Sage Journals: Discover world-class research

Abstract

Exposure treatment involves systematic confrontation with fear-inducing stimuli, effectively reducing fear and anxiety. However, a significant number of clients still experience a return of fear (ROF) after treatment. This study investigates whether incorporating an approach component during fear extinction, a laboratory exposure analog, could mitigate this return of fear. Furthermore, we explored the underlying mechanisms by drawing on predictions from the inhibitory learning theory and the reflective-impulsive model of behavior. In a within-subjects design, we compared instructed active approach of a stimulus during extinction to more passive non-avoidance. Contrary to expectations, our findings revealed that performing approach behavior during extinction did not reduce ROF when compared to non-avoidance. Furthermore, valence and action tendencies, which were potential mechanisms based on the reflective-impulsive model of behavior, remained unaltered. Still, a noteworthy discovery emerged in the form of increased threat expectancies for the approached stimulus during extinction, suggesting a heightened level of expectancy violation, as predicted on the basis of the inhibitory learning theory. These findings offer valuable insights into the intricate relationship between approach behavior, ROF, and underlying mechanisms, highlighting the need for further research to assess the potential benefits of emphasizing approach in exposure treatment.

Keywords

Exposure fear extinction approach avoidance anxiety

Introduction

Exposure, a cornerstone of cognitive-behavioral therapy, involves systematic confrontation with fear-eliciting stimuli. This method is highly effective in reducing fear and anxiety (Olatunji et al., 2010) and is considered to be the first-choice intervention for anxiety disorders (Kaczkurkin & Foa, 2015). However, despite its efficacy, a significant number of clients experience a return of fear (ROF) after treatment, which for some results in a full relapse (Springer et al., 2018; Vervliet et al., 2013). Consequently, a key focus of exposure-related research lies in the identification of strategies that reduce ROF, thereby optimizing the retention of exposure outcomes (for a recent review, see Craske et al., 2022).

During exposure, clients are typically encouraged to actively and intentionally engage with fear-eliciting stimuli that were previously avoided, either physically (e.g., moving closer), or cognitively (e.g., by focusing attention on the stimulus). This shift in behavior from avoidance to approach is regarded by some to be a primary target in clinical treatment (Hayes et al., 1999). Even though approach behavior is an inherent part of exposure therapy, only a limited number of studies have examined to what extent it contributes to the effects of exposure and whether encouraging additional approach during exposure can further optimize treatment. Wolitzky and Telch (2009), for instance, found in individuals with fear of heights that adding fear antagonistic action (incongruent with avoidance and involving active approach) during exposure resulted in noteworthy reductions in fear, compared to exposure therapy without such actions. These reductions were observed both immediately post-treatment and 1 month after treatment (i.e., ROF). In contrast, in a study by Van Uijen et al. (2015), no evidence was found supporting the augmentation of exposure effects through active approach. It should be noted that this study focused solely on immediate posttest effects, with no measurement of long-term outcomes. These contrasting outcomes between the studies of Wolitzky and Telch (2009) and Van Uijen et al. (2015) emphasize the importance of continued research in this area.

Additionally, an open question remains which underlying mechanisms can explain the proposed beneficial effects of approach behavior in exposure therapy. We discuss two theoretical models that generate hypotheses about the mechanisms through which active approach may enhance the efficacy of exposure. A first candidate mechanism is rooted in the inhibitory learning model of exposure (Bouton, 1993; Craske et al., 2008, 2014, 2022). According to this model, repeated exposure to a fear-inducing stimulus (conditioned stimulus; CS) without occurrence of the expected aversive consequences (unconditioned stimulus; US) leads to the formation of new inhibitory associations (CS-noUS) that compete with and suppress the original threat association (CS-US), resulting in fear reduction. Within this framework, expectancy violation is presumed to drive the formation of inhibitory associations: the greater the degree of expectancy violation or discrepancies between the expected and actual outcomes, the stronger the inhibitory associations.

Approaching a stimulus can potentially increase opportunities to violate expectancies, as it is incompatible with avoidance and safety behaviors (i.e., protection from extinction; Lovibond et al., 2009). Indeed, if individuals actively approach a feared stimulus, for example, by coming closer and focusing attention to the stimulus, it becomes more difficult to avoid at the same time. In support of this notion, using laboratory exposure analogs, Pittig (2019) and Pittig and Wong (2021) have found evidence that instructing and incentivizing non-avoidance during exposure leads to significant improvements in inhibitory learning, as indicated by weaker ROF (decreased US-expectancy ratings and fear responses) when safety behavior became unavailable

Arguably, actively approaching a feared stimulus goes beyond mere non-avoidance. It involves an increase in proximity to and/or an increase in emotional engagement with it. This heightened engagement may increase the perceived probability of the aversive outcome (US). For example, the perceived probability of being bitten by a dog might increase when actively approaching the dog. Similarly, one might perceive the probability of not being able to tolerate anxiety as higher while maintaining focused attention on the feared stimulus. As a result, heightened US-expectancies when approaching a fear stimulus can result in more substantial expectancy violation when the anticipated US does not occur, thereby facilitating stronger inhibitory learning (Tolin, 2019; Weisman & Rodebaugh, 2018).

Two additional candidate working mechanisms draw from the reflective-impulsive model of behavior (RIM; Deutsch & Strack, 2020). In this model, valence and action tendencies are inseparably related. In particular, approach tendencies reflect an automatic inclination (generated by an implicit system) to engage with positive or rewarding stimuli. Furthermore, the RIM suggests that any input signaling approach can induce a corresponding positive valence and approach tendency. Engaging in approach behavior can be considered a powerful implicit approach-signaling input. Early accounts reserve a central role for arm movements in this process. In essence, as arm flexion is strongly associated with acquisition or consumption of desired stimuli, flexing the arm when confronted with a stimulus could signal that stimulus is positive and worthy of approach (Cacioppo et al., 1993). More recent accounts propose that simply providing evaluative labeling of the behavior may suffice, for example, framing the approach response as positive and/or distance-altering (Mertens et al., 2018). As such, the model suggests that incorporating adequately framed approach training in exposure therapy may offer dual benefits, implicitly impacting both the valence of fear-inducing stimuli and automatic action tendencies.

The first potential working mechanism derived from this model concerns the positive impact of approach on the valence of fear-inducing stimuli. Research has demonstrated that stimuli that were approached were rated more positively than stimuli that were avoided (Van Dessel et al., 2016). This effect has also been observed for fear-related stimuli, as demonstrated by Jones et al. (2013). In their study, spider fearful and non-fearful individuals who approached subliminally administered images of spiders reported more positive evaluations of spiders. In turn, enhanced positive valence of CSs may reduce susceptibility to return of fear (Dour et al., 2016; Zbozinek et al., 2015; although see: Van Dis et al., 2019), therefore improving exposure outcomes.

