costly repetitive threat‐neutralization behavior and identify its neural substrates with functional magnetic resonance imaging (fMRI). The present paradigm is similar to existing avoidance tasks in a number of ways. First, the paradigm involves conflicting motivations to avoid threat and approach reward, such that avoiding threat confers loss of reward (e.g., van Meurs et al. 2014; Pittig et al. 2014). This mirrors the motivational conflict inherent in many instantiations of maladaptive threat‐related behavior; for example, an individual who engages in repetitive hand‐washing might sacrifice opportunities for social engagement, productivity, rest, or other positive outcomes. Second, the paradigm draws on and extends existing methods for assessing non‐dichotomous avoidance behavior. Tasks in which participants can modulate the probability of a negative outcome by moving an onscreen slider (e.g., Aupperle et al. 2015; Wong and Pittig 2022) or squeezing a handheld grip bar (Nord et al. 2017) offer dimensional measurement of avoidance, but do not constitute repetitive behavior. Another existing task includes avoidance options of high‐effort or low‐effort repeated button‐tapping (Dymond et al. 2024), enabling assessment of repetitive behavior at two levels. The present behavioral paradigm is the first to our knowledge to include free‐choice repetitive threat‐neutralization behavior (i.e., participants are free to choose the number of repetitions they would like to perform within a discrete time period), such that the number of repetitions is negatively correlated with both (a) probability of a negative outcome and (b) magnitude of reward. This enables fine‐grained assessment of the decision‐making processes underlying costly repetitive threat‐neutralization behavior.

Our hypotheses relating to the neural correlates of this common and clinically‐relevant behavior pattern stem from prior animal and human literature on threat appraisal and decision‐making (see Figure 1). Neural systems underlying threat appraisal are likely to underlie any threat‐related behavior, including the dorsal anterior cingulate cortex (dACC), anterior insula, amygdala, and ventromedial prefrontal cortex (vmPFC). The dACC and anterior insula have been identified as cortical hubs of the ‘salience network’ (Seeley et al. 2007), with the dACC also implicated in detection of motivational conflict (Brockett et al. 2020; Shahnazian and Holroyd 2018); the anterior insula in integrating interoception and emotional awareness (Simmons et al. 2012; Zaki et al. 2012); and both regions implicated in fear expression (Fullana et al. 2016; Milad et al. 2007; Webler et al. 2021). The amygdala has been robustly implicated in fear learning in rodent models (LeDoux 1993), and activates to threat cues in some human neuroimaging studies (Lissek et al. 2014; Phelps et al. 2004), though not in recent meta‐analyses (Fullana et al. 2016; Webler et al. 2021); these mixed findings may reflect rapid habituation of amygdala activity to a new threat cue (LaBar et al. 1998; Sehlmeyer et al. 2009). The vmPFC, by contrast, has been implicated in safety signaling and extinction learning (Cha et al. 2014; Harrison et al. 2017).

FIGURE 1.

Theoretical model of neural substrates of repetitive threat‐neutralization. AIns, Anterior insula; dACC, Dorsal anterior cingulate cortex; lOFC, Lateral orbitofrontal cortex; vmPFC, Ventromedial prefrontal cortex. Figure created with BioRender.

In addition to threat and safety circuitry, neural systems involving decision‐making and action selection are also hypothesized to underlie threat‐related behavior. The orbitofrontal cortex (OFC) plays a putative role in stimulus valuation (Camille et al. 2011; Gottfried 2003; Kringelbach and Rolls 2003) and threat‐related behavior (Elliott et al. 2010; Hollerman et al. 2000; O'Doherty et al. 2003). The dACC has also been associated with action valuation and effort devaluation (Klein‐Flügge et al. 2016). These regions are in turn thought to influence action selection via inputs to the dorsal striatum (Haber 2016); dorsal striatal regions have also been robustly implicated in repetitive behavior in rodents (Schilman et al. 2010).

In the present study, we provide initial characterization of in‐person and online versions of a novel task, termed the Tap‐to‐Safety (TTS) Task, including behavioral and neural response patterns. We first tested whether the task would elicit the expected behavioral effects in online and in‐person versions. We hypothesized that (1) the task would elicit repetitive threat‐neutralization behavior to a threat cue that would generalize across safe stimuli with partially overlapping features to the threat‐cue; (2) that neutralization behavior would gradually decrease over time after the threat was removed; and (3) that neutralization behavior would be followed by post‐neutralization decreases in perceived risk and anxiety. We then used fMRI to identify neural correlates of conditioned fear and repetitive threat‐neutralization behavior. We predicted that (4) activity in dACC, anterior insula, amygdala, and dorsal striatum (caudate, putamen) would scale with threat‐relevance of stimuli during stimulus onset. We also hypothesized that (5) pre‐neutralization activity (i.e., during putative deliberation or decision‐making) in these regions as well as the OFC would scale with threat‐relevance of stimuli, given the OFC's putative role in decision‐making. Finally, we hypothesized that (6) activity in dACC, anterior insula, amygdala, OFC, and dorsal striatum would correlate with the degree of repetitive threat‐neutralization behavior.

2. General Methods

In Study 1, a sample of participants completed the task in‐person. In Study 2, behavioral effects were then examined in a separate online sample. Finally, in Study 3, neural correlates of the task were examined using fMRI. These experiments were not preregistered.

3. Study 1: In‐Person Behavioral Effects

3.1. Methods

3.1.1. Participants

Adults ages 18–65 were recruited from the community and local mental health clinics. The sample included participants either with no psychiatric disorder or with a current anxiety disorder and/or obsessive compulsive disorder diagnosis. Exclusion criteria included: current DSM‐5 diagnosis of a psychosis spectrum disorder, bipolar disorder, moderate or severe substance use disorder, PTSD; active suicidal ideation; current use of anxiolytics, antipsychotics, mood stabilizers, or antidepressants, with the exception of selective serotonin reuptake inhibitors (SSRIs) at a stable dose for at least 6 weeks prior to enrollment; stimulant use or caffeine intake (> 1000 mg/day), MRI contraindications, and medical conditions that could interfere with participation or be compromised by participation.

