Individual Differences in Vulnerability Factors for Anxiety Disorders: A Study on the Acquisition, Extinction, and Renewal of Avoidance, and the Concomitant Dynamics of Relief

The last decade has seen a surge in experimental research on fear extinction and the return of fear in humans due to its potential to improve our understanding of the effects of exposure therapies on fear reduction and fear relapse after treatment for anxiety-related disorders (Craske et al., 2018; Delamater & Westbrook, 2014; Vervliet et al., 2013). The procedure used in many studies involves participants learning a fear reaction to a conditioned stimulus (CS) through a Pavlovian learning phase based on CS-unconditional stimulus (US) pairings. Then, participants undergo a fear extinction phase in which the CS is presented, and the US is withheld, which usually leads to fear reduction. Finally, the return of fear is usually induced in a new extinction test phase following different procedures such as inserting a time interval between the extinction and the test phase (usually leading to a spontaneous fear-recovery effect; Pavlov, 1927; Quirk, 2002) or conducting the test phase in a context different from that in which the extinction phase took place (known to produce a renewal-of-fear effect; Bouton & Bolles, 1979; Laborda et al., 2011; see also the reviews by Craske et al., 2018; and Vervliet et al., 2013).

Despite the extensive literature on fear extinction and recovery, very few studies have focused on avoidance as a behavioural measure of fear, and even fewer have been concerned with studying individual differences in the extinction and return of avoidance (see, however, Krypotos & Engelhard, 2018; Papalini et al., 2021; Pittig et al., 2018; San Martín et al., 2020; Sheynin et al., 2014; Vervliet et al., 2017; Vervliet & Indekeu, 2015). Regarding anxiety-related disorders, avoidance is considered one of the most important maladaptive reactions due to its dysfunctional consequences for patients, and its causal role in the development and maintenance of anxiety (Cameron et al., 2015; Engelhard et al., 2015; Krypotos et al., 2015; López-Moraga et al., 2022; Lovibond, 2006; Pittig et al., 2018; van Dis et al., 2022; van Uijen, Dalmaijer, et al., 2018; van Uijen, Leer, et al., 2018; Vervliet et al., 2017; Vervliet & Indekeu, 2015; Xia et al., 2019). Moreover, evidence suggests that avoidance may persist even after successful fear extinction (Bravo-Rivera et al., 2015; Hendrikx et al., 2021; Urcelay et al., 2019; Vervliet & Indekeu, 2015; Zuj et al., 2020), and that the mere availability of avoidance may promote the relapse of fear after extinction (van Uijen, Leer, et al., 2018). These considerations have led to a growing interest in studying the effects of extinction procedures on avoidance behaviour, as well as revealing the underlying mechanisms responsible for the persistence and return of avoidance (Dymond, 2019; Krypotos et al., 2015; Pittig et al., 2018).

Recent studies on the extinction and return of avoidance have benefited from the analysis of individual differences associated with risk factors for anxiety-related disorders (see, for example, Arnaudova et al., 2013; Flores et al., 2018; Krypotos & Engelhard, 2018; Papalini et al., 2021; Pittig et al., 2018; San Martín et al., 2020; Sheynin et al., 2014; Vervliet et al., 2017; Vervliet & Indekeu, 2015; Wake et al., 2021). In general, the study of the acquisition, extinction, and return of fear from an individual-difference perspective holds considerable promise for helping unveil the mechanisms responsible for the persistence and return of fear and improving the efficacy of fear reduction treatments in the context of pathological anxiety (Lonsdorf & Merz, 2017). In the present study, we assessed the impact of trait anxiety (TA) and intolerance of uncertainty (IU) on the extinction and recovery of avoidance induced by a change in the training context (i.e., a renewal procedure).

Trait anxiety is “considered a measure of general vulnerability to emotional disorder and distress” (Nordahl et al., 2019), and has been found to be related to enhanced generalisation of conditioned fear to safety (i.e., never followed by the US) and ambiguous signals in experiments based on Pavlovian paradigms (Gazendam et al., 2013; Wong & Lovibond, 2018; see also Sep et al., 2019, for a systematic review and meta-analysis), generalisation of avoidance to safety signals (Vervliet & Indekeu, 2015; see also Vervliet et al., 2017, for related results), and anxiety-related mental disorders (Knowles & Olatunji, 2020). Consequently, it is reasonable to expect TA to have an impact on persistent avoidance, and hence, to be related to impaired avoidance extinction and enhanced renewal. Intolerance of uncertainty (IU) has been defined as “an individual’s dispositional incapacity to endure the aversive response triggered by the perceived absence of salient, key, or sufficient information, and sustained by the associated perception of uncertainty” (Carleton, 2016, p. 31). This personality trait is considered a vulnerability factor for several anxiety-related mental disorders (Boswell et al., 2013; Carleton, 2012; Carleton et al., 2012; Gentes & Ruscio, 2011; Hong & Cheung, 2015; Mahoney & McEvoy, 2012; McEvoy & Erceg-Hurn, 2016; McEvoy & Mahoney, 2013; Norr et al., 2013; see also Carleton, 2016, for a review), and has shown to be associated with disrupted threat extinction (Morriss et al., 2015; Morriss, Christakou, et al., 2016; Morriss et al., 2019; see also the review by Morriss, Zuj, et al., 2021, and the meta-analysis by Morriss, Wake, et al., 2021) and generalisation of avoidance responses (San Martín et al., 2020). Freeston et al.’s (1994) IU scale — one of the most widely used IU questionnaires — has been found to have a factorial structure of two components (Birrell et al., 2011), that is, a prospective intolerance of uncertainty component (P-IU), defined as the “desire for predictability and an active engagement in seeking certainty”, and an inhibitory intolerance of uncertainty component (I-IU), defined as the “paralysis of cognition and action in the face of uncertainty”. Given that previous experiments conducted in our laboratory (see Cobos et al., 2022, and Flores et al., 2018, 2020) have found P-IU to be related to excessive and persistent avoidance (even when controlling for TA and I-IU), we decided to assess the impact of this specific IU dimension on avoidance extinction and renewal. Consequently, we expected P-IU to be related to impaired avoidance extinction and enhanced avoidance renewal.