As a second potential working mechanism, the RIM suggests that approach training can change automatic action tendencies toward approach (Cacioppo et al., 1993). This approach orientation is, according to the model, one of two key elements in shaping overt behavior (the other being cognitive decision-making). Hence, by approaching a fear-eliciting stimulus, action tendencies toward that stimulus may change from avoidance to approach tendencies, which in turn play a part in facilitating future overt approach behavior toward the stimulus or novel stimuli. Supporting evidence comes from Amir and colleagues’ (2013) investigation of training approach behavior in individuals with contamination fear. They observed that participants who were consistently prompted to approach contamination-related images with a joystick showed stronger approach tendencies toward novel contamination-related images and a greater overt approach toward a variety of physical contaminants. Notably, their study design did not directly assess the impact of approach on fear levels. Furthermore, Krypotos and colleagues (2015) added approach training to the CS after fear extinction (a laboratory exposure analog). They found that this training successfully altered approach latencies during the task, which partially generalized to an assessment of action tendencies at the end of the session. However, no impact on return of fear was observed.

The theoretical models discussed thus far provide hypotheses for understanding the impact of approach training on exposure outcomes. The inhibitory learning model emphasizes a mechanistic impact of approach behavior on learning processes, particularly through enhanced expectancy violation. In contrast, the reflective-impulsive model focuses on the influence of approach behavior on the implicit system, suggesting that it affects both the valence and automatic action tendencies toward fear-inducing stimuli. However, empirical evidence regarding these proposed mechanisms, as well as evidence regarding the extent to which approach actually enhances exposure effects, remains limited.

To address this gap in knowledge, the present study aims to investigate the influence of instructing physical approach behavior (specifically: active engagement with the stimulus, resulting in increased proximity) on ROF (spontaneous recovery and reinstatement) in a fear extinction procedure. We achieve this by comparing extinction in three distinct scenarios. In one scenario, participants were instructed to actively approach a CS (CS_zoom) by moving a joystick, resulting in the CS becoming larger. Half of the participants approached by flexing their arm muscles (“pulling the image closer”), and the other half approached them by extending their arm muscles (“zooming in”). We contrasted this approach scenario to two non-avoidance scenarios, where participants did not actively engage with the CS. In the first non-avoidance scenario, the CS (CS_yoked) became larger without any active intervention by the participant. This allowed us to contrast active approach and more passive non-avoidance in two scenarios where the visual input was the same (i.e., the CS increased in size), with only behavior differing. In the second non-avoidance scenario, the CS (CS_unmoving) did not increase in size, remaining in its original size throughout the procedure. Additionally to assessing differential effects on ROF, our study aimed to explore three candidate mechanisms previously introduced: expectancy violation, change in valence, and change in action tendencies.

We hypothesized that the stimulus that was actively approached during extinction would evoke (1) reduced spontaneous recovery and (2) reduced reinstatement after extinction, compared to both non-avoidance stimuli. Furthermore, we predicted that approaching this stimulus would implicitly (3) increase the perceived valence of the stimulus and (4) modify conditioned action tendencies, facilitating quicker responses when prompted to approach the approached CS in comparison to the other stimuli. Finally, we conducted exploratory analyses to investigate three aspects: (5) whether flexing the arm muscle to approach the CS was essential for eliciting an effect on extinction outcomes, (6) whether US-expectancies during the extinction phase progressed differently for the three stimuli (which may be indicative of differential expectancy violation), and (7) whether the proposed mechanisms (valence, action tendencies, and US-expectancies during extinction) significantly correlated with spontaneous recovery and reinstatement, providing preliminary insight into their role in modifying return of fear.

Method

Participants

A sample of 77 participants (Table 1) was recruited through university-related Facebook groups and an online participant database of KU Leuven. Exclusion criteria were provided during recruitment and reassessed during the first session. These included being under 17 years of age, being pregnant, having heart conditions, other cardiovascular problems, neurological conditions, or severe medical conditions in general, having been recommended by a medical professional to avoid stressful situations, having an electronic implant, and having pain or a condition of the hand or wrist.

Table 1.

Participant demographics.

Characteristic	n	%
Gender
Male	16	20.78
Female	60	77.92
Declined to answer	1	1.30
Age
17–20	57	74.03
20–29	16	20.78
30–39	1	1.30
40–49	1	1.30
60+	1	1.30
Declined to answer	1	1.30

Five participants did not return after one (N = 3) or two (N = 2) of the three sessions, resulting in a sample size of N = 74 for the final research question (6), and N = 72 for the other research questions. The minimal sample size of 72 was chosen in order to maximize counterbalancing across subjects and because it offers adequate power to detect an effect of F = 0.18 (small to medium in size) in the primary analyses; repeated measures ANOVA’s containing three levels (with a conservative correlation of 0.3), when aiming for 80% power at a significance level of α = .05 (N = 71).

Participants were compensated with a monetary reward or course credit for their participation in the study. Ethical approval for this study was obtained from the KU Leuven Social and Societal Ethics Committee (G-2021–4428). Preregistration of the hypotheses and analyses, original data, and analysis scripts are available at https://osf.io/6dtuv/.

Materials and measures

Stimuli

Throughout the experiment, images of a snake, a bear, and a shark served as CSs. These images were selected from the International Affective Picture System (respectively image 1120, 1321, and 1930; Lang et al., 2008), based on their fear-relevance and similar valence and arousal ratings, with the aim to intensify acquisition (Carpenter et al., 2019) and to elicit defensive responses. Additionally, all three animals are known to attack, allowing for the framing of the electrical stimulus as a bite, and making the approach response more face valid and meaningful. During the experiment, the images were sized 75 mm × 75 mm, with a maximum expansion to 1650 mm × 1650 mm during extinction. The precise stimulus category that specific images were assigned to (CS_zoom, CS_yoked, and CS_unmoving) was counterbalanced across participants.

A 2-ms electrical stimulus, delivered via a constant-current stimulator (Digitimer DS7A) to the wrist of the left hand, served as the unconditioned stimulus (US). The intensity of the stimulus was individually chosen in a work-up procedure (see: 2.3. Procedure). Conductive K-Y gel was applied between the electrode bar and the skin.

US-expectancy and distress ratings

Participants rated their expectancy about the occurrence of the electrical stimulus in a 7-second window, 2–4 seconds after CS onset. These ratings were provided on a visual analog scale (VAS) ranging from 0 (“certainly no bite”), over 5 (“uncertain”), to 10 (“certainly a bite”). Distress was rated in a subsequent 5-second window, on a VAS scale ranging from 0 (“no distress at all”) to 10 (“a large amount of distress”).

CS valence

The valence of each CS was rated on a VAS ranging from 0 (“not unpleasant at all”), over 5 (“neutral”), to 10 (“very unpleasant”). Valence ratings were assessed at the start and end of each session, and after the reinstatement phase (see Figure 1).

Figure 1.

Overview of the experimental phases. Note. CS, conditional stimulus; US, unconditional stimulus.

Skin conductance (SC)

SC was assessed continuously throughout the three sessions using two disposable EL507 Ag/AgCl snap electrodes attached to the hypothenar eminence (Biopac Systems Inc., USA). The signal was recorded in microsiemens (µS) at a 2000 Hz sampling rate, using a Biopac MP160 system with Biopac Acqknowledge 5.0 software (Biopac Systems Inc., USA). In subsequent processing, it was downsampled to 1000 Hz and filtered through a 10 Hz low-pass zero-phase Butterworth filter.