3.1.2. Procedure

Participants completed a shock calibration procedure in which they self‐administered brief presentations of the shock starting at a low level, and shock level was increased in intensity gradually until it reached a level that the participant rated as uncomfortable but tolerable; this level of shock was then used throughout the study. Participants then completed the initial Acquisition phase of TTS Task outside of the MRI environment, followed by the remaining three phases of the TTS Task with fMRI. All study procedures were conducted in accordance with the Declaration of Helsinki and were approved by the WCG Institutional Review Board (#20192170, #2024006, #20200161). Written informed consent was obtained from all participants after a detailed explanation of the study procedure was provided. Participants were compensated for taking part in the study.

3.1.3. Equipment

The task was programmed with a custom internally‐developed application used for developing and hosting tasks using 2020.1.3 PsychoPy 3. Participants made risk and anxiety ratings and performed repetitive threat‐neutralization using a Current Designs bimanual response pad; the right‐hand pad had four buttons in a curved formation and was used for risk and anxiety ratings, and the left‐hand pad included one button for the little finger and was used for the speed test and repetitive threat‐neutralization (described below). Electric shocks were delivered to the right ankle by a constant current stimulator (Digitimer DS7A, Digitimer North America, Fort Lauderdale, FL). (Of note, participants were not excluded based on handedness).

3.1.4. The Tap‐to‐Safety Task

3.1.4.1. Task Procedure

The TTS Task was designed to elicit competing motivations for threat‐avoidance and reward accrual. The task begins with Pavlovian conditioning with a threat cue paired with shock and safety cues never paired with shock. Next, participants are instructed that on future trials they will be able to repeatedly tap a button to reduce their risk of shock, but that doing so will interfere with the accrual of reward points. Threat cues, safety cues, and safe generalization stimuli varying in similarity to the threat cue are presented sequentially, with the neutralization option available on some trials. In the final phase of the task, the threat cue and safety cues are presented without shock, to assess extinction learning.

3.1.4.2. Stimuli

Task stimuli included ring shapes along a continuum of similarity and one outlier shape. One extreme stimulus (CS+) was paired with the unconditioned stimulus (US), a mild electric shock delivered to the right ankle, and the other extreme was a safety cue (oCS−) never paired with shock. The CS+ was counterbalanced across participants. The V‐shaped outlier stimulus (vCS−) was also never paired with shock, and was included to control for generalization to all ring‐shaped stimuli. The intermediate‐sized cues were used as generalization stimuli (GSs), which were also never paired with shock (Figure 2A). Each class of generalization stimuli (GS1, GS2, GS3) consisted of two rings, such that GSs formed a continuum of similarity consisting of six rings in total between oCS− and CS+. Rings increased in size by increments of 20% of the smallest ring, such that diameters of circular stimuli in inches were: 0.80, 0.96, 1.12, 1.28, 1.44, 1.60, 1.76, 1.92. The V‐shaped stimulus was 1.36 in. high.

FIGURE 2.

The Tap‐to‐Safety task: Stimuli and trial‐types. (A) Sizes of task stimuli were counterbalanced across participants such that for half of participants, the largest ring was paired with shock (A1) and for the remaining participants the smallest ring was paired with shock (A2). The task included (B) passive trials and (C) choice trials. Durations of each event in the trial sequence are shown in seconds. We assessed neural activity corresponding to stimulus onset and tap preparation, indicated with bold and underlined text. At the end of all trials, a red frame around the edges of the screen is presented for 1 s to signal the outcome period during which shock may or may not occur. On choice trials, a yellow frame and a verbal prompt signal a 4‐s period during which button‐tapping has an effect.

3.1.4.3. Trial Types

Passive Trials: Passive trials were included to enable initial Pavlovian acquisition of CS+/US contingency, and ongoing retention of the contingency throughout the task. The stimulus was displayed onscreen for the duration of each passive trial. Following initial stimulus onset, after 1–2 s (jittered for fMRI), onscreen text prompted participants to rate either (a) their risk of shock on the current trial or (b) their anxiety in the current moment. Participants were given 4 s to make the rating using an onscreen slider, on a scale from 1 (“none”) to 5 (“a lot”). Next, participants viewed a red frame around the edges of the screen for 1 s, signaling the outcome period of the trial when a shock might be delivered (the red frame was displayed on all trials regardless of shock delivery). On some trials, a 2‐s pause was included between the rating and the outcome, with the stimulus still onscreen, to avoid perception of contingency between the rating and outcome. (Figure 2B).

Choice Trials: Choice trials were included to assess repetitive threat‐neutralization behavior. Prior to the introduction of choice trials (see Task Phases, below), participants were explicitly instructed that they would be able to tap a button in their left hand to reduce their risk of shock, such that more tapping on a given trial conferred more protection from shock on that trial. On choice trials (Figure 2C), following initial stimulus onset (jittered 1–2 s), participants were given 4 s to make a risk or anxiety rating, as was done in passive trials. Following the rating, the prompt “You will soon have the option to tap the button…” appeared for 2 s. Then, the prompt “You now have the option to tap the button!” appeared for 4 s, along with a yellow frame around the edges of the screen, to clearly signal the duration of the button‐tapping period. Participants were then given 4 s to make a second risk or anxiety rating, reflecting any change in ratings after the tapping period. Finally, a red frame appeared around the screen to signal the outcome (shock or no shock). As in passive trials, a 2‐s pause was intermittently included between the second rating and the outcome to avoid any perceived contingency between the rating and outcome.

To elicit motivational conflict regarding repetitive threat‐neutralization behavior, a points system was introduced. Points in the task were not linked with any monetary or other external reward. Participants learned in neutralization training that the amount of tapping on each choice trial was linearly associated with more protection from shock and fewer points (see “Training” section below and see Supplement for instructions). An avatar was displayed running across the screen collecting points, with a blue “shield” displayed around the avatar. Participants could see that each time they tapped: (a) the blue “shield” grew stronger, resulting in greater protection from shock, and (b) the avatar paused, resulting in fewer total points collected. Participants were told that they gained points cumulatively across all choice trials, which were interspersed throughout the task. Participants could win from 0 to 900 points per choice trial; the total possible points throughout the task ranged from 0 to 84,600. Point totals were displayed onscreen after each “run” involving choice trials: once after the Discrimination phase, after each of 3,8‐min intervals throughout the Generalization phase, and after the Extinction phase.

3.1.4.4. Task Phases

Table 1 summarizes stimulus‐types and trial‐types in each task phase. Instructions for each task phase are detailed in the Supplement.

TABLE 1.

Counts of stimulus types and trial types.