Most of the (very few) studies related to avoidance extinction and recovery from an individual-difference perspective have used extinction with response prevention (ExtRP) to reduce avoidance frequency (Krypotos & Engelhard, 2018; Papalini et al., 2021; Vervliet et al., 2017; Vervliet & Indekeu, 2015). This procedure involves the presentation of the CS while the aversive US is withheld and the avoidance response is not available. One commonly argued rationale for choosing the ExtRP procedure is its similarity to exposure-like treatments in clinical practice, where patients are usually exposed to the feared situation after being deterred from engaging in avoidance behaviour. However, ExtRP also has several shortcomings, one of which is that it does not allow for assessing the effect of extinction on avoidance during the very course of the extinction procedure. Instead, the effect of extinction on avoidance is measured through an avoidance test in which the avoidance response is again made available to participants while the aversive US is still withheld. This switch from the extinction to the avoidance test phase is typically signalled by means of instructions or a specific stimulus indicating that the avoidance response is available, which has been interpreted by some experts as a context change that may, per se, lead to a renewal effect (Papalini et al., 2021). In fact, several studies have found very modest reductions in avoidance frequency after ExtRP as well as fear recovery on the early trials of re-extinction through ExtRP after the avoidance test phase (Hendrikx et al., 2021; Urcelay et al., 2019; Vervliet & Indekeu, 2015). In our experiment, we used an alternative extinction procedure in which making the avoidance response did not prevent participants from noticing that the US no longer followed the mere occurrence of the CSs+ (see also Sheynin et al., 2014, for another alternative procedure to extinguish passive avoidance). Specifically, in our experiment, avoidance responses led to a reduction in US intensity, rather than to its absolute prevention. Thus, the complete omission of the US during the extinction phase could not be attributed to the avoidance response but to a change in the CS-US predictive relationship. An interesting advantage of our procedure is that it allows us to establish an appropriate avoidance baseline in the later trials of the extinction phase to accurately measure the magnitude of avoidance recovery after extinction resulting purely from a context change.

Another aspect in which our experiment differs from most previous studies is the use of a learning task in which participants must make multiple responses to ensure the avoidance of the US on a single trial. In most studies, avoidance of the US on a given trial depends on whether or not a single avoidance response is made (Krypotos & Engelhard, 2018; Lovibond et al., 2008; Papalini et al., 2021; Vervliet et al., 2017). Although this strategy has proved to be fruitful for studying avoidance behaviour from an individual-difference perspective (Papalini et al., 2021; San Martín et al., 2020), an alternative strategy based on a free operant procedure may be more sensitive to individual differences in avoidance frequency (see Krypotos et al., 2018, for similar arguments). For this purpose, as in Flores et al. (2018) and Cobos et al. (2022), our learning task included trials in which the aversive US could occur according to a variable time schedule within the CS duration interval. To avoid the US, the avoidance response had to precede the occurrence of the US by no more than one second. This procedure forces participants to respond many times per trial and leads to a wide between-subject variation in avoidance frequency. Additionally, the occurrence of the US was affected by several sources of uncertainty: a) The US occurred on 50% of CS+ trials, b) the US occurred at different and unpredictable time points on each trial, and c) on many trials, participants remained uncertain about whether they had succeeded in avoiding the US until the CS offset. As in previous experiments in our laboratory, we expected these sources of uncertainty to have a differential impact on participants’ avoidance reactions depending on their scores on anxiety-predisposing factors such as TA and P-IU.

Finally, we measured participants’ post-trial relief ratings on each trial on which the US was withheld throughout the whole learning task as a measure of their degree of distress during the preceding trial. Relief has been defined as a pleasant emotion resulting from a sudden reduction in distress (Hoerl, 2015). Accordingly, the degree of relief experienced by participants after trials on which the US had been omitted is assumed to generally reflect their level of distress during the previous trial. Following this logic, high relief ratings might be taken to indicate high distress during the preceding trial, whereas low relief ratings would indicate a small degree of distress. Previous studies have shown that relief ratings may be very useful for detecting individual differences associated with anxiety-predisposing factors (Cobos et al., 2022; Papalini et al., 2021; San Martín et al., 2020; Vervliet et al., 2017) and may shed some light on the mechanisms responsible for excessive and persistent avoidance. In general, we expected P-IU and TA to be related to poorer CS+ vs CS- discrimination in terms of avoidance frequency and relief ratings, impaired extinction of avoidance and relief ratings, and enhanced renewal of avoidance and relief ratings.

The participants were 195 undergraduate students from the Faculty of Psychology and Speech Therapy at the University of Málaga (Spain), who volunteered to take part in the study in exchange for course credits. The sample size included the maximum possible number of participants that could be recruited from the convenience samples of students available at the time of the experiment. The data from 43 participants were excluded from the analysis for the following reasons: 33 participants were either not able to effectively avoid the aversive sound on any trial (30 participants) or provided the same relief ratings across all CS types and trial blocks during any of the different phases of the experiment (4 participants; one of which was also excluded due to the former criterion). In addition, the data from 10 participants were excluded as they did not complete any of the questionnaires used in the study. Therefore, the final sample consisted of 152 participants (120 females; M_age = 18.97, SD_age = 2.05; age range = 17 - 30).

The entire procedure was performed using IBM-compatible PCs, and the experimental task was programmed in Psychopy 2020.2.6 (Peirce et al., 2019). The task files are available in the Open Science Framework (https://osf.io/ew52z/). Visual stimuli were presented on a 21-inch monitor with a 1920x1080 pixels resolution, while sounds were delivered via headphones (Manufacturer: Audio-Technica model ATH-M20x). Participants’ responses were registered through a standard QWERTY keyboard and the PC mouse. After completion of all questionnaires, the various phases of the task began.

The participants completed three personality questionnaires to evaluate IU, trait anxiety, and distress tolerance. The instruments used were the Spanish adaptations of the Intolerance of Uncertainty Scale: IUS (Freeston et al., 1994; adaptation: González-Rodríguez et al., 2006), the State-Trait Anxiety Inventory: STAI (Spielberger et al., 1983; adaptation: Seisdedos, 1990), and the Distress Tolerance Scale: DTS (Simons & Gaher, 2005; adaptation: Sandín et al., 2017). All the participants gave written informed consent before completing the questionnaires.

The IUS (internal consistency ranged from .91 to .94 and test-retest reliability ranged from .74 to .83) is a 27-item self-report measure that assess the degree to which individuals find uncertainty to be distressing and undesirable (i.e., “I must get away from all uncertain situations” and “Uncertainty makes me uneasy, anxious, or stressed”). This scale includes two subscales to assess the following two factors: Prospective Intolerance of Uncertainty (11 items) and Inhibitory Intolerance of Uncertainty (16 items). Items are rated on a five-point Likert scale ranging from 1 (not at all characteristic of me) to 5 (extremely characteristic of me).

The STAI (internal consistency ranging from .86 to .95 and test-retest reliability coefficients ranging from .65 to .75) is a 40-item self-report measure for assessing TA (i.e., I worry too much over something that really doesn’t matter) and State Anxiety (i.e., I feel calm and secure). Given our focus on TA, we only administered the 20 items corresponding to the trait anxiety subscale. Items are rated on a four-point Likert scale ranging from 0 (hardly ever) to 3 (always).

The DTS (internal consistency .82 and test-retest reliability .7) is a 15-item self-report measure for assessing how well people tolerate feelings of distress (i.e., Being distressed or upset is always a major ordeal for me). Items are rated on a five-point scale ranging from 1 (strongly agree) to 5 (strongly disagree).