Fear potentiated startle (FPS). Movement of the left orbicularis oculi was recorded continuously at a sampling rate of 2000 Hz via electromyography (EMG). Two 4 mm Ag/AgCl electrodes (Biopac Systems Inc., USA) filled with electrolyte gel (Signa Gel; Parker Labs, USA) were positioned approximately 1 cm under the pupil and 1 cm below the lateral canthus. A third electrode was served as the ground electrode and was placed on the middle of the forehead. The raw EMG signal (measured in microvolts; μV) was amplified online, and then filtered (28–500 Hz; zero-phase Butterworth), rectified, and smoothed (40 Hz low-pass filter) in accordance to published guidelines (Blumenthal et al., 2005). Startle responses were evoked by administering a white-noise probe (50 ms, 100 dB) bilaterally through a set of headphones.

Procedure

The study involved three 45-minute sessions that were conducted over three consecutive days. On the first day, participants underwent fear acquisition training, followed by a fear extinction procedure on the second day. The third day included a spontaneous recovery test, a reinstatement test, and a manikin approach-avoidance test. An overview of the procedure is displayed in Figure 1.

Day 1: Preparation

At the start of the first session, participants provided written informed consent and were screened for the exclusion criteria. Subsequently, electrodes were attached, followed by a US calibration procedure (also work-up procedure). Starting with a light stimulus of 2 mA, the intensity was increased gradually, and participants were instructed to stop at a level that was “highly uncomfortable but not painful.” Subsequently, participants rated the valence of the CS images. The session continued with a startle habituation phase, containing 8 startle probes. Then, the acquisition phase started.

Day 1: Fear acquisition training

At the start of acquisition training, participants received instructions that during the task, photos of animals would be presented and that some of these animals might bite (as evidenced by the electrical stimulus). They were further instructed to try to predict if the depicted animal would bite or not.

During acquisition, each CS was presented eight times. Stimuli were presented in eight blocks of four trials: Each block consisted of one trial per stimulus (CS_zoom, CS_yoked, and CS_unmoving), and one noise-alone (NA) trial presented in a random order. In the first and last two blocks, each of the CSs was always paired with the US. In the four middle blocks, two of the four presentations of each CS were reinforced randomly, resulting in a 75% reinforcement rate (i.e., per CS, 6 out of 8 presentations were reinforced). Each trial had a duration of 16 seconds. Two seconds after CS onset, the expectancy rating scale was presented (7 seconds), followed by the distress rating scale (5 seconds). A startle probe was presented 14.5 seconds after CS onset. On reinforced trials, a US followed 15.5 seconds after CS onset. Inter-trial intervals (ITIs) ranged from 17 to 23 seconds, with a mean of 20 seconds. On NA trials, the startle probe was presented immediately at trial onset, in the presence of a fixation cross, and followed by a 1.5-second interval and the subsequent ITI. CS valence was reassessed at the end of the session, after the acquisition training.

Day 2 and 3: Preparatory steps

The second and third sessions commenced with the preparation and attachment of electrodes (Ag/AgCl electrodes and electrical stimulation bar), followed by the startle habituation phase. The US was not recalibrated, in order to avoid inducing unwanted reinstatement effects (Haaker et al., 2014).

Day 2: Extinction

The extinction procedure started with a practice block, in order to be able to provide feedback on whether they performed the approach behavior by using the joystick correctly. This practice block featured a novel stimulus (a set of meerkats) that was not used in the rest of the experiment. This part was followed by 12 extinction blocks. Each extinction block consisted of one unreinforced CS trial per stimulus (CS_zoom, CS_yoked, and CS_unmoving) and one NA trial, presented in a random order.

During the extinction phase, the CSs prompted different actions from the participants. For one of the images (CS_zoom), participants were cued to zoom in on the image during the first four seconds of its presentation using a joystick, causing the image to increase in size. Half of the participants zoomed in by pushing the joystick away from themselves (arm extension), while the other half did so by pulling the joystick toward themselves (arm flexion). Respective instructions regarding the cue (blue arrow) at the start of the task read: “zoom in, in order to come closer to the animal” and “pull the animal toward you.” A second image served as a yoked stimulus (CS_yoked). It zoomed in without action by the participant, at a speed that was yoked to their response on the last CS_zoom trial. A third image lacked the prompt to move and remained stationary (CS_unmoving). Notably, not all blocks contained a prompt to approach the CS_zoom. In the first two blocks plus one randomly chosen block, this prompt was lacking, in order to prevent that participants would attribute the non-occurrence of the US to the approach behavior (conditional inhibitor; Rescorla, 1969). This resulted in an extinction phase with approach prompts in 75% of CS_zoom trials¹.

After the four-second window allowing for approach behavior, the trial continued as described under “Day 1: Fear Acquisition Training.”

Day 3: Spontaneous recovery and reinstatement test

The third session commenced with a spontaneous recovery test, consisting of four blocks, each containing one unreinforced trial per CS and one NA trial. Stimulus order was counterbalanced across participants in the first block (always commencing with an NA trial, followed by a fixed order of CSs), and random in the three subsequent blocks. The trial flow was identical to that of the acquisition phase.

Immediately after the spontaneous recovery test, three unsignaled USs were delivered, spread 5 seconds apart, followed by a reinstatement test phase. The design of this phase (including randomization), mimicked the spontaneous recovery phase.

Day 3: Manikin approach-avoidance test

After the reinstatement test phase, the participants underwent a manikin approach-avoidance test (AAT; Krypotos et al., 2015) to assess action tendencies. The task consisted of two phases, each containing 2 practice blocks (to train correct responses) and 4 test blocks. The practice blocks consisted of 2 trials, displaying a vertical and a horizontal empty gray frame. Test blocks contained six trials, displaying each stimulus (CS_zoom, CS_yoked, and CS_unmoving) twice. One stimulus was presented in a horizontal white frame, and the other in a vertical white frame. Stimulus order within practice and test blocks was randomized.

Each trial commenced with the appearance of a white manikin figure, situated on the left or right side of the screen. After 1.5 seconds, the empty gray (practice) or white frame with stimulus (test) was presented on the opposite side of the screen, and the manikin could be moved. Participants were instructed beforehand to move the manikin toward the landscape-oriented frame and away from the portrait-oriented frame, or vice versa (counterbalanced). After the first phase, halfway through the task, the instructions were switched. Participants could move the manikin to the left or right by pressing the corresponding arrows on the keyboard three times. The manikin and the frame disappeared after 3.5 seconds. If the participant’s response was incorrect, a red cross appeared at the manikin’s starting position for 0.5 seconds. The duration of the ITIs was 2 seconds. Reaction times assessed action tendencies.

Data reduction

Skin conductance responses (SCRs) were computed by subtracting the mean value in a 2-second period before CS onset (i.e., baseline) from the maximum value obtained in the 0.5 to 14.5–16.5-second window following CS, ending at the presentation of the startle probe (Boucsein et al., 2012). The latter window was used for analyses of the extinction trials and accommodated for the additional 2 seconds that were provided to zoom in on the stimulus. Trials with negative responses were recoded to zero and included in all subsequent analyses. A square root transformation normalized the distribution (Dawson et al., 2007). Responses from two consecutive trials of a stimulus were averaged in order to reduce noise (Pittig, 2019).