	Acquisition phase	Training	Discrimination phase	Generalization phase	Extinction phase
Stimuli
US (% of CS+)	6 (75%)	—	3–9 (33%–100%)	6–16 (6%–89%)	—
CS+	8	—	9	18	12
GS1	—	—	—	12	—
GS2	—	—	—	12	—
GS3	—	—	—	12	—
vCS−	8	—	8	12	12
oCS−	8	—	8	12	12
Trial types
Passive trials	24	—	7	20	18
Choice trials	—	—	18	58	18

Note: Number of US presentations in Discrimination and Generalization phases can vary because participants can cancel the US by engaging in neutralization behavior. Therefore, ranges of possible total US presentations are shown.

Acquisition: The first phase of the task was the Acquisition phase, which included presentation of vCS−, oCS−, and CS+ only, with 75% of the CS+ presentations co‐terminating with shock. Participants completed 2 practice instrumental trials with squiggle shapes and anxiety/risk ratings/to orient them to the timing and sequence of these trials. Then, real trials were presented, with stimulus type and rating type presented in quasi‐randomized order. A red frame was displayed at the end of each trial, regardless of whether or not a US was presented, which provided a visual cue for when shock was possible.

Training: The remainder of the task was completed in the magnetic resonance imaging (MRI) environment. Before training, a speed test was administered to establish the individual's maximum tapping rate. Participants were given 10 speed‐test trials and asked to tap a button using their left little finger as fast as they could for 4 s. The highest number of taps was recorded as that participant's maximum rate.

Next, participants received training on how various levels of tapping increased protection from shock but reduced accrual of points. This training consisted of training trials in which participants watched the avatar run across the screen collecting points, and were instructed to tap a specific amount (via text onscreen reading “Tap”, “Tap More”, “Last Tap”, and “Don't Tap”). Stars were displayed in a scatter formation across the bottom of the screen to discourage rhythmic tapping (e.g., tapping once each time the avatar reaches a star). The opacity of the blue shield increased linearly with each button‐tap, such that the shield reached 100% opacity once the participant had tapped at 100% of his or her maximum effort (established via the speed test, described above). After each trial of practice, participants were shown the outcome of their tapping rate, including the amount of protection, displayed as the opacity of the blue shield, and the amount of points, displayed as a vertical stack of stars. Participants completed 5 trials for each of the 3 training anchors (see Supplement). Shock likelihoods were selected to enable a linear reduction in shock likelihood across binned levels of tapping. Practice anchors were chosen to lie in the middle of each binned range.

Discrimination and Generalization Phases: Participants first completed four practice choice trials with squiggle shapes and anxiety/risk/ratings. The Discrimination phase included presentation of vCS−, oCS−, and CS+, with 100% of the CS+ presentations co‐terminating with shock unless neutralized by the participant. The Generalization phase included all stimulus types with 88% of the CS+ presentations co‐terminating with shock unless neutralized by the participant. Passive trials and choice trials were interspersed throughout, including reinforced CS+ passive trials without ratings, to help participants retain the CS+/US contingency and to heighten affective arousal at initial stimulus onset.

Extinction Phase: The Extinction phase included presentation of vCS−, oCS−, and CS+, without shock. The first six CS+ trials of the Extinction phase were passive trials, so as to mitigate “protection from extinction” via neutralization (Lovibond et al. 2000).

3.1.4.5. Task Measures

Task outcome measures assessed for each stimulus type in each phase were: (a) risk and anxiety ratings taken following stimulus onset on all trials and following the tapping period on choice trials, on a scale of 1–5 with 1 = “none”, 3 = “some”, and 5 = “a lot”; and (b) extent of tapping behavior, indexed as a percentage of the participant's maximum demonstrated speed during the speed test.

3.1.4.6. Post‐Task Questionnaire

To aid in interpretation of behavioral results, after completing the task participants were asked to complete a survey about their experience, rating questions such as “How important was not getting shocked?” and “How important was gaining points?” on a 0–10 scale (see Supporting Information).

3.2. Statistical Analysis

To assess task responding across stimuli, linear mixed‐effects models (LMEs) were conducted in each phase separately, regressing ratings, neutralization behavior, and neural activity to onset and tap‐preparation on stimulus‐type (e.g., Neutralization ~ Stimulus). Post hoc pair‐wise comparisons (Welch two‐sample t‐tests) were conducted comparing CS+ to oCS− and vCS− in the Acquisition phase, to test for successful Pavlovian conditioning to the CS+. To assess task responding across trials in the Extinction phase, LMEs were conducted regressing task measure on trial, for CS+ trials only (e.g., Neutralization ~ Trial). To assess effects of repetitive threat‐neutralization on post‐neutralization risk and anxiety ratings, LMEs were conducted in each phase separately (Discrimination, Generalization, Extinction), regressing post‐neutralization rating on trial‐level tapping, stimulus‐type, and their interaction, with pre‐neutralization rating as a covariate (i.e., Post‐neutralization rating ~ Tapping × Stimulus + pre‐neutralization rating). All analyses included random effects for participant ID. For all analyses of repetitive threat‐neutralization, the percentage of maximum tapping rate was entered as the primary variable of interest, and maximum tapping rate was entered as a covariate. For behavioral analyses, a significance threshold of α = 0.05 was used.

As an exploratory analysis, to determine data‐driven profiles of task behavior, unsupervised k‐means clustering was performed using the k‐means function in R (R Core Team 2021); variables entered were stimulus‐specific means for anxiety ratings, risk ratings, and neutralization behavior in Discrimination, Generalization, and the first and second half of Extinction. Within‐cluster sums of squares were computed using the fviz_nbclust function from the factoextra package (Kassambara and Mundt 2020) in R to determine the optimal number of clusters.

3.3. Results

3.3.1. Participants

A total of 55 participants completed the task as part of a larger ongoing study. Of these, 6 were excluded for not displaying conditioning effects (risk ratings during Acquisition phase to CS+ < vCS−). A total of n = 49 participants were included in behavioral analyses presented here. Sensitivity analyses indicated that this sample would enable 95% power to detect small to medium effect sizes (f = 0.19) when comparing task behavior across stimuli. To test whether inclusion of non‐conditioner participants would substantially alter the results, primary analyses were repeated with the full sample of n = 55. Participant demographics are reported in Table 2.

TABLE 2.

Sample demographics.