The CSs consisted of three geometric figures: a blue square with a base of 212 pixels (RGB values of 0, 113, 192), a grey circle with a diameter of 212 pixels (RGB values of 191, 191, 191) and a yellow triangle with a base and height of 212 pixels (RGB values of 191, 144, 0). The shapes were presented at the center of a computer screen against a black background square with a base of 268 pixels. CSs were presented for 6 s. The black background square included a 4-pixel red frame (RGB values of 253, 56, 55). The aversive US was a 2-second high-pitch beep of 44100 Hz and high-volume (97 ± 3 dB) presented simultaneously to the left and the right ear. Pictures of a kitchen, an office room, and a bedroom served as contexts X, Y, and Z (counterbalanced). Context images occupied the full size of the computer screen as a background when the CSs were presented. Figure 1 depicts the different phases of the experimental design: A Pavlovian phase, an Avoidance phase, an Extinction phase, and finally, a Re-extinction phase.

The dependent measures used in our experiment were the participants’ relief ratings given on every trial on which the US was omitted and the frequency of avoidance responses on each trial. Additionally, participants also gave unpleasantness ratings for the US before starting the experimental task and at the end of the experiment.

The procedure used in this experiment complied with the Helsinki Declaration and was approved by the local ethical committee of the University of Málaga (registry code 46-2020-H). The participants entered the experimental room in groups of 10 and sat at a minimum distance of 2 m apart from each other. Additional measures to protect the participants from COVID-19 infection included hand washing with hydroalcoholic gel and keeping the windows and door open to ensure adequate ventilation. The participants and experimenters wore face masks at all times during the experiment. The participants started by carefully reading an informed consent document on the computer screen after which they were invited to accept (or decline) the invitation to participate. Then, they completed the IUS, followed by the STAI (only the TA subscale), and the DTS. After completing the questionnaires, they began the experimental task. Participants within the same session could only leave the experimental room once everyone had completed the task.

Participants began by putting on the headphones and reading the instructions concerning the Pavlovian learning phase. The instructions told the participants that they would be presented with a series of geometrical figures, and that some of them would occasionally be paired with an upcoming unpleasant sound presented to both ears. They experienced the aversive sound during the instructions and were requested to rate its unpleasantness on a 9-point horizontal scale with the anchors “0 = Not unpleasant at all” and “9 = Extremely unpleasant”.

Each trial began with a fixation cross displayed at the center of the screen on an image of Context X, which served as a permanent background. After 1 s, the fixation cross was immediately replaced by the CS for 6 s. Two different geometric figures (CS+_Av and CS+_Unav) were paired with the aversive sound (US) 50% of the time, whereas the other geometric figure (CS-) was never paired with the US. A total of 8 trials per CS type was programmed. The onset of the sound could occur between 2 and 5 seconds after the onset of the CS and was programmed according to a variable interval schedule that followed a rectangle distribution with an amplitude of 3 seconds. That is, the aversive sound could randomly appear at second 3, 4, or 5 from the onset of the CS.

Subjective relief ratings were measured immediately after the offset of the CS on each trial in which the US was omitted. Participants used an analogue and numerical, horizontally displayed scale to give their ratings. The onset of the scale occurred immediately after the offset of the CS. The question “How much relief did you feel at the offset of the image?” prompted their ratings. The rating scale ranged from 0 to 100 (“0 = No relief at all”, “50 = Moderate relief”, and “100 = A lot of relief”). A message below the scale indicated to the participants that they had to place a triangular marker with the mouse on the position of the scale that best reflected their degree of relief. Changes in the marker position were immediately translated into the corresponding numerical expression (from 0 to 100), which was displayed just below the scale. By clicking on the numerical rating, the participants confirmed their acceptance, and the scale was removed from the screen. The fixation cross of the following trial appeared two seconds after the scale removal or after the CS offset on those trials in which the US was delivered. Overall, this phase comprised four blocks of six trials: 2 x CS+_Av trials (only one reinforced), 2 x CS+_Unav trials (only one reinforced), and 2 x CS- non-reinforced trials.

Participants were instructed that during this phase they could attenuate the volume of the unpleasant sound by pressing the space bar. The instructions stressed that the inhibitory effect of every single bar press only lasted one second. Thus, to ensure the avoidance of the aversive sound, they were advised to press again before one second after the last response. Participants were explicitly told that pressing at a rate of one response per second or more will ensure the attenuation of the sound volume. However, they were not told that the avoidance response was ineffective in the presence of one of the CSs+. The instructions encouraged the participants to learn when the avoidance response was effective and when it was not. No immediate feedback was provided regarding the consequences of individual responses, and therefore, participants were not certain whether they had avoided the aversive sound until the end of the trial. The temporal sequence of events was the same as in the Pavlovian phase, and relief ratings were requested only on those trials in which no US (in its aversive or mild form) was delivered. This phase comprised six blocks of six trials: 2 CS+_Av trials (only one reinforced or attenuated if successfully avoided), 2 CS+_Unav trials (only one reinforced regardless of whether or not participants made the avoidance response), and 2 CS- (non-reinforced trials).

Following the avoidance phase, the extinction phase took place. During this phase, a different context (Context Y) appeared in the form of a permanent screen background until the end of the phase, and CSs were presented without being followed by the US. Participants were not given any further instructions about the effect of the space bar and were unaware that they would not receive the aversive reinforcement during this phase. To ensure that participants noticed the context change, the previous context remained on the screen for two seconds after the end of the last trial of the previous phase and was replaced by Context Y after an inter-stimulus interval of one second during which a black background was displayed on the screen. Additionally, the first trial of the extinction phase started 6 seconds after the onset of Context Y. Thus, the avoidance response was available throughout this phase. A total of eight blocks of six trials: 2 CS+_Av non-reinforced trials, 2 CS+_Unav non-reinforced trials, and 2 CS- non-reinforced trials were programmed. Relief ratings were requested at the end of each trial of this phase, and the temporal sequence of events on each trial was the same as the previous phases.

This phase was identical to the previous phase except for the display of a different context (Context Z) as a screen background until the end of the learning task. Again, no instructions were given about the change of phase and the absence of the US, and phase transition proceeded as explained in the previous section. This phase comprised four blocks of six trials: 2 CS+_Av non-reinforced trials, 2 CS+_Unav non-reinforced trials, and 2 CS- non-reinforced trials. Relief ratings were requested at the end of each trial of this phase.

Once participants had completed the different phases, they were requested to rate the level of unpleasantness of the aversive sound on the same 9-point horizontal scale as described above.

The raw data file has also been uploaded to the Open Science Framework repository and can be found by clicking on the link https://osf.io/ew52z/. Statistical analyses were conducted using IBM SPSS (version 23; IBM Corp, 2015).

In all the repeated-measures ANOVAs, sphericity was evaluated using Mauchly’s test, and the degrees of freedom were corrected using the Greenhouse–Geisser correction, if necessary. The effect size statistics reported are partial eta squared (⁠$ηp2$⁠) and Cohen’s d for t tests.