Startle amplitudes were computed per trial by subtracting the mean value during a 20-ms baseline (0–20 milliseconds following probe onset) from the peak value in a 21–200 millisecond window following probe onset (Blumenthal et al., 2005). Negative responses were scored as 0 and included in all analyses. Responses from two consecutive trials of a stimulus were averaged in order to reduce noise.

For the AAT, incorrect responses were excluded from the analysis. Subsequently, median reaction times (RTs) were calculated for each stimulus (CS_zoom, CS_yoked, and CS_unmoving) by prompt (approach, avoid; Krypotos et al., 2015).

Statistical analyses

The main analyses focused on the four outcome measures (i.e., US-expectancies, distress ratings, SC, and FPS) during the spontaneous recovery (1) and reinstatement (2) test phases. For each phase and per outcome measure, a separate repeated measures ANOVA was conducted on the first trial, with Stimulus (CS_zoom, CS_yoked, and CS_unmoving) as the within-subject factor. After conducting these frequentist analyses, Bayesian analyses were performed to complement them. The inclusion Bayes factor (BF_incl) quantified the evidence for the Stimulus effect, compared to the null model (Outcome ∼ Participant). We expected to find a significant effect of Stimulus in both phases, with follow-up contrasts revealing lower fear responding to the CS_zoom compared to the other CSs.

Furthermore, exploratory analyses were conducted to test for return of fear responding by comparing the first trial of spontaneous recovery and reinstatement test with the end of extinction. To this end, a repeated measures ANOVA was conducted for each of the phases (spontaneous recovery and reinstatement), with Trial (last trial of extinction, first trial of the respective phase) and Stimulus (CS_zoom, CS_yoked, and CS_unmoving) as within-subject factors. The complementary BF_incl quantified the evidence for the Stimulus and Trial effect, respectively, compared to the null model (Outcome ∼ Participant), as well as the evidence for the Stimulus × Trial interaction, compared to a model containing only the main effects (Outcome ∼ Stimulus + Trial + Participant). SC was not included in these analyses, as the responses on the last extinction trial were corrupted by movement toward the CS_zoom.

To examine potential working mechanisms by which approach influences extinction, we assessed differences between the stimuli in terms of valence ratings (3) and action tendencies (4). Valence ratings across the entire experiment were analyzed using a repeated measures ANOVA with factors Stimulus (CS_zoom, CS_yoked, and CS_unmoving) and Time (pre-acquisition, post-acquisition, pre-extinction, post-extinction, pre-ROF, post-ROF, and post-AAT), with complementary inclusion Bayes factors (as described above). We expected to find significant main effects of both factors, as well as a significant interaction effect. Post-hoc exploratory pairwise comparisons with Bonferroni-corrected p-values were performed to assess changes in valence ratings from each notable assessment point to the next (pre- to post-acquisition, pre- to post-extinction, and post-extinction to pre-ROF). To assess the transfer of approach training from the extinction phase to participants’ action tendencies, we compared reaction times on the AAT using a repeated measures ANOVA with factors Stimulus (CS_zoom, CS_yoked, and CS_unmoving) and Response (approach and avoid), as well as inclusion Bayes factors. We expected to find facilitated approach (reflected in lower reaction times) toward the CS_zoom in comparison to CS_yoked and CS_unmoving.

Furthermore, a first exploratory analysis (5) assessed if the effect of approach on return of fear, valence and action tendencies was dependent on the specific arm movement to zoom in or approach (arm flexion vs. arm extension). Therefore, the four main analyses were rerun with an additional between-subjects factor: Group (pull-CS_zoom vs. push-CS_zoom).

A second exploratory analysis focused on comparing US-expectancy ratings toward the CSs throughout the extinction phase (6) using a nonlinear regression model with predictors Trial (3–12), Stimulus (CS_zoom, CS_yoked, and CS_unmoving), and their interactions. An F-test compared the fit (i.e., residual sum of squares) of different regression models: initially comparing models with and without the Stimulus variable, and subsequently assessing the significance of including interaction terms between Trial and Stimulus variables. Subsequently, in the best-fitting model, regression coefficients were examined to compare the intercepts and slopes of responses to the stimuli throughout extinction. The first trial and second trial of extinction were not included in these analyses, as they lacked approach prompts.

In a final exploratory analysis (7), we examined the correlation between summary measures of valence, approach tendencies, and expectancies during extinction (referred to as proposed mechanisms) and measures of return of fear (ROF). First, we computed summary indices for each participant for all proposed mechanisms and ROF measures. The valence index represented the average change in valence from pre- to post-extinction. The approach tendency index was calculated as the average response time to approach minus the average response time to avoid. The extinction expectancy index was derived from the average expectancy rating throughout extinction. ROF indices were computed in various ways, aligning with the primary analyses described earlier. Eight main ROF indices captured responses on the first trial of the spontaneous recovery and reinstatement test phase for each measure (i.e., US-expectancies, distress ratings, SC, and FPS). Additionally, six secondary ROF indices measured the increase in responding (for US-expectancies, distress ratings, and FPS) from the final trial of extinction to the first trial of the spontaneous recovery and reinstatement test phase.

Results

Graphs for US-expectancy, distress ratings, FPS, and SCR per stimulus and experimental phase are shown in Figure 2.

Figure 2.

US-expectancy, distress, FPS, and SCR throughout the phases. Note. Error bars represent the 95% confidence interval.

Spontaneous recovery

Responding on the first trial of the spontaneous recovery test phase did not significantly differ between stimuli on any of the measures. Additionally, Bayes factors favored the absence of stimulus differences: US-expectancy, F (1.77, 125.78) = 0.30, p = .714, BF_incl = 0.06; distress, F (1.82, 129.39) = 0.06, p = .925, BF_incl = 0.05; FPS, F (1.78, 126.68) = 2.85, p = .067, BF_incl = 0.58; and SCR, F (2, 142) = 0.90, p = .409, BF_incl = 0.11.

Subsequent exploratory analyses compared the last trial of extinction to the first trial of spontaneous recovery. For each of the three measures, these analyses revealed a significant main effect of Trial: US-expectancy, F (1, 70) = 128.37, p < .001, η² = .30, BF_incl > 1000, distress, F (1, 70) = 79.71, p < .001, η² = .22, BF_incl > 1000, and FPS, F (1, 71) = 51.99, p < .001, η² = .07, BF_incl > 1000, suggesting a significant return of fear at the start of the spontaneous recovery phase (across CSs). Additionally, these analyses revealed a main effect of Stimulus, although only for US-expectancy ratings, F (1.82, 127.46) = 4.18, p = .021, η² = .00, BF_incl = 0.07, and FPS, F (1.76, 125.31) = 3.27, p = .047, η² = .00, BF_incl = 0.13, with Bayes factors arguing against an effect of Stimulus. Most importantly, no significant interaction effect was found between Trial and Stimulus (for distress: p = .058, BF_incl = 0.11; for all other measures p > .10, BF_incl < 0.10). This suggests that the increase in fear responding from the end of extinction to the first trial of spontaneous recovery did not differ significantly between the CS_zoom and control stimuli.