	Study 1 (n = 49)		Study 2 (n = 88)		Study 3 (n = 31)
	Mean	SD	Mean	SD	Mean	SD
Age	36.24	13.38	31.2	10.08	33.11	12.01

	N	%	N	%	N	%
Biological sex
Female	40	70.18	25	71.59	22	70.97
Male	17	29.82	63	28.35	9	29.03
Race
American Indian or Alaska Native	3	5.26	—	—	1	3.23
Asian	9	15.79	6	7.05	7	22.58
Black or African American	2	3.51	2	2.35	1	3.23
Multiracial	7	12.28	0	0	7	22.58
Native Hawaiian or Pacific Islander	0	0	—	—	0	0
White	33	57.90	70	82.35	13	41.94
Other	3	5.26	5	5.88	2	6.45

Note: In the online study, the demographic survey from the Prolific platform was used, which did not offer “American Indian or Alaska Native” and “Native Hawaiian or Pacific Islander” as options to participants. In the online study, 1 participant did not provide data on biological sex and 2 participants did not provide data on race.

3.3.2. Behavioral Results

3.3.2.1. Perceived Threat, Anxiety, and Neutralization Across Stimuli

In the Acquisition phase, prior to the introduction of choice trials, significant main effects of stimulus‐type were found for risk ratings (p = 0.001, η _p ² = 0.66) and anxiety ratings (p = 0.001, η _p ² = 0.49) (Figure 3A). Post hoc pair‐wise comparisons (Welch two‐sample t‐tests) revealed that risk ratings were significantly higher for the CS+ compared to the oCS− (t[111.31] = 9.33, p < 0.001) and the vCS− (t[110.05] = 12.34, p < 0.001), as were anxiety ratings (oCS−: t[111.99] = 4.88, p < 0.001; vCS−: t[111.20] = 7.45, p < 0.001), indicating expected conditioning effects across the sample (see Supporting Information).

FIGURE 3.

Risk ratings, anxiety ratings, and neutralization behavior across stimuli. Black squares reflect the sample mean within each phase and stimulus‐type; points in color reflect each trial‐level response; violin plots are probability density plots. CS+, Threat‐cue; GS, Generalization stimulus; oCS−, Ring‐shaped safety cue; vCS−, V‐shaped safety cue.

In the Discrimination, Generalization, and Extinction phases, significant main effects of stimulus‐type were found across all task indices (ps < 0.001, η _p ² > 0.32). (Table S4, Figure 3B–D), indicating retention of conditioning effects after choice trials were introduced. In the Generalization phase, gradients of generalization (Figure 3C) indicated gradually decreasing risk ratings, anxiety ratings, and tapping as stimuli decreased in similarity from the CS+. Of note, these task effects were also seen when including those who did not show conditioning effects at Acquisition (ps < 0.001, η _p ² > 0.30; Table S14).

3.3.2.2. Changes in CS+ Responding Across Acquisition and Extinction

In the Acquisition phase, risk ratings and anxiety ratings to the CS+ increased across trials, as expected, indicating successful acquisition of aversive conditioning (see Supporting Information).

In the Extinction phase, risk (p < 0.001, η _p ² = 0.21) and anxiety (p = 0.001, η _p ² = 0.04) ratings to the former CS+ gradually decreased across trials, indicating successful extinction. Neutralization to the former CS+, which occurred on choice trials following a series of CS+ passive trials never resulting in shock, also numerically declined across trials in the expected pattern, but these effects were not significant (p = 0.100) (Figure 4).

FIGURE 4.

Effects of trial on task measures. Learning slopes are shown for the two task phases in which novel CS‐US contingencies were introduced: (A) the Acquisition phase, and (B) the Extinction phase. Trials are numbered within each phase and stimulus type; for example, in the Acquisition phase, the first CS+ trial in the Acquisition phase has an anxiety rating, the second has a risk rating, etc. In the Extinction phase, the first six CS+ trials are passive trials, so that participants have the opportunity to observe (without neutralizing) that the CS+ is no longer paired with shock.

3.3.2.3. Change in Risk and Anxiety Ratings With Repetitive Threat‐Neutralization

Regarding risk ratings, more trial‐level neutralization was associated with lower post‐neutralization risk ratings across stimuli in the Discrimination (p < 0.001, η _p ² = 0.13), Generalization (p < 0.001, η _p ² = 0.27), and Extinction (p = 0.001, η _p ² = 0.15) phases, indicating participants' perceptions that neutralization behavior reduced likelihood of shock. Significant Tapping by Stimulus‐type interactions in Discrimination and Generalization phases indicated that this effect of more neutralization on lower post‐neutralization risk ratings was strongest for CS+ (ps < 0.002). Regarding anxiety ratings, more neutralization led to lower post‐neutralization anxiety ratings in the Discrimination (p < 0.001, η _p ² = 0.10), Generalization (p < 0.001, η _p ² = 0.19), and Extinction phases (p = 0.011, η _p ² = 0.15) (Table S7). This pattern of findings also emerged when including non‐conditioner participants (Table S16).

3.3.2.4. Profiles of Behavioral Responding

K‐means clustering revealed four profiles of responding (Figure 5). Profile 1 (n = 18) was characterized by consistent differentiation between CS+ and safe stimuli in risk ratings, anxiety ratings, and neutralization, which persisted into the Extinction phase (i.e., poor extinction). Profile 2 (n = 5) was characterized by low risk and anxiety ratings and above‐average tapping across stimuli. Profile 3 (n = 24) was characterized by slightly above‐average CS+ risk ratings during Discrimination, reflecting strong acquisition of the CS+/US contingency, and relatively low ratings and neutralization across stimuli. Profile 4 (n = 10) was characterized by relatively high ratings of perceived risk and anxiety across stimuli, and below‐average neutralization behavior across stimuli.

FIGURE 5.

Profiles of task behavior. Unsupervised k‐means clustering revealed four distinct profiles. CS+, Threat cue; Disc, Discrimination phase; Ext1, Early extinction (first 6 CS+ trials); Ext2, Late extinction (second 6 CS+ trials); Gen, Generalization phase; GS, Generalization stimulus; oCS−, Ring‐shaped safety cue; vCS−, V‐shaped safety cue.