Bivariate correlational analyses were conducted to assess the relationship between P-IU, I-IU, TA, and DT scores.

Repeated measures ANOVAs were conducted on relief ratings and avoidance frequency for each phase of the task. As our study is mainly focused on the acquisition, extinction, and renewal of avoidance, we refer the reader to the Supplementary Materials for the analyses of data corresponding to the Pavlovian phase. In the avoidance phase, a 3 (CS Type: CS+_Av, CS+_Unav, and CS-) x 6 (Trial Block: 1–6) was performed. In the extinction phase, a 3 (CS Type: CS+_Av, CS+_Unav, and CS-) x 8 (Trial Block: 1–8) analysis was conducted; and finally, in the re-extinction phase, a 3 (CS Type: CS+_Av, CS+_Unav, and CS-) x 8 (Trial: 1–8) analysis was carried out. Additionally, we conducted a 3 (CS Type: CS+_Av, CS+_Unav, and CS-) x 2 (Phase: Last block of extinction phase, first re-extinction trial) RM-ANOVA on relief ratings and avoidance frequency to assess the renewal effect.

To assess the role of individual differences, we included each personality trait as a covariate in separate ANCOVAs. Specifically, we focused on the role of TA and P-IU given that these were the only traits that yielded significant effects (see Supplementary Materials for the results regarding DT and I-IU, as well as intolerance of uncertainty measured through the short version of the IU scale). Thus, we conducted separate ANCOVAs on relief ratings and the frequency of avoidance responses for each of the different phases of the experiment. When one or more personality traits showed a significant interaction with a factor or a significant main effect, we repeated the analysis considering all the different traits together to assess the specificity of their significant effects. When a personality trait showed a significant moderation effect, we conducted follow-up analyses to further examine the source of this interaction. For these analyses, we used the Pearson coefficient to analyze the correlation between the target significant personality trait scores and the corresponding significant main and/or interaction effects from the ANCOVAs. However, to help visualize the effects of personality traits, we have included figures in which traits are represented as median-based grouping variables.

As stated in the Method section, most of the participants whose data were excluded from the analyses did not avoid the US on any of the avoidance learning trials. To look more closely at this absence of avoidance, we computed the excluded participants’ mean response on CS+ trials throughout the avoidance learning phase. The results (M = 0.49, SD = 0.43, Max = 1.92, Min = 0) revealed that these participants, especially when compared with the included participants (M = 10.89, SD = 8.46, Max = 33.33, Min = 0.5) showed a greater tendency to not respond on CS+ trials. Estimating the US as non-aversive may be one reason for not avoiding the US. However, we did not find any significant difference between the included and excluded participants in terms of aversiveness ratings [M = 6.79, M = 7, t(143) = 0.79, p = .432]. We did not explore alternative explanations for the low avoidance response rates of the excluded participants such as a failure to understand the task instructions or a lack of attention. In any case, even though the results of the statistical analyses are practically the same when these participants are included, we decided to exclude them because studying the effects of extinction and renewal on avoidance necessarily requires the prior acquisition of the avoidance response.

Participants judged the aversiveness produced by the sound used as a US on a 0 to 9 rating scale on two occasions, once before starting the task (M = 7.14; SD = 1.55), and once again at the end of the task (M = 6.63; SD = 1.55). Both ratings differed significantly, t(114) = 2.21, p = .029. Even at the end of the task, the aversiveness of the sound was of a moderate-to-high magnitude.

As expected, the correlation between P-IU and TA was significant (r_p = .533, p < .001). See Table S1 in the supplementary materials for additional results concerning I-IU, DT, and the short version of the IU scale (IU-12).

Figure 2 shows participants’ mean relief ratings as a function of CS Type and Trial Block. Again, as in the Pavlovian Phase, ratings after the presentation of the safety cue (CS-) were lower than those reported after the two threat cues, CS+_Av and CS+_Unav. During this phase, in which participants had the possibility to avoid the US, the decrease in ratings across trials was of a greater magnitude for the safety cues than the threat cues (see Figure 3). A repeated measures ANOVA conducted on these ratings confirmed these impressions, yielding a significant main effect of CS Type and a significant CS Type x Trial Block interaction, F(1.40, 210.66) = 19.38, p < .001, $ηp2$ = .114 and F(8.14, 1229.57) = 3.34, p = .001, $ηp2$ = .022, respectively.

Further analyses served to help interpret these significant effects. First, we compared CS+_Av and CS+_Unav to evaluate participants’ discrimination between both cues. This analysis revealed a significant CS Type x Trial Block interaction, F(4.71, 710.64) = 3.19, p = .009, $ηp2$ = .021 (all other F’s < 2.40, and p’s > .124). This interaction might suggest that the participants learned to discriminate between CS+_Av and CS+_Unav trials. However, inspection of Figure 2 suggests that the CS Type x Trial Block interaction may have been driven by spurious and non-systematic differences between CS Type conditions on some trial blocks. Second, we compared participants’ ratings on CS+_Av with those on CS-, which yielded a significant effect of CS Type, F(1, 151) = 20.63, p < .001, $ηp2$ = .12, and Trial Block, F(3.8, 574.36) = 4.09, p = .001, $ηp2$ = .026. Finally, we compared participants’ CS+_Unav and CS- ratings, which revealed a significant main effect of CS Type, F(1, 151) = 22.67, p < .001, $ηp2$ = .131, and a CS Type x Trial Block interaction, F(4.47, 675) = 5.49, p < .001, $ηp2$ = .035. Overall, these results strongly suggest that the participants experienced higher relief after CS+ than after CS- trials, which presumably resulted from experiencing more distress during the former than the latter trial type.

Individual differences. Two independent repeated measures ANCOVAs were conducted including P-IU and TA as covariates. The only significant effect was the CS Type x Trial Block x P-IU interaction, F(8.16, 1224.30) = 2.10, p = .032, $ηp2$ = .014 (all other Fs < 2.81, and ps > .081). This interaction was still significant even after including TA as an additional covariate, F(8.16, 1215.25) = 1.99, p = .043, $ηp2$ = .013. Follow-up analyses were conducted to analyse this three-way interaction. In order to help visualise the modulating impact of P-IU on participants’ ratings, we treated this trait as a grouping variable by using the median to split the sample into two groups, and the CS+_Av and CS+_Unav ratings were collapsed into a single mean per participant and trial block. Inspection of Figure 3 suggests that the decrease in P-IU was associated with a steeper increase in the discrimination between CS+ and CS- throughout the trial blocks. Consistent with this impression, we found a significant negative correlation between P-IU and the difference between participants’ CS+_Av vs CS- discrimination (i.e., CS+_Av minus CS-) in the last trial block and the first trial block (last minus first), r = -.201, p = .013. When participants’ ratings for CS+_Av and CS+_Unav were collapsed into a single mean per participant and trial block, we also found a significant negative correlation between P-IU and the difference between the CS+ vs CS- discrimination in the last and first trial block, r = -.173, p = .033. The correlations involving the remaining CS pairs failed to reach significance.