Reinstatement

Similarly to extinction retention, responding did not significantly differ between stimuli on the first trial of the reinstatement test phase for US-expectancy, F (1.73, 122.88) = 0.76, p = .454, BF_incl = 0.10; distress, F (1.69, 120) = 0.23, p = .759, BF_incl = 0.06; FPS, F (1.83, 130.01) = 1.50, p = .228, BF_incl = 0.18; or SCR, F (2, 142) = 0.98, p = .377, BF_incl = 0.11. Again, Bayes factors favored the absence of stimulus differences.

Subsequent exploratory analyses on the final trial of extinction and the first trial of reinstatement revealed significant main effects of Trial: for expectancy, F (1, 70) = 128.37, p < .001, η² = .30, BF_incl > 1000; distress, F (1, 70) = 79.71, p < .001, η² = .22, BF_incl > 1000; and FPS, F (1, 71) = 51.99, p < .001, η² = .07 BF_incl > 1000, suggesting significant reinstatement in fear responding. No significant effect of Stimulus was found (all p-values >.10), with Bayes factors favoring the absence of an effect (all BF_incl < 0.14). Notably, a significant interaction between Trial and Stimulus was found for distress ratings, F (1.66, 116.05) = 4.71, p = .016, η² = .00, BF_incl = 0.20. Here, the increase in responding was smallest for the CS_zoom and largest for the CS_unmoving². However, this interaction effect was not supported by the corresponding Bayes factor.

CS valence

A two-way repeated measure ANOVA was performed to assess differences between stimuli regarding valence over time (Figure 3). There was a significant main effect of Time, F (3.11, 214.86) = 51.15, p < .001, η² = .19, BF_incl > 1000; but no significant main effect of Stimulus, F (1.73, 119.33) = 1.13, p = .320, BF_incl = 0.03; nor a significant interaction between the two, F (6.03, 415.87) = 0.88, p = .510, BF_incl < 0.001. Exploratory pairwise comparisons with Bonferroni adjustments revealed a significant decrease in valence across the three stimuli from pre-to post-acquisition; t (213) = −7.92, p_adj < .001, d = −0.54, and a significant increase from pre-to post-extinction; t (214) = 10.82, p_adj < .001, d = 0.74. It then significantly decreased again, from post-extinction to pre-return of fear test (spontaneous recovery and reinstatement); t (215) = −7.13, p_adj < .001, d = −0.49.

Figure 3.

CS valence ratings throughout the experiment.

Action tendencies

A repeated measure ANOVA assessed reaction times when approaching versus avoiding the different stimuli during the AAT. This analysis revealed significant a main effect of Response, F (1, 70) = 49.99, p < .001, η² = .04, BF_incl > 1000, with longer reaction times for avoidance compared to approach, suggesting a general motivational inclination to approach instead of avoid the three stimuli. No significant main effect was found for Stimulus, F (2, 140) = 2.99, p = .053, BF_incl = 0.32, nor a significant interaction between Response and Stimulus, F (2, 140) = 0.20, p = .821, BF_incl = 0.06.

Specific approach movement

In order to assess if the effect of approach on return of fear, valence and action tendencies was dependent on the specific arm movement that was made to approach, the main analyses were rerun with an additional between-subjects factor: Group (arm flexion vs. extension). For each of the measures during the spontaneous recovery test, these analyses did not reveal any main effect for Group, nor a significant interaction between Group and Stimulus (all p-values >.100). In line with this, related Bayes factors favored the absence of stimulus differences over their presence (all BF_incl < 0.62).

Similarly, no significant main effect for Group was found on any of the measures during reinstatement test (all p-values >.4, all BF_incl < 0.46), and no significant interactions between Group and Stimulus were found for US-expectancy, distress, and FPS (all p-values >.40, all BF_incl < 0.18). However, a significant interaction was found in SC, F (2, 140) = 3.82, p = .024, η² = .02, BF_incl = 2.09. Post-hoc analyses with Bonferroni corrections revealed that the effect of Stimulus on SC was only significant when the joystick had been pushed (arm extension) to zoom in on the image F (2, 70) = 5.19, p_adj = 0.016, η² = .04, BF_incl = 4.98, with significantly lower SC to the CS_zoom compared to the CS_yoked t (35) = 3.03, p_adj = .014, d = 0.51. Other comparisons were not significant.

Furthermore, we did not find a significant main effect of Group, F (1, 68) = 2.26, p = .137, BF_incl = 0.31, nor a significant interaction between Group and Stimulus, F (1.73, 117.83) = 0.32, p = .694, BF_incl = 0.04, in the valence ratings throughout the experiment.

Finally, neither effect was found for action tendencies measured at the end of the experiment: Group, F (1, 60) = 0.03, p = .863, BF_incl = 0.31; Group x Stimulus, F (2, 120) = 1.35, p = .264, BF_incl = 0.04.

US-expectancies throughout extinction

In a second exploratory analysis, we investigated the differences between stimuli on expectancy during extinction using a quadratic regression model. We compared several models to assess the significance of incorporating the stimulus variable. Initially, we compared two models: one that solely included the Trial variable (both Trial and Trial^²; Model 1) and another that additionally incorporated the Stimulus variable (Model 2). Model 2 demonstrated a significantly improved model fit compared to Model 1, F (2, 2213) = 24.78, p < .001. Subsequently, we introduced a more complex model (Model 3), which additionally included interaction terms between both terms of Trial and the Stimulus variable. Model 3 exhibited a significant improvement over Model 2, F (2, 2209) = 4.47, p = .001, indicating that both the intercepts and slopes of the expectancy ratings throughout extinction differed significantly between stimuli.

Upon closer examination of the coefficients in Model 3 (Table 2), we observed some noteworthy findings. Specifically, the intercept for CS_zoom was significantly higher than that of both CS_yoked and CS_unmoving, suggesting higher rates of US-expectancies overall. Additionally, both interaction terms between Trial and CS_unmoving were significant, albeit with opposite signs. This implies that while the linear slope was steeper for CS_unmoving compared to CS_zoom, the quadratic slope (or “initial drop”) in expectancies was less steep. Conversely, no significant interaction was found between Trial and CS_yoked, indicating that its slope did not differ significantly from that of the CS_zoom.

Table 2.

Coefficients for the prediction of expectancy ratings during extinction based on trial, stimulus, and interactions.

	β	SE	t	p
Intercept	3.11	0.09	35.04	<.001***
Trial	−61.53	4.19	−14.69	<.001***
Trial²	25.88	4.19	6.18	<.001***
CS_yoked	−0.45	0.13	−3.62	<.001***
CS_unmoving	−0.89	0.13	−7.06	<.001***
Trial x CS_yoked	5.19	5.92	0.88	.381
Trial² x CS_yoked	0.48	5.92	0.08	.935
Trial x CS_unmoving	18.74	5.92	3.17	.002**
Trial² x CS_unmoving	−13.46	5.92	−2.27	.023*

*p < .05 **p < .01 ***p < .001.