3.3.2.5. Post‐Task Questionnaire

Participants reported US aversiveness as well as motivations to win points and avoid shocks on the post‐task questionnaire. Participants' response to the question “How uncomfortable were the shocks?” indicated moderate aversiveness (M = 5.41 out of 10; SD = 2.60). Exploratory pairwise t‐tests revealed that participants reported that they were happier to receive points than to avoid shock (t[47] = 2.72, p = 0.009), found it more important to get points than to avoid shock (t[48] = 5.16, p < 0.001), liked getting points better than tapping the button (t[48] = 9.89, p < 0.001), and felt upset about both missing points and getting shocked (p = 0.706). Questionnaire items and descriptive statistics can be found in the Supporting Information.

3.4. Study 1 Summary

In this preliminary sample, the TTS Task elicited the expected behavioral pattern of graded threat‐neutralization behavior that gradually decreased across trials of the Extinction phase. We next aimed to assess the replicability of these results in a larger online sample.

4. Study 2: Online Behavioral Effects

4.1. Methods

4.1.1. Participants

A separate sample of participants completed the TTS Task online, recruited via the Prolific platform. Inclusion criteria were: ages 18–65, right‐handed, and English fluency. Participants were provided with a detailed informed consent document describing the study and providing contact information for the study team, and all participants indicated informed consent before beginning study procedures. During questionnaires administered prior to the task, three attention check items (e.g., “Please select ‘3 – a lot’”) were included; participants who responded incorrectly to two or more of these items were excluded. All online study procedures were approved by the WCG (#20234728).

4.1.2. Tap‐to‐Safety Task

The online TTS task is similar in overall structure to the in‐person task. The online study had two notable differences from the in‐person study. First, the online study used aversive pictures from the International Affective Picture System (IAPS) (Lang et al. 1997) instead of shocks as the US. Second, the online study was programmed at an earlier stage of task development and did not use reward points. Instead, participants were simply instructed that they could tap the button to reduce their risk of seeing the picture. A detailed description of the online study can be found in the Supplement. LMEs were conducted to identify differential responding across stimuli and across trials, as well as to identify changes in ratings pre‐ to post‐neutralization, as the in‐person study.

4.2. Results

Of the 217 participants who enrolled in the online study, 118 completed the task. The n = 88 who passed questionnaire attention checks and demonstrated conditioning effects (risk ratings during Acquisition phase to CS+ < vCS−) were included in the present analyses. Given that risk ratings served as an additional measure of participant attention during the online task, analyses were not repeated to include non‐conditioner participants. Sensitivity analyses indicated that this sample would enable 95% power to detect small effect sizes (f = 0.13) when comparing task behavior across stimuli. Participant demographics are shown in Table 2.

Behavioral effects largely echoed the results from the in‐person sample. Risk ratings, anxiety ratings, and neutralization behavior differed across stimuli in all task phases (Table S10), following successful acquisition of elevated risk and anxiety ratings to CS+ versus vCS− (although anxiety ratings did not significantly differ between CS+ and oCS−; Table S11). Anxiety ratings gradually increased across trials of Acquisition. Anxiety ratings, risk ratings, and neutralization behavior all gradually decreased across trials of Extinction (Table S12), consistent with expectations. In the Generalization phase, more neutralization behavior was associated with lower post‐neutralization ratings across stimuli. In the Extinction phase, this effect was seen for both post‐neutralization risk ratings and post‐neutralization anxiety ratings (Table S13).

4.3. Study 2 Summary

The online study replicated findings from the in‐person sample, indicating repetitive threat‐neutralization behavior that scaled with the threat‐relevance of stimuli. Given these promising results, we next examined the neural correlates of neutralization behavior.

5. Study 3: Neuroimaging

5.1. Methods

5.1.1. Participants

A subset of participants who completed the task as described in Study 1 also underwent brain imaging during the TTS task (n = 40). Inclusion and exclusion criteria are as described above.

5.1.2. Imaging Acquisition, Preprocessing, and Regions of Interest

A 3T Siemens MAGNETOM Prisma‐NX with a 32‐channel head coil was used for neuroimaging. Functional data were collected using an echo‐planar imaging sequence with the following parameters: TR/TE/flip = 1.5 s/27 ms/72°; matrix 96 on a 230 mm FOV with 54 slices resulting in 2.4 mm isotropic voxels; GRAPPA = 2 and multiband acceleration = 2. Phase reversed spin echo anatomic images were acquired for distortion correction; a standard T1‐MPRAGE sequence (1 mm, TR/TE/TI = 1.69 s/24 ms/1.06 s) was acquired for registration. In order to better resolve orbitofrontal structures, subjects' heads were tilted to reduce through‐plane dephasing at the ventral surface. Participants completed 5 runs of the TTS Task during a single scanning visit: the Discrimination phase (328 TR, ~8 min), the Generalization phase (3 runs of 288–312 TR each, ~7.5 min each), and the Extinction phase (432 TR's, ~11 min). T1‐weighted structural images were also collected using a 3D MPRAGE sequence for anatomical localization with the following parameters: TR = 1.69 s, TE = 24 ms, flip angle = 8°, voxel size 1 × 1 × 1 mm, FOV 25.6 cm², 192 contiguous slices.

Data were preprocessed and analyzed using Analysis of Functional NeuroImages (AFNI) (Cox 1996). A standard preprocessing pipeline was used including correcting for head motion, coregistration, normalization to MNI space, denoising, and spatial smoothing (4 mm full width half maximum (FWHM)). The Brainnetome atlas (Fan et al. 2016) was used to extract blood oxygen‐level dependent (BOLD) signal time series from seven bilateral regions of interest (ROIs) based on a priori hypotheses: dACC, anterior insula, lateral OFC, caudate, putamen, vmPFC, and amygdala (see Supplement for Brainnetome regions used for each ROI mask). Neural activity in these regions was assessed with a hemodynamic response function (4–6 s peak) in each task phase separately (Discrimination, Generalization, Extinction). Neural activity was assessed corresponding to initial stimulus presentation (1‐ or 2‐s duration between stimulus onset and rating) vs. baseline and tap‐preparation (2‐s duration corresponding to the prompt “You will soon have the option to tap the button”) versus baseline for each stimulus‐type (vCS−, oCS−, CS+ for all phases; and each GS for Generalization phase). Regressors of no interest included: rating periods, taps, shocks, motion‐related regressors, and the first four polynomial terms used to eliminate slow signal drifts. For neuroimaging analyses, the Bonferroni‐corrected significance threshold of α = 0.007 was used to correct for multiple comparisons across the seven ROIs examined.