Figure 4 shows the frequency of avoidance responses as a function of CS Type and Trial Block. Participants responded more frequently to the threat cues (CS+_Av and CS+_Unav) than the CS- cue. Though the number of responses to the threat cues increased across trials, this number remained relatively stable across trials for the CS- cue. A repeated measures ANOVA conducted on avoidance frequency served to confirm this impression, yielding significant main effects of CS Type, F(1.15, 227.99) = 90.03, p < .001, $ηp2$ = .374, Trial Block, F(3.36, 506.74) = 31.22, p < .001, $ηp2$ = .171, and the CS Type x Trial Block interaction, F(7.42, 111.81) = 9.11, p < .001, $ηp2$ = .057. A repeated measures ANOVA focused only on CS+_Av and CS+_Unav revealed only a significant effect of Trial Block, F(3.62, 546.82) = 35.103, p < .001, $ηp2$ = .19. Neither the main effect of CS Type [F(1, 151) = 2.66, p = .105] or the CS Type x Trial Block interaction effect [F(4.26, 642.80) = 1.14, p = .339] were significant. Thus, even though avoidance responses were unable to prevent the US on CS+_Unav trials, participants showed a similar pattern of responding to that observed on CS+_Av trials.

Individual differences. Two independent repeated measures ANCOVAs were conducted including P-IU and TA as covariates. These analyses revealed only a significant CS Type x TA interaction, F(1.54, 230.97) = 8.33, p = .001, $ηp2$ = .053 (all other Fs< 1.21, and ps > .291). This interaction was still significant even after including P-IU as an additional covariate, F(1.54, 228.77) = 7.8, p = .001, $ηp2$ = .05. Figure 5 helps visualise the origin of this interaction by showing the participants’ avoidance response frequency as a function of CS Type (CS+ vs CS-), and TA group (TA was treated as a grouping variable by using the median to split the sample into two groups). It appears that the increase in TA was positively associated with an increase in avoidance response frequency only in the CS- condition. This impression was confirmed by a significant positive correlation between TA and avoidance frequency throughout the whole instrumental training phase on CS- trials, r = .223, p = .006, but not CS+_Av trials (r = -.107, p = .19), or CS+_Unav trials (r = -.053, p = .517).

Figure 6 shows participants’ mean relief ratings as a function of CS Type and Trial Block. A repeated measures ANOVA revealed a significant main effect of CS Type, F(1.30, 196.16) = 32.15, p < .001, $ηp2$ = .176, and Trial Block, F(1.98, 298.88) = 78.56, p < .001, $ηp2$ = .342, as well as a significant CS Type x Trial Block interaction, F(6.46, 974.66) = 3.15, p = .004, $ηp2$ = .020. Further analyses served to explore the origin of this interaction effect. First, relief ratings after CS+_Av trials were lower than those after CS+_Unav trials, F(1, 151) = 10.54, p = .001, $ηp2$ = .065, and both cases showed a similar decrease, since only the main effect of Trial Block was significant, F(2.06, 310.68) = 77.82, p < .001, $ηp2$ = .340, but not the CS Type x Trial block interaction [F(5.72, 862.95) = 0.63, p = .702]. Thus, paradoxically, the experimental manipulation of the effectiveness of the avoidance response affected participants’ performance during the Extinction but not the Avoidance learning phase. Unlike the Avoidance phase, relief ratings were higher after CS+_Unav trials than CS+_Av trials, suggesting that the former trial type was regarded as more distressful than the latter. Second, we compared participants’ CS+_Av and CS- ratings, finding a significant main effect of CS Type, F(1, 151) = 35.21, p < .001, $ηp2$ = .189, Trial Block, F(2.03, 307.19) = 75.35, p < .001, $ηp2$ = .333, and a significant CS Type x Trial Block interaction, F(3.82, 577.26) = 3.97, p = .004, $ηp2$ = .026. Finally, a comparison of CS+_Unav and CS- ratings yielded a significant main effect of CS Type, F(1, 151) = 35.82, p < .001, $ηp2$ = .192, Trial Block, F(2.05, 310.05) = 71.35, p < .001, $ηp2$ = .321, and a CS Type x Trial Block interaction, F(3.85, 581.6) = 3.71, p = .006, $ηp2$ = .024. These results indicate that relief ratings were higher for both threat cues than for the CS- cue. Additionally, and as can be seen in Figure 6, relief ratings decreased to a greater extent throughout CSs+ than CS- trials, as confirmed by a post-hoc analysis comparing the decrement in the ratings from the first to the last block, t(151) = 2.43, p = .016, d = 0.13.

Figure 6 also shows that, even at the end of this phase, not only was extinction incomplete (or had not reached asymptote), but relief ratings also differed between the trial types. A final repeated measures ANOVA on relief ratings in the last block of extinction training showed a significant main effect of CS Type, F(1.49, 224.39) = 12.41, p < .001, $ηp2$ = .076. Post-hoc comparisons confirmed that ratings after CS+_Unav were higher than after CS+_Av trials, t(151) = 3.96, p < .001, d = 0.08. Additionally, CS+_Av and CS+_Unav trials both differed from those after the safety CS- trials, t(151) = 2.22, p = .028, d = 0.18 and t(151) = 4.11, p < .001, d = 0.33, respectively.

Individual differences. None of the two independent repeated measures ANCOVAs conducted revealed a significant effect involving any of the covariates (all F’s < 1.74, and p’s > .188).

Figure 7 shows the frequency of avoidance responses as a function of CS Type and Trial block. The overall number of avoidance responses diminished across the extinction phase, as expected. This decrease was more rapid for the two threat cues than the safety cue, which was also expected. These impressions were confirmed by a repeated measures ANOVA conducted on participants’ avoidance responses, which yielded a main effect of CS Type, F(1.74, 262.50) = 51.34, p < .001, $ηp2$ = .254, Trial Block, F(2.36, 356.72) = 92.33, p < .001, $ηp2$ = .379, and CS Type x Trial Block, F(5.06, 764.51) = 24.24, p < .001, $ηp2$ = .138. No significant difference was found between avoidance responses on CS+_Av and CS+_Unav trials [F(1, 151) = 2.96, p = .088]. Therefore, we compared avoidance responses to both threat cues (i.e., collapsing responses across CS+_Av and CS+_Unav trials) with those on CS- trials. Again, the main effects of CS Type and Trial Block were significant, F(1, 151) = 75.15, p < .001, $ηp2$ = .332 and F(2.29, 346) = 82.31, p < .001, $ηp2$ = .353, respectively, as well as the CS Type x Trial block interaction, F(2.43, 367.43) = 37.94, p < .001, $ηp2$ = .201.