Correlations between proposed mechanisms and return of fear

Finally, we investigated how the proposed mechanisms (valence, approach tendencies, and expectancies during extinction) related to ROF (spontaneous recovery and reinstatement). Correlation matrices are provided in Supplementary Material 2. We found significant correlations only between expectancies during extinction and verbal measures of return of fear. Specifically, we observed positive correlations for the primary ROF outcomes, indicating that higher expectancy ratings throughout extinction were associated with higher expectancy ratings (for spontaneous recovery, r (70) = .61, p < .001, and reinstatement, r (70) = .59, p < .001), and distress ratings (for spontaneous recovery, r (70) = .30, p = .011, and reinstatement, r (70) = .42, p < .001) during the ROF test phases. Contrastingly, negative correlations were observed for the secondary ROF outcomes, suggesting that higher expectancy ratings throughout extinction were associated with a smaller increase in expectancy from the end of extinction to the ROF test phases (for spontaneous recovery, r (70) = −.28, p = .019, and reinstatement, r (70) = −.25, p = .036). All other correlations, including those relating valence and approach tendencies to ROF, were non-significant (all p-values >.058).

Discussion

In this study, we aimed to examine the effects of instructed approach toward the conditional stimulus (CS) during fear extinction on the extinction process and return of fear. In addition, we tested hypotheses about the working mechanisms through which approach might affect exposure/extinction outcomes, drawing on predictions from the inhibitory learning theory (Bouton, 1993; Craske et al., 2008, 2014, 2022) and the reflective-impulsive model of behavior (Deutsch & Strack, 2020). To this aim, we employed a within-subjects design, in which a stimulus that was repeatedly approached during an extinction procedure was compared to a yoked moving and an unmoving control stimulus. Return of fear was examined by a spontaneous recovery test (1 day after extinction) and a reinstatement test.

We primarily assessed the impact of instructed approach behavior on spontaneous recovery and reinstatement, expecting to find both measures to be reduced for the stimulus that was repeatedly approached. We found a significant return of fear in both phases. However, this return of fear was not reduced for the approached stimulus compared to the others on the first trial of spontaneous recovery and reinstatement test. Exploratory analyses did reveal very limited evidence for a differential increase in fear responding from the end of extinction to the start of the spontaneous recovery and reinstatement test, with a smaller increase for the CS_zoom, but this effect was only significant for distress ratings from the end of extinction to reinstatement test. Moreover, this effect was not supported by the complementary Bayesian analysis (with the Bayesian factor estimating the absence of an effect five times more likely than its presence), and it proved sensitive to stimulus order effects during reinstatement.

Secondarily, we assessed a number of predictions made about the driving mechanisms derived from the reflective-impulsive model of behavior (RIM). Based on this model, it was hypothesized that valence and action tendencies could be altered as a result of approach training during extinction. Our results demonstrated significant changes in valence throughout the procedure. Most importantly, there was a significant shift toward more positive valence after extinction. However, no significant differences in valence were observed between the stimulus that was approached and those that were not. Bayesian analysis further supported these findings, indicating that a stimulus-based difference in valence was highly unlikely. With regard to action tendencies, we observed that participants needed more time to avoid the three stimuli in comparison to approaching them. This suggests a motivational inclination to approach instead of avoid all three stimuli. Again, however, no differences in approach tendencies were observed between the stimuli, as supported by frequentist (p-values) and Bayesian analyses, suggesting that approach during extinction did not affect action tendencies. Furthermore, early accounts relating approach to automatic stimulus evaluation suggested that a specific arm movement may be necessary to constitute an approach action (flexion vs. extension). However, no frequentist or Bayesian evidence was found for differences in valence ratings or action tendencies when comparing flexing versus extending the arm muscles to approach the CS, and similarly, no effect was found on the return of fear. Additionally, it is noteworthy that we did not find significant correlations between changes in valence nor approach tendencies and return of fear measures. These findings suggest that, irrespective of condition, changes in valence and action tendencies were not associated with return of fear.

A plausible explanation for the absence of evidence aligned with the reflective-impulsive model may lie in its expectation that approach behavior automatically elicits positive valence. Research conducted by Mertens and colleagues (2018) suggests that the evaluation of approach actions is highly context-dependent, and thus not unequivocally positive. Specifically, the authors argue that when approaching a threatening stimulus, the approach action may not be interpreted as a positive action, even when it is clearly labeled as such. Consequently, there may be no positive valence associated with the approach behavior to then transfer to the stimulus. This highlights an important nuance: when participants perceive the approach action as threatening, engaging in approach behavior during exposure or extinction may not have the expected beneficial effects on valence and action tendencies and could even have unfavorable consequences.

Subsequently, to assess predictions based on the inhibitory learning theory (ILT), we conducted exploratory analyses to examine the potential beneficial impact of approach behavior on US-expectancies during extinction. We hypothesized that the approached stimulus would elicit higher US-expectancy ratings compared to the other stimuli, which could be indicative of a greater violation of fear-related expectations as the unconditional stimulus (US) was not delivered during extinction. Consistent with this hypothesis, we observed significantly higher US-expectancies throughout the extinction phase, for the approached stimulus compared to both the yoked stimulus and the unmoving control stimulus. It is important to note that the former difference may be impacted by the design of the extinction phase. In order to align the movement of the yoked stimulus with the participant’s earlier responses, in blocks with an approach prompt, this stimulus was always preceded by the approached stimulus. Therefore, it is possible that the US-expectancies for these yoked stimuli were reduced compared to the accompanying approached stimulus, simply due to the fact that they were consistently presented later in the extinction phase. However, this does not detract from the differences found between the approached and the unmoving control stimulus, which was presented in a random order within the blocks. Additionally, the initial exponential slope (or “drop”) in expectancies was found to be larger for the approached stimulus compared to the unmoving control stimulus, suggesting that there may have been a stronger expectancy violation, specifically at the start of the phase. Of note, is that an additional exploratory analysis revealed mixed findings on the relationship between expectancies during extinction and return of fear. While higher expectancies during extinction correlated with increased primary ROF indices (expectancies during ROF test phases), they were associated with decreased secondary ROF indices (expectancy change from the end of extinction to ROF test phases). These results need to be interpreted with caution: positive correlations with primary indices may suggest a potential response bias for some participants (i.e., higher expectancy ratings throughout the procedure), an issue that the secondary ROF indices mitigate.

In summary, our findings demonstrate some evidence of heightened expectancies for the approached stimulus during extinction, suggesting a potential for enhanced expectancy violation and increased inhibitory learning. Moreover, some preliminary evidence was found that heightened expectancies may in turn relate to reductions in return of fear. However, we did not consistently observe any effect of approach on return of fear. Furthermore, contrary to the predictions of the reflective-impulsive model, we did not find evidence of enhanced valence or approach tendencies toward the approached stimulus. These findings provide initial insights into the potential role of approach behavior in exposure therapy. However, the proposed underlying mechanisms driving these effects partially remain unclear and further investigation is necessary before extending these conclusions to clinical recommendations. With that in mind, we propose some avenues for future research.