5.2. Results

5.2.1. Participants

Neuroimaging data were available for 40 participants; of these, 5 did not display conditioning effects (risk ratings during Acquisition phase to CS+ < vCS−), 4 were excluded for fMRI quality issues (alignment, motion, extreme outlier values across task conditions). The remaining n = 31 participants were included in the present fMRI analyses. Sensitivity analysis indicated that this sample would enable 95% power to detect medium effect sizes (f = 0.23) when comparing neural activity across stimuli. This subset of participants also reported moderate US aversiveness in retrospective ratings (M = 5.65 out of 10, SD = 2.29). Participant demographics are reported in Table 2. To test whether inclusion of non‐conditioner participants would substantially alter the results, primary analyses were repeated for all n = 36 participants with usable fMRI data. Analyses were also repeated with anxiety disorder status as a covariate, which did not change any of the findings reported below.

5.2.2. Neural Activity Across Stimuli

Activity to stimulus onset in the anterior insula was elevated to the CS+ in all three phases (ps < 0.005, η _p ² > 0.15), with gradually decreasing activity across GSs in the Generalization phase (Table 3, Figure 6A). Activity to stimulus onset in other ROIs (dACC, OFC, caudate, putamen) was not significantly different across stimuli. When non‐conditioner participants were included, a similar pattern of results emerged, with heightened anterior insula activity to CS+ in Generalization and Extinction (ps = 0.002, η _p ² > 0.09; Table S17).

TABLE 3.

Effects of stimulus‐type on neural activity to stimulus onset.

	Main effect of stimulus‐type
	DF	F	p	η p 2
Discrimination
dACC	2, 60	0.86	0.426	0.03
Anterior Insula	2, 60	5.97	0.004**	0.17
lOFC	2, 60	0.25	0.781	0.01
Caudate	2, 60	4.64	0.013*	0.13
Putamen	2, 60	2.10	0.131	0.07
vmPFC	2, 60	0.18	0.838	< 0.01
Amygdala	2, 60	0.62	0.540	0.02
Generalization
dACC	5, 150	1.98	0.085	0.06
Anterior Insula	5, 150	4.30	0.001**	0.13
lOFC	5, 150	0.44	0.818	0.01
Caudate	5, 150	0.77	0.573	0.02
Putamen	5, 150	1.09	0.369	0.04
vmPFC	5, 150	2.04	0.076	0.06
Amygdala	5, 150	1.35	0.246	0.04
Extinction
dACC	2, 60	0.29	0.750	0.01
Anterior Insula	2, 60	5.85	0.005**	0.16
lOFC	2, 60	1.20	0.308	0.04
Caudate	2, 60	1.00	0.374	0.03
Putamen	2, 60	0.36	0.701	0.01
vmPFC	2, 60	2.25	0.114	0.07
Amygdala	2, 60	1.47	0.238	0.05

Note: Models took the form Task measure ~ Stimulus with random effects for participant. For Discrimination and Extinction phases, stimulus types included were CS+, oCS−, and vCS−; for Generalization phase, stimulus types included were CS+, oCS−, and vCS−, and GSs.

Abbreviations: dACC, dorsal anterior cingulate cortex; lOFC, lateral orbitofrontal cortex.

^* p < 0.05.

^** p < 0.01.

FIGURE 6.

Differential neural activity to threat cues versus safety cues. Percent signal change across task stimuli is shown for the Discrimination and Generalization phases of the task, corresponding to neural activity to (A) initial stimulus onset and (B) tap‐preparation. Brain regions showing significant CS+ versus CS− contrast during the Discrimination phase are shown. Ant. Insula, Anterior insula; dACC, Dorsal anterior cingulate cortex.

Activity to tap‐preparation in dACC, insula, caudate, and putamen in the Discrimination phase differed across stimuli (ps < 0.001, η _p ² > 0.21), with greatest activity to CS+ and gradually decreasing activity across GSs (Table 4, Figure 6B). In the Generalization phase, this effect was seen in anterior insula and putamen only (ps < 0.002, η _p ² > 0.11). The vmPFC showed an unexpected pattern of activity, involving gradually decreasing activity from vCS− across GSs, and elevated activity to CS+ (p = 0.006, η _p ² > 0.10; Figure S2). In Extinction, no significant effects of stimulus were found. Similarly, when non‐conditioner participants were included, similar patterns emerged in the caudate at Discrimination, and in dACC, anterior insula, putamen, and vmPFC at Generalization (ps < 0.007, η _p ² > 0.08; Table S18).

TABLE 4.

Effects of stimulus‐type on neural activity to tap‐preparation.

	Main effect of stimulus‐type
	DF	F	p	η p 2
Discrimination
dACC	2, 60	10.26	< 0.001***	0.25
Anterior Insula	2, 60	9.03	< 0.001***	0.23
lOFC	2, 60	4.87	0.011*	0.14
Caudate	2, 60	22.35	< 0.001***	0.43
Putamen	2, 60	12.92	< 0.001***	0.30
vmPFC	2, 60	2.05	0.137	0.06
Amygdala	2, 60	0.15	0.858	< 0.01
Generalization
dACC	5, 150	3.04	0.012*	0.09
Anterior Insula	5, 150	5.14	< 0.001***	0.15
lOFC	5, 150	1.87	0.103	0.06
Caudate	5, 150	3.23	0.008**	0.10
Putamen	5, 150	4.20	0.001**	0.12
vmPFC	5, 150	3.42	0.006**	0.10
Amygdala	5, 150	2.00	0.082	0.06
Extinction
dACC	2, 60	1.11	0.336	0.04
Anterior Insula	2, 60	0.98	0.380	0.03
lOFC	2, 60	0.58	0.560	0.02
Caudate	2, 60	0.71	0.497	0.02
Putamen	2, 60	1.67	0.197	0.05
vmPFC	2, 60	1.37	0.262	0.04
Amygdala	2, 60	0.21	0.815	< 0.01

Note: Models took the form Task measure ~ Stimulus with random effects for participant. For Discrimination and Extinction phases, stimulus types included were CS+, oCS−, and vCS−; for Generalization phase, stimulus types included were CS+, oCS−, and vCS−, and GSs.

Abbreviations: dACC, dorsal anterior cingulate cortex; lOFC, lateral orbitofrontal cortex.

^* p < 0.05.

^** p < 0.01.

^*** p < 0.001.