As in the case of relief ratings, Figure 7 also shows that even at the end of the Extinction phase, the frequency of avoidance responses differed according to trial type. A post-hoc comparison revealed that participants responded more frequently to the threat cues (i.e., collapsing CS+_Av and CS+_Unav responses) than the safety cue, t(151) = 3.56, p = .001, d = 0.29. Thus, the extinction of avoidance responses was not asymptotic or complete at the end of extinction training.

Individual differences. Two independent repeated measures ANCOVAs were conducted including P-IU and TA as covariates. These analyses yielded three significant effects, namely, a Trial Block x P-IU interaction, F(2.40, 359.74) = 2.95, p = .044, $ηp2$ = .019; a Trial Block x TA interaction, F(2.39, 357.74) = 4.77, p < .001, $ηp2$ = .031; and a second order CS Type x Trial Block x TA interaction, F(5.20, 779.31) = 2.56, p = .024, $ηp2$ = .017.

We then introduced both covariates in another ANCOVA to assess whether any of the effects found was specific to either P-IU or TA. In this case, only the CS Type x Trial Block x TA interaction effect was significant, F(5.18, 770.5) = 2.43, p = .032, $ηp2$ = .016.

Follow-up analyses were conducted to further explore this three-way interaction. First, we conducted an ANCOVA comparing avoidance responses on the CS+_Av and CS+_Unav trials, which yielded only a significant Trial Block x TA interaction, F(2.47, 370.09) = 5.5, p = .002, $ηp2$ = .035 (other Fs < 0,552, and ps > .692). Considering that we did not find a significant main or interaction effect of CS type, we decided to collapse the avoidance responses to the threat cues (i.e., CS+_Av and CS+_Unav) and compare these with the responses evoked on CS- trials. Again, we found a significant three-way interaction, F(2.49, 373.55) = 3.84, p = .015, $ηp2$ = .025. To further explore this effect, we then conducted a correlation analysis between TA and the magnitude of the decrement in avoidance responses (i.e., responses in the first block minus the last block), finding a significant correlation only in the CSs+ trial type, r = -0.261, p = .001. As Figure 8 shows, participants scoring low in TA showed a larger decrease of avoidance responses across trials for the threat cues than their high TA counterparts, whereas the decrement on CS- trials seems to be unrelated to TA scores. Again, note that TA is treated as a grouping variable using the median to split the sample only for visualization purposes.

Figure 9 shows participants’ mean relief ratings as a function of CS Type and Trial. This figure shows a clear decrease in relief ratings across trials, with a less sharp decline throughout CS- trials. Ratings after CS+_Av and CS+_Unav did not appear to differ. Consistent with this impression, a repeated measures ANOVA revealed main effects of CS Type, F(1.28, 192.79) = 17.65, p < .001, $ηp2$ = .105, and Trial, F(2.19, 330.56) = 56.13, p < .001, $ηp2$ = .271, as well as a significant CS Type x Trial interaction, F(8.66, 1307.16) = 2.93, p = .002, $ηp2$ = .019. The difference between ratings after CS+_Av and CS+_Unav trials was far from significant [F(1, 151) = .030, p = .863]. The source of the CS Type x Trial interaction is the steeper decline in ratings on the threat cue trials compared with CS- trials. A repeated measures ANOVA revealed a significant main effect of CS Type (i.e., threat cues vs. safety cue), F(1, 151) = 21.18, p < .001, $ηp2$ = .123, Trial, F(2.20, 332.07) = 56.32, p = .000, $ηp2$ = .265, and a significant CS Type x Trial interaction, F(4.82, 727.47) = 3.72, p = .003, $ηp2$ = .024.

Individual differences. Two independent repeated measures ANCOVAs were conducted on relief ratings including P-IU and TA as covariates. These analyses revealed only a significant Trial x P-IU interaction, F(2.21, 331.25) = 3.23, p = .002, $ηp2$ = .021 (all other Fs < 1.87, and ps < .054). An additional ANCOVA with P-IU and TA as covariates showed that this interaction was still significant, F(2,19, 325,76) = 3.18, p = .039, $ηp2$ = .021. To further explore this interaction, a correlation analysis was conducted between P-IU and the magnitude of the decrement in relief ratings (relief ratings from the first trial block minus the last trial block) collapsed across all CSs. Although the correlation failed to reach significance (r = -.12, p = .139), Figure 10 shows that relief ratings tended to decrease across trials at a faster rate in low P-IU participants when compared with the high P-IU group.

Figure 11 shows the frequency of avoidance responses as a function of CS Type and Trial. The overall number of avoidance responses diminished across the re-extinction phase, although the slopes differed depending on the trial type. Repeated measures ANOVAs conducted on participants’ avoidance responses yielded a main effect of CS Type, F(1.88, 283.19) = 11.22, p < .001, $ηp2$ = .069, Trial, F(2.81, 424.11) = 13.63, p < .001, $ηp2$ = .083, and a significant CS Type x Trial interaction, F(7.17, 1082.18) = 3.54, p = .001, $ηp2$ = .023. Avoidance frequency on CS+_Av trials was higher than on CS+_Unav trials, F(1, 151) = 4.67, p = .032, $ηp2$ = .030 and declined more sharply in the former than in the latter condition, F(4.19, 632.16) = 2.46, p = .042, $ηp2$ = .016. This is consistent with the idea that participants were more confident in the effectiveness of the avoidance response on CS+_Av trials than CS+_Unav trials, even though the US was withheld on every trial of the re-extinction phase. Unsurprisingly, avoidance frequency differed between CS+_Av and CS- trials both in magnitude and the rate at which it diminished across trials. The number of responses after CS+_Av was higher than after the safety cue and diminished to a greater extent across trials, F(1, 151) = 17.71, p < .001, $ηp2$ = .105, F(4.44, 670.85) = 5.44, p < .001, $ηp2$ = .035; respectively. Finally, the number of responses after CS+_Unav was higher than after the safety cue, F(1, 151) = 8.32, p = .004, $ηp2$ = .052, although no difference was found in the rate of extinction [F(3.83, 578.82) = 2.28, p = .062].

Individual differences. None of the two independent repeated measures ANCOVAs showed a significant effect of any of the covariates (all Fs < 1.20, and ps > .211).

In this section, we report the results of the analyses of the renewal effect by examining how the behavioral measures differed between the last block of the extinction phase and the first trial of the re-extinction phase.

Figure 12 shows the participants’ mean relief ratings as a function of CS Type and Phase. A repeated measures ANOVA conducted on these ratings revealed significant main effects of CS Type, F(1.62, 244.71) = 14.16, p < .001, $ηp2$ = .086, and Phase, F(1, 151) = 51.73, p <.001, $ηp2$ = .255, but no significant CS Type x Phase interaction [F(1.90, 286.14) = 1.46, p = .234, $ηp2$ = .010]. Thus, relief ratings showed an overall (and non-differential) recovery from the last block of the extinction phase to the first trial of the re-extinction phase. However, as such recovery also affected relief ratings in the CS- condition, it is not possible to confirm that a renewal effect occurred, at least when measured by participants’ relief ratings.