A first potential avenue for future research lies in the inclusion of goal-directed approach behavior during extinction. In our procedure, approach was simply instructed. No reasoning or incentive was tied to performing the approach behavior. This decision aligned with the proposed working mechanisms outlined by the Reflective-Impulsive Model (RIM), suggesting that valence and action tendencies can be changed through implicit/automatic processes, regardless of specific goals. However, this decision had its limitations. As mentioned in the introduction, according to the RIM, the implicit system is only one of two key elements in shaping overt behavior, the other being cognitive decision-making. The instructed nature of our approach manipulation may have neglected the role of goal-directed actions in shaping behavior. Additionally, the lack of reasoning or incentives for approach behavior may have overlooked the potential benefits associated with positive affect and feelings of mastery linked to goal attainment, which may facilitate both the reduction of avoidance behaviors and fear. Relevant to these points, a study by Pittig and Wong (2021) found that while both incentives and instructions for non-avoidance had a similar impact on fear and avoidance, incentives had a stronger effect on reducing self-reported avoidance motivation when participants imagined continuing the paradigm. This might suggest that incentives, and, by extension, goal-directed approach may offer benefits to extinction that the current procedure missed.

Second, a potential limitation of our study stems from exclusively implementing an approach manipulation during the extinction phase. Because of this decision, for the stimulus that was approached, certain aspects of the extinction phase differed from the preceding and subsequent phases, potentially leading to a partial protection from extinction effect. While we took precautions to mitigate this risk, such as delaying the approach component and interspersing trials without approach, a residual bias may persist, particularly increasing the return of fear for the approached stimulus. Hence, we recommend that future studies consider allowing approach throughout the entire experiment procedure, only stimulating it for one stimulus during extinction, to address these concerns effectively.

Third, to further enhance the applicability of our findings to clinical practice, we recommend considering the inclusion of an avoidance phase during the acquisition stage in future research. This adjustment aligns with the typical experiences of individuals with anxiety disorders, who often engage in extensive avoidance behaviors before undergoing exposure therapy (Craske et al., 2022; Pittig et al., 2018). For such individuals, introducing an alternative action through approach behavior training may hold particular promise. Consequently, we advise to further investigate the impact of approach behavior within a design that incorporates additional avoidance training. Alternatively, future research could directly focus on clinical populations, characterized by a history of rigid avoidance behavior.

Fourth, in future research, it could be beneficial to incorporate overt behavior as a primary outcome measure. As predicted by the reflective-impulsive model, modifying the impulsive system, including valence and action tendencies, may not only influence fear but also overt behavior. Therefore, including an overt avoidance measure would provide valuable additional insights into the effects of approach training. Moreover, this addition would enhance the clinical validity future studies, by addressing an important objective of exposure therapy, which is to alleviate the burden of avoidance behavior (Craske et al., 2022).

In conclusion, our study did not provide evidence supporting the predicted effects of instructing approach on return of fear. Furthermore, contrary to the predictions of the reflective-impulsive model (RIM), we did not observe enhanced valence or approach tendencies toward the approached stimulus. Notably, we did observe that enhanced expectancy ratings when approaching the fear-eliciting stimulus during extinction, which may indicate a potential for increased inhibitory learning. That being said, these findings should be interpreted with caution due to a number of limitations. Addressing these limitations in future research, such as enhancing ecological validity, including avoidance behavior measures, considering incentives, and exploring the context-dependent nature of approach actions, will further contribute to our understanding of the effects of approach training in exposure therapy and its potential application in clinical populations.

Footnotes

Acknowledgments

We would like to thank Mathijs Franssen for his help in processing the physiological data, and Noor Demeulenaere for her role in data collection.

Author contributions

Contributed to conception and design: NC, SS, and DH. Contributed to acquisition of data: NC. Contributed to analysis and interpretation of data: NC. Drafted and revised the article: NC, SS, and DH.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by a C1 research grant of the KU Leuven, awarded to Dirk Hermans, Tom Beckers, Bram Vervliet, and Laura Luyten (3H190245). Additionally, a large part of the equipment was supported by an infrastructure grant from the FWO and the Research Fund KU Leuven, Belgium (AKUL/19/06; I011320N).

ORCID iD

Naomi Carpentier

Notes

Author Biographies

Naomi Carpentier is a PhD candidate at KU Leuven.

Sara Scheveneels an assistant professor at Maastricht University and a postdoctoral researcher at KU Leuven.

Dirk Hermans is a full professor at KU Leuven.

References

Amir

Kuckertz

J. M.

Najmi

(2013). The effect of modifying automatic action tendencies on overt avoidance behaviors. Emotion, 13(3), 478–484. https://doi.org/10.1037/a0030443

Blumenthal

T. D.

Cuthbert

B. N.

Filion

D. L.

Hackley

Lipp

O. V.

Van Boxtel

(2005). Committee report: Guidelines for human startle eyeblink electromyographic studies. Psychophysiology, 42(1), 1–15. https://doi.org/10.1111/j.1469-8986.2005.00271.x

Boucsein

Fowles

D. C.

Grimnes

Ben-Shakhar

Roth

W. T.

Dawson

M. E.

Filion

D. L.

(2012). Publication recommendations for electrodermal measurements. Psychophysiology, 49(8), 1017–1034. https://doi.org/10.1111/j.1469-8986.2012.01384.x

Bouton

M. E.

(1993). Context, time, and memory retrieval in the interference paradigms of Pavlovian learning. Psychological Bulletin, 114(1), 80–99. https://doi.org/10.1037/0033-2909.114.1.80

Cacioppo

J. T.

Priester

J. R.

Berntson

G. G.

(1993). Rudimentary determinants of attitudes. Ii: Arm flexion and extension have differential effects on attitudes. Journal of Personality and Social Psychology, 65(1), 5–17. https://doi.org/10.1037/0022-3514.65.1.5

Carpenter

J. K.

Pinaire

Hofmann

S. G.

(2019). From extinction learning to anxiety treatment: Mind the gap. Brain Sciences, 9(7), 164. https://doi.org/10.3390/brainsci9070164

Craske

M. G.

Kircanski

Zelikowsky

Mystkowski

Chowdhury

Baker

(2008). Optimizing inhibitory learning during exposure therapy. Behaviour Research and Therapy, 46(1), 5–27. https://doi.org/10.1016/j.brat.2007.10.003

Craske

M. G.

Treanor

Conway

C. C.

Zbozinek

Vervliet

(2014). Maximizing exposure therapy: An inhibitory learning approach. Behaviour Research and Therapy, 58(1), 10–23. https://doi.org/10.1016/j.brat.2014.04.006

Craske

M. G.

Treanor

Zbozinek

T. D.

Vervliet

(2022). Optimizing exposure therapy with an inhibitory retrieval approach and the OptEx Nexus. Behaviour Research and Therapy, 152(11), Article 104069. https://doi.org/10.1016/j.brat.2022.104069

10.

Dawson

M. E.

Schell

A. M.

Filion

D. L.

(2007). The electrodermal system. Handbook of Psychophysiology, 2, 200–223. https://doi.org/10.1017/9781107415782.010

11.

Deutsch

Strack

(2020). Changing behavior using the reflective-impulsive model. In Hagger

M. S.

Cameron

L. D.

Hamilton

Hankonen

Lintunen

(Eds.), The handbook of behavior change (1st ed., pp. 164–177). Cambridge University Press. https://doi.org/10.1017/9781108677318.012

12.