Whole‐brain effects of stimulus type on task activity are available at https://osf.io/thdsu/?view_only=be4d29bb672447afbd9b3355247b3984 and echo ROI results, including anterior insula activation to CS+ onset and dorsal striatal activation to CS+ tap‐preparation.

5.2.3. Neural Activity Associated With Tapping

Stimulus Onset: Neural activity to stimulus onset was not associated with tapping behavior across any regions of interest (see Table S8).

Tap‐Preparation: Neural activity to tap‐preparation in putamen was positively associated with tapping across stimuli in Discrimination and Generalization phases. This effect was also seen in the caudate in the Discrimination phase, and in the dACC and anterior insula in the Generalization phase (ps < 0.007, η _p ² > 0.05; Table 5, Figure 7). When non‐conditioner participants were included, a similar pattern emerged (ps < 0.004, η _p ² > 0.08; Table S19).

TABLE 5.

Effects of neural activity to tap‐preparation on neutralization behavior.

	Main effect of neural activity
	DF	F	p	η p 2
Discrimination
dACC	1, 57	7.56	0.008**	0.13
Anterior Insula	1, 57	0.26	0.611	0.01
lOFC	1, 57	2.36	0.130	0.03
Caudate	1, 57	10.68	0.002**	0.17
Putamen	1, 57	12.56	0.001***	0.18
vmPFC	1, 57	0.44	0.508	< 0.01
Amygdala	1, 57	3.95	0.052	0.07
Generalization
dACC	1, 144	20.24	< 0.001***	0.14
Anterior Insula	1, 144	7.81	0.006**	0.06
lOFC	1, 144	0.08	0.780	< 0.01
Caudate	1, 144	1.12	0.291	0.02
Putamen	1, 144	19.9	< 0.001***	0.16
vmPFC	1, 144	0.03	0.870	< 0.01
Amygdala	1, 144	0.15	0.696	< 0.01
Extinction
dACC	1, 57	2.30	0.135	0.08
Anterior Insula	1, 57	0.27	0.606	< 0.01
lOFC	1, 57	0.07	0.789	< 0.01
Caudate	1, 57	0.11	0.737	0.01
Putamen	1, 57	7.33	0.009**	0.15
vmPFC	1, 57	0.10	0.751	< 0.01
Amygdala	1, 57	0.51	0.480	< 0.01

Note: Models took the form Tapping (% max) ~ neural activity × stimulus with random effects for participant. For Discrimination and Extinction phases, stimulus types included were CS+, oCS−, and vCS−; for Generalization phase, stimulus types included were CS+, oCS−, and vCS−, and GSs.

Abbreviations: dACC, dorsal anterior cingulate cortex; lOFC, lateral orbitofrontal cortex.

^** p < 0.01.

^*** p < 0.001.

FIGURE 7.

Association of neural activity with neutralization behavior. Horizontal axes show neural activity to tap‐preparation, and vertical axes show average tapping (as percentage of maximum tapping speed); all values reflect averages for each stimulus‐type within each participant, in the Generalization phase. Ant. Insula, Anterior insula; dACC, Dorsal anterior cingulate cortex; %SC, Percent signal change.

5.3. Study 3 Summary

The fMRI study extended our behavioral findings, indicating that the TTS Task elicited expected activity in brain regions associated with fear excitation and decision‐making. These results also identified a pattern of action‐preparatory activity in dACC, anterior insula, and striatum that positively correlated with task behavior.

6. Discussion

This study provides an initial validation of the TTS Task as an experimental paradigm eliciting repetitive threat‐neutralization behavior. Results from in‐person and online samples indicate that the task successfully evokes graded repetitive threat‐neutralization behavior that persists even in the absence of true threat. Pre‐neutralization neural activity in the anterior insula, dACC, and dorsal striatum scaled with the threat‐relevance of stimuli and with the magnitude of neutralization, highlighting the task's utility in probing underlying neural mechanisms. Taken together, these findings provide initial evidence that the TTS Task can quantify the behavioral and neural mechanisms underlying repetitive threat‐neutralization.

6.1. Differential Behavior Across Threat and Safety Cues

Consistent with our hypotheses, we observed elevated repetitive threat‐neutralization to the threat cue relative to safety cues (during which protection from shock is unnecessary). Greater trial‐level neutralization was followed by lower post‐neutralization threat expectancy, in line with existing frameworks that emphasize the role of avoidance behavior in reducing threat expectancy (Lovibond 2006). Neutralization also generalized across safe stimuli with varying similarity to the threat cue, echoing findings from existing avoidance paradigms (Dymond et al. 2012; Meulders et al. 2020; van Meurs et al. 2014; Norbury et al. 2018; Vandael et al. 2023). Our finding that participants exhibited heightened neutralization behavior in threatening contexts, and reduced their neutralization behavior when not under threat, suggests that the TTS Task may represent an apt analogue for real‐world situations in which threat‐neutralization behaviors occur.

6.2. Fear Extinction and Reduction of Neutralization Behavior

When aversive outcomes were removed at the end of the task, threat expectancy declined accordingly, but neutralization behavior persisted, in contrast to the extinction of avoidance observed in other tasks (Urcelay and Prével 2019). This may reflect that repetitive threat‐neutralization is relatively resistant to extinction, pointing to the importance of better understanding this behavior pattern in relation to anxiety‐related psychopathology. However, it is also important to consider the relatively low number of trials in the Extinction phase of the current task, with only six passive CS+ trials followed by six choice CS+ trials (interspersed with safety cues), in contrast to previous extinction tasks with 8–20 trials of the former threat cue (e.g., Boeke et al. 2017; Krypotos and Engelhard 2018; Volders et al. 2012).

6.3. Profiles of Neutralization and Threat Reactivity

In an exploratory analysis, unsupervised k‐means clustering was applied to reveal four distinct behavioral profiles. These profiles provide additional insight into patterns of maladaptive repetitive threat‐neutralization. Two profiles were characterized by above‐average neutralization. In one of these profiles, elevated neutralization occurred only in the Extinction phase and was coupled with poor extinction of risk and anxiety ratings. In the other, elevated neutralization occurred across stimuli, despite relatively low risk and anxiety ratings. Future analyses with larger samples will test the generalizability of these profiles.