Individual differences. Two independent repeated measures ANCOVAs were conducted including P-IU and TA as covariates. These analyses revealed that only the second order CS Type x Phase x TA interaction was significant, F(1.87, 281.03) = 3.44, p = .036, $ηp2$ = .022. This interaction was still significant after including both P-IU and TA as covariates in an additional ANCOVA, F(1.87, 278.63) = 3.64, p = .030, $ηp2$ = .024. Figure 13 shows participants’ relief ratings as a function of CS Type, Phase, and TA group. Again, TA was treated as a grouping variable to help visualize the modulating impact of this trait. To better understand the origin of this interaction, we compared relief ratings after CS+_Av trials with those after CS+_Unav trials. The analysis revealed a significant CS Type x Phase x TA interaction, F(1, 150) = 4.39, p = .038, $ηp2$ = .028. To further explore this interaction, we then conducted a correlation analysis between TA and the difference in the magnitude of renewal in participants’ relief ratings (relief ratings on the last trial of the extinction phase minus those on the first trial in the re-extinction phase) between CS+_Av and CS+_Unav trials (CS+_Av minus CS+_Unav). This analysis revealed a significant negative correlation, r = -.169, p = .038. As Figure 13 shows, participants scoring low in TA showed a greater renewal effect to CS+_Av compared to CS+_Unav. Second, we compared participants’ CS+_Av and CS- ratings, finding a significant CS Type x TA interaction, F(1, 150) = 4.89, p = .029, $ηp2$ = .032, and CS Type x Phase x TA interaction, F(1, 150) = 6.99, p = .009, $ηp2$ = .045. It is clear from Figure 13 that the difference in the renewal effect between the CS+_Av and CS- condition tended to be greater for those participants scoring low in TA. Consistent with this impression, we found a negative correlation between TA and the difference between participants’ ratings on CS+_Av and CS- trials in the last block of extinction and the first trial of re-extinction (last minus first), r = -.211, p = .009. Finally, we compared participants’ CS+_Unav and CS- ratings, which revealed no significant effect or interaction involving TA (all Fs < 2.14, and ps > .145).

Figure 14 shows the frequency of avoidance responses as a function of CS Type and Phase. A repeated measures ANOVA revealed a significant main effect of CS Type, F(1.87, 283.39) = 12.18, p < .001, $ηp2$ = .075, and Phase, F(1, 151) = 46.73, p < .001, $ηp2$ = .236, as well as a significant CS Type x Phase interaction, F(1.74, 263.15) = 3.28, p = .046, $ηp2$ = .021. These findings indicate a general increase in the frequency of avoidance responses due to the context change, but that the magnitude of this effect differed depending on the specific CS involved. To further explore this interaction, we conducted an ANOVA comparing avoidance responses on the CS+_Av trials to those on CS+_Unav trials, finding a significant main effect of Phase, F(1, 151) = 47.45, p < .001, $ηp2$ = .239, and a marginally significant main effect of CS type, F(1, 151) = 3.57, p = .061, $ηp2$ = .023.

Then, we compared the increase in avoidance frequency on CS+_Av and CS+_Unav trials with the increase on CS- trials. The difference between the response frequency on CS+_Av and CS- trials indicates a renewal effect, that is, avoidance responses increased to a greater extent on CS+_Av than CS- trials, F(1, 151) = 8.10, p = .005, $ηp2$ = .051. However, the renewal effect for CS+_Unav was marginal [F(1, 151) = 3.24, p = .074, $ηp2$ = .021].

Individual differences. Two independent repeated measures ANCOVAs were conducted including P-IU and TA as covariates. These analyses revealed that only the Phase x P-IU interaction was significant, F(1, 150) = 4.31, p = .040, $ηp2$ = .028 (for the remaining comparisons involving both covariates, all Fs < 3.52, and ps > .060). Nevertheless, this interaction lost significance after including TA as an additional covariate (all the comparisons involving both covariates: Fs < 1.61, and ps > .206). To understand the origin of this P-IU modulatory effect, a correlation analysis was conducted between P-IU and the magnitude of the renewal effect (i.e., differences in the mean frequency of responses collapsed across all CS types between the first re-extinction trial and the last trial block of the extinction phase). This analysis showed that the magnitude of renewal was negatively correlated with P-IU, r = -.167, p = .040, which suggests a stronger renewal effect in participants scoring low in P-IU compared with their high P-IU counterparts.

The present study assessed the role of P-IU and TA in the extinction and relapse of avoidance and distress (inferred from post-trial relief ratings) by using a discriminated, free operant procedure in which participants had to respond repeatedly to avoid an uncertain aversive US. Our learning task consisted of an initial Pavlovian learning phase in which participants could learn that two CSs+ predicted the US, whereas a CS- predicted its absence. In a subsequent avoidance learning phase, participants learned to avoid the aversive US on CS+_Av, but not CS+_Unav trials. During a subsequent extinction phase, and unlike previous studies, participants could still perform the avoidance response, which did not prevent them from noticing that the US no longer followed the mere occurrence of the CSs+. Consequently, response rates and relief ratings decreased on CSs+ and CS- trials and tended to converge throughout the course of extinction. Relapse of avoidance and distress was assessed through a re-extinction phase conducted in a different context from that in which the avoidance and extinction phases took place. An overall recovery of avoidance was found when comparing participants’ responses on the last trial block of extinction and the first trial block of the re-extinction phase. Moreover, a renewal effect was evidenced by a stronger recovery of avoidance on CS+_Av than CS- trials. Concerning relief ratings, we only found an overall recovery effect that did not interact with CS type. Finally, analyses of the data from the re-extinction phase revealed similar results to the extinction phase, that is, both avoidance response rates and relief ratings decreased on CS+_Av, CS+_Unav, and CS- trials, and tended to converge throughout this phase.

Regarding the modulating role of P-IU and TA in the acquisition, extinction, and renewal effects, the results are somewhat unclear when examined in detail. Nonetheless, a clear picture seems to emerge when considering the entire body of results. In general, we found evidence to clearly support the hypothesis that both P-IU and TA are associated with individual differences in the acquisition and extinction of avoidance and distress (as measured through post-trial relief ratings). For instance, in the avoidance acquisition phase, an increase in P-IU was found to be associated with impaired discrimination in terms of differences in relief ratings on CSs+ and the CS- trials, whereas TA was found to have the same modulating effect on avoidance responses. In both cases, impaired discrimination was mainly due to heightened relief ratings and avoidance rates on CS- trials. This impairment in learned safety in individuals scoring high in anxiety-predisposing traits is consistent with previous findings on avoidance (Flores et al., 2018; San Martín et al., 2020) and other fear-related measures (Klingelhöfer-Jens et al., 2022; Morriss, Macdonald, et al., 2016; Wong & Lovibond, 2018: Wroblewski et al., 2022). Crucially, increased TA was associated with an overall less steep decline in avoidance rates, and slower extinction of avoidance. In other words, avoidance rates on CSs+ and CS-trials converged more slowly in high TA participants. In a similar vein, increased P-IU was associated with an overall less steep decline in distress (as measured through post-trial relief ratings) in the re-extinction phase. To the best of our knowledge, this is the first clear demonstration that impaired active avoidance extinction is associated with a risk factor for anxiety using a procedure in which avoidance reduction is assessed during the course of extinction, similar to Pavlovian fear conditioning preparations (see Sheynin et al., 2014, for a comparable procedure based on passive avoidance).