Dour

H. J.

Brown

L. A.

Craske

M. G.

(2016). Positive valence reduces susceptibility to return of fear and enhances approach behavior. Journal of Behavior Therapy and Experimental Psychiatry, 50, 277–282. https://doi.org/10.1016/j.jbtep.2015.09.010

13.

Haaker

Golkar

Hermans

Lonsdorf

T. B.

(2014). A review on human reinstatement studies: An overview and methodological challenges. Learning & Memory, 21(9), 424–440. https://doi.org/10.1101/lm.036053.114

14.

Hayes

S. C.

Strosahl

K. D.

Wilson

K. G.

(1999). Acceptance and commitment therapy (Vol. 6). Guilford Press.

15.

Jones

C. R.

Vilensky

M. R.

Vasey

M. W.

Fazio

R. H.

(2013). Approach behavior can mitigate predominately univalent negative attitudes: Evidence regarding insects and spiders. Emotion, 13(5), 989–996. https://doi.org/10.1037/a0033164

16.

Kaczkurkin

A. N.

Foa

E. B.

(2015). Cognitive-behavioral therapy for anxiety disorders: An update on the empirical evidence. Dialogues in Clinical Neuroscience, 17(3), 337–346. https://doi.org/10.31887/DCNS.2015.17.3/akaczkurkin

17.

Krypotos

A.-M.

Arnaudova

Effting

Kindt

Beckers

(2015). Effects of approach-avoidance training on the extinction and return of fear responses. PLOS ONE, 10(7), Article e0131581. https://doi.org/10.1371/journal.pone.0131581

18.

Lang

P. J.

Bradley

M. M.

Cuthbert

B. N.

(2008). International affective picture system (IAPS): Affective ratings of pictures and instruction manual. University of Florida. Technical Report A-8/.

19.

Lovibond

P. F.

Mitchell

C. J.

Minard

Brady

Menzies

R. G.

(2009). Safety behaviours preserve threat beliefs: Protection from extinction of human fear conditioning by an avoidance response. Behaviour Research and Therapy, 47(8), 716–720. https://doi.org/10.1016/j.brat.2009.04.013

20.

Mertens

Van Dessel

De Houwer

(2018). The contextual malleability of approach-avoidance training effects: Approaching or avoiding fear conditioned stimuli modulates effects of approach-avoidance training. Cognition and Emotion, 32(2), 341–349. https://doi.org/10.1080/02699931.2017.1308315

21.

Olatunji

B. O.

Cisler

J. M.

Deacon

B. J.

(2010). Efficacy of cognitive behavioral therapy for anxiety disorders: A review of meta-analytic findings. Psychiatric Clinics of North America, 33(3), 557–577. https://doi.org/10.1016/j.psc.2010.04.002

22.

Pittig

(2019). Incentive-based extinction of safety behaviors: Positive outcomes competing with aversive outcomes trigger fear-opposite action to prevent protection from fear extinction. Behaviour Research and Therapy, 121, Article 103463. https://doi.org/10.1016/j.brat.2019.103463

23.

Pittig

Treanor

LeBeau

R. T.

Craske

M. G.

(2018). The role of associative fear and avoidance learning in anxiety disorders: Gaps and directions for future research. Neuroscience & Biobehavioral Reviews, 88, 117–140. https://doi.org/10.1016/j.neubiorev.2018.03.015

24.

Pittig

Wong

A. H. K.

(2021). Incentive-based, instructed, and social observational extinction of avoidance: Fear-opposite actions and their influence on fear extinction. Behaviour Research and Therapy, 137, Article 103797. https://doi.org/10.1016/j.brat.2020.103797

25.

Rescorla

R. A.

(1969). Pavlovian conditioned inhibition. Psychological Bulletin, 72(2), 77–94. https://doi.org/10.1037/h0027760

26.

Springer

K. S.

Levy

H. C.

Tolin

D. F.

(2018). Remission in CBT for adult anxiety disorders: A meta-analysis. Clinical Psychology Review, 61, 1–8. https://doi.org/10.1016/j.cpr.2018.03.002

27.

Tolin

D. F.

(2019). Inhibitory learning for anxiety-related disorders. Cognitive and Behavioral Practice, 26(1), 225–236. https://doi.org/10.1016/j.cbpra.2018.07.008

28.

Van Dessel

De Houwer

Gast

(2016). Approach–Avoidance training effects are moderated by awareness of stimulus–Action contingencies. Personality and Social Psychology Bulletin, 42(1), 81–93. https://doi.org/10.1177/0146167215615335

29.

Van Dis

E. A. M.

Hagenaars

M. A.

Bockting

C. L. H.

Engelhard

I. M.

(2019). Reducing negative stimulus valence does not attenuate the return of fear: Two counterconditioning experiments. Behaviour Research and Therapy, 120(3), Article 103416. https://doi.org/10.1016/j.brat.2019.103416

30.

Van Uijen

S. L.

Van Den Hout

Engelhard

I. M.

(2015). Active approach does not add to the effects of in vivo exposure. Journal of Experimental Psychopathology, 6(1), 112–125. https://doi.org/10.5127/jep.042014

31.

Vervliet

Craske

M. G.

Hermans

(2013). Fear extinction and relapse: State of the art. Annual Review of Clinical Psychology, 9(1), 215–248. https://doi.org/10.1146/annurev-clinpsy-050212-185542

32.

Weisman

J. S.

Rodebaugh

T. L.

(2018). Exposure therapy augmentation: A review and extension of techniques informed by an inhibitory learning approach. Clinical Psychology Review, 59, 41–51. https://doi.org/10.1016/j.cpr.2017.10.010

33.

Wolitzky

K. B.

Telch

M. J.

(2009). Augmenting in vivo exposure with fear antagonistic actions: A preliminary test. Behavior Therapy, 40(1), 57–71. https://doi.org/10.1016/j.beth.2007.12.006

34.

Zbozinek

T. D.

Hermans

Prenoveau

J. M.

Liao

Craske

M. G.

(2015). Post-extinction conditional stimulus valence predicts reinstatement fear: Relevance for long-term outcomes of exposure therapy. Cognition and Emotion, 29(4), 654–667. https://doi.org/10.1080/02699931.2014.930421

Approach and Conquer: Optimizing Fear Extinction by Adding Approach?

Abstract

Keywords

Introduction

Method

Participants

Materials and measures

Stimuli

US-expectancy and distress ratings

CS valence

Skin conductance (SC)

Procedure

Day 1: Preparation

Day 1: Fear acquisition training

Day 2 and 3: Preparatory steps

Day 2: Extinction

Day 3: Spontaneous recovery and reinstatement test

Day 3: Manikin approach-avoidance test

Data reduction

Statistical analyses

Results

Spontaneous recovery

Reinstatement

CS valence

Action tendencies

Specific approach movement

US-expectancies throughout extinction

Correlations between proposed mechanisms and return of fear

Discussion

Footnotes

Acknowledgments

Author contributions

Declaration of conflicting interests

Funding

ORCID iD

Notes

Author Biographies

References