6.4. Neural Activity Associated With Threat Cues and Neutralization Behavior

Neuroimaging results shed light on the neural basis of repetitive threat‐neutralization behavior, supporting the hypothesized roles of the dACC, anterior insula, and dorsal striatum, but not amygdala, vmPFC, or OFC (Figure 1). Activity in anterior insula to initial stimulus onset scaled with threat‐relevance of presented stimuli, as has been seen in previous fear‐conditioning tasks (meta‐analyses: Fullana et al. 2016; Webler et al. 2021). The anterior insula's activity to tap preparation also scaled with threat‐relevance, and this activity was positively associated with the degree of neutralization behavior. This supports the anterior insula's hypothesized role in maintaining attention toward threats (Figure 1) (Gasquoine 2014; Uddin et al. 2013) to support adaptive threat‐neutralization behavior. Furthermore, given the anterior insula's role integrating interoception and emotional awareness (Simmons et al. 2012; Zaki et al. 2012), anterior insula activity in the present study may reflect awareness of anxiety or anticipation of the feeling of shock.

Activity in dACC and dorsal striatum (a) tracked threat‐relevance and (b) scaled with the degree of neutralization behavior. These findings are in line with prior literature on the dACC's role in conflict detection, error monitoring and action selection (Botvinick et al. 1999; Brockett et al. 2020; Camille et al. 2011; Swick and Turken 2002), its interactions with insula in visceral regulation (Namkung et al. 2017; Smith et al. 2017), and its interactions with dorsal striatum to guide decision‐making (Balleine et al. 2007; Haber 2016). The vmPFC during tap‐preparation followed an unexpected pattern of heightened activity to both threat and safety cues, and decreased activity to safe generalization stimuli; this may reflect deactivation during decision‐making in novel contexts, consistent with the vmPFC's role in the default network (Buckner and DiNicola 2019). Contrary to our hypotheses, neither the OFC nor the amygdala showed threat‐differentiated activity, nor was their activity correlated with tapping behavior. Our null findings regarding the amygdala may reflect rapid habituation of amygdala activity to threat‐cues, as has been previously observed in fear conditioning studies (LaBar et al. 1998; Sehlmeyer et al. 2009), and/or the co‐occurrence of both threat and reward processing within the a priori amygdala mask we examined (Smith and Torregrossa 2021). Regarding the OFC, a limitation of the present TTS task is that reward value (i.e., points) did not vary across trials, due to the large number of trials needed to assess generalization and extinction of conditioned fear. Future studies manipulating both threat and reward independently could better disentangle the potential role of OFC in valuation of opposing motivational cues in the task.

6.5. Evidence for Construct Validity

Expanding the validity of experimental avoidance paradigms will increase their ability to generate clinically‐useful insights (Krypotos et al. 2018; Vervliet and Raes 2013). The present findings support the construct validity, or etiologic validity, of the behavior elicited by the TTS Task, that is, its concordance with the theorized function of a real‐world behavior (Krypotos et al. 2018; Nestler and Hyman 2010). Specifically, we found that the task elicits neutralization behavior that increases in increasingly threatening contexts, and serves to momentarily lessen threat‐expectancy and anxiety, mirroring the putative function of real‐world repetitive threat‐neutralization behaviors such as checking or reassurance‐seeking. Our finding that the task activates the expected neural circuitry associated with fear excitation and decision‐making further supports construct validity (though expected results in amygdala and vmPFC were not observed). Future research in clinical and longitudinal samples is needed to establish the predictive and diagnostic validity of this task. To the extent that task behavior maps onto real‐world clinical symptoms, it can be used to test the efficacy of interventions (e.g., pharmacological, cognitive‐behavioral, neuromodulatory) that aim to reduce repetitive threat‐neutralization behavior.

6.6. Limitations

While the TTS task models repetitive threat‐neutralization behaviors that interfere with pursuit of rewards, the extent to which task performance maps onto real‐world neutralization behaviors remains unknown. In particular, the neutralization behavior in the task is confined to a relatively short duration (4 s per trial). Whether the behavioral patterns observed in the task generalize to behavioral phenomena that play out over minutes or hours remains an open empirical question that warrants further investigation. Future analyses with clinical populations will establish the clinical relevance of the task by linking neurobehavioral task indices with specific symptom presentations. An additional limitation is that the current task does not manipulate threat probability and reward value independently. The separate roles of reward motivation (i.e., to gain points) and threat motivation (i.e., to avoid shock) are therefore difficult to disentangle in the present data. Finally, extinction findings are limited by the relatively brief duration of the Extinction phase; more trials of the former threat cue, and more choice trials following extinction learning, could enable a greater degree of extinction and more opportunity for examination of individual differences in extinction.

7. Conclusions

These findings support the use of the TTS Task to assess repetitive threat‐neutralization behavior and its neural underpinnings. In particular, this task can enable investigations of maladaptive repetitive threat‐neutralization in anxiety disorders, obsessive‐compulsive disorder, and other conditions. Future studies with clinical populations are needed to establish the clinical relevance of the task and identify modifiable targets for treatment.

Author Contributions

H.B. led the conceptualization and design of the Tap‐to‐Safety Task, interpreted findings, and wrote the main manuscript text. H.B., R.R., R.K., A.R., and N.M. performed data analysis. A.R. programmed the Tap‐to‐Safety Task. All authors contributed to the design of the Tap‐to‐Safety Task; All authors reviewed the manuscript.

Funding

This work was funded by the Laureate Institute for Brain Research and the National Institute for General Medical Sciences (P20 GM121312).

Ethics Statement

My article reports human subjects. Recruitment meets scientific requirements & HBM's expectation of inclusivity.

Conflicts of Interest

Dr. Martin Paulus has received royalties for an article about methamphetamine published in UpToDate. In the last 3 years, Dr. Simpson has received a stipend from the American Medical Association to serve as Associate Editor of JAMA Psychiatry and royalties from Cambridge University Press and UpToDate Inc. The other authors declare no conflicts of interest.

Supporting information

Data S1: Supporting Information.

Berg, H. , Rozniarek R., Robinson A., et al. 2026. “The Tap‐to‐Safety Task: A Novel fMRI Paradigm Assessing Repetitive Threat‐Neutralization.” Human Brain Mapping 47, no. 3: e70473. 10.1002/hbm.70473.

Data Availability Statement

De‐identified data and analysis code are available at https://osf.io/thdsu/?view_only=be4d29bb672447afbd9b3355247b3984. Code for programming the Tap‐to‐Safety Task may be made available upon request.