Another detected regularity is related to the specific contribution of each personality factor. With the exception of one case, P-IU was related to effects on relief ratings, whereas TA was associated with effects on avoidance rate. This result is rather surprising considering previous results found in our laboratory. Recently, we have merged the data from all the experiments based on Flores et al.’s (2018) procedure conducted in our laboratory (N = 212), with very slight differences between them, and included P-IU, I-IU, and TA as covariates in a repeated measures ANCOVA. This analysis yielded a significant positive association only between P-IU and avoidance frequency (observed 1-β = .86), which runs counter to the results of the present study. An important aspect in which our procedure differs from that used in Flores et al.’s is the trial duration. In Flores et al. (2018) (see also Cobos et al., 2022), the CS was presented for 20 seconds, which is almost four times the duration of the CS used in the present experiment. A long CS duration may have promoted over-engagement in avoidance reactions as a strategy to cope with the distress caused by waiting in uncertainty for the termination of the trial, especially in participants scoring high in P-IU. In contrast, in our experiment, the trials were 6 s in duration, which may have been insufficient to increase avoidance reactions to cope with the distress produced by waiting in uncertainty. Of course, this explanation is speculative and falls short of explaining why our procedure was still effective in detecting individual differences in distress (as measured through post-trial relief ratings) associated with P-IU (see the previous paragraph). Therefore, further research is needed to better understand the results found in the present study regarding the relationship between P-IU and avoidance rate, as well as the discrepancies found between the present results and those of our previous studies based on Flores et al’s (2018) procedure.

In addition to extinction, the other main objective of our study was to explore individual differences in renewal of avoidance and relief. Contrary to our expectations, we found evidence that might suggest a weaker renewal effect associated with an increase in TA, as confirmed by a three-way CS Type x Phase x TA interaction for relief ratings. This interaction was mainly due to greater differential recovery of relief ratings (i.e., the amount of recovery in CS+_Av vs. CS- trials) in low TA participants (see Figure 12). Additionally, an increase in P-IU was associated with less overall recovery of avoidance. However, these findings should be interpreted with caution. Although our extinction procedure seems to have been effective in reducing avoidance, the number of extinction trials was insufficient to achieve complete extinction. This is evidenced by significantly higher avoidance rates and relief ratings in the CS+ condition compared with the CS- condition on the last extinction trial block. Incomplete extinction may hamper the assessment of the renewal effect and leaves open the possibility that the renewal effect found is conditional upon residual differential avoidance or relief. Moreover, two further limitations should be considered when analysing individual differences in renewal. First, the avoidance and relief rating baselines on the last extinction trial block tend to correlate positively with TA and P-IU. Second, as stated above, avoidance extinction was slower in the high TA group than their low TA counterparts. In fact, inspection of Figure 8 (trial blocks 7 and 8) reveals that the avoidance rates on CS+_Av and CS+_Unav trials appear to reach almost maximum convergence with avoidance rates on CS- trials in the low TA participants, whereas such convergence is clearly weaker in participants in the high TA participants. These differences found at the end of the extinction phase pose serious problems for drawing any firm conclusion regarding individual differences in renewal associated with anxiety-predisposing traits. Future experiments based on our procedure should include more extinction trials to overcome this limitation. At the same time, our results highlight the importance of assessing the effects of extinction on avoidance during the extinction phase to ensure that participants’ scores in anxiety-predisposing traits do not correlate with effects that may hinder an accurate assessment of renewal effects. In this regard, the main message to emerge from our analyses could be that it is very difficult to draw clear conclusions regarding individual differences in extinction and recovery effects from previous studies measuring extinction with a response-prevention procedure.

Another interesting issue concerns the consistency between the results found in avoidance responses and relief ratings. In the avoidance acquisition phase, the avoidance responses followed an expected pattern, since responding tended to increase throughout CSs+ trials and remained low on CS- trials. In principle, one would expect a decrease in relief ratings throughout CS+_Av trials (see Cobos et al., 2022; Papalini et al., 2021; San Martín et al., 2020; Vervliet et al., 2017), but our results showed a rather flat distribution, with a level of responding that was not significantly different from the relief ratings observed throughout CS+_Unav trials. This pattern of results suggests that participants may have found it difficult to discriminate between CS+_Av and CS+_Unav. Consequently, many participants may have remained somewhat uncertain about the occurrence of the US despite making avoidance responses on CS+_Av trials throughout most of the avoidance acquisition phase, which may explain the flat distribution of relief ratings. In the subsequent training phases, however, some differences between the CS+_Av and the CS+_Unav conditions emerged, which seem to reflect some residual effect of the different training conditions administered in the previous avoidance acquisition phase. Specifically, relief ratings were higher on CS+_Unav than CS+_Av trials in the extinction phase, which strongly suggests that participants felt less distress on CS+_Av trials because they were more confident of avoiding the US than on CS+_Unav trials. Additionally, participants made more avoidance responses on CS+_Av than CS+_Unav reextinction trials (particularly on the early trials), which suggests that participants had, to some extent, learned that avoidance responses were effective in the presence of CS+_Av but not in the presence of CS+_Unav. Finally, in both the extinction and the re-extinction phases we found a reduction in relief ratings and avoidance rates consistent with an extinction effect, which has also been reported in previous studies (Papalini et al., 2021; Vervliet et al., 2017). Overall, however, it seems that avoidance responses extinguished more rapidly than relief ratings.

Another limitation concerns the high percentage of female participants (79%) in our sample. This gender bias could have impacted the results, especially given that our aim was to evaluate individual differences related to anxiety-predisposing traits. Moreover, although most of our participants were white European, we did not request information regarding ethnicity.

To conclude, our results provide further support for the role of vulnerability traits in anxiety-related disorders, as measured by heightened distress and frequency of avoidance responses to safety signals during avoidance acquisition. In addition, our study is the first to demonstrate a renewal effect on active avoidance after an extinction procedure not based on response prevention, as well as impaired avoidance extinction associated with an increase in TA, and a slower overall reduction in relief ratings during re-extinction associated with increase in P-IU. Overall, our experimental model based on scenarios in which there is uncertainty about future threats has shown to be useful for analysing individual differences in vulnerability traits for anxiety disorders, as revealed by relevant learning measures.

Preparation of this manuscript was supported by Grants PGC2018-096863-B-I00 and UMA18-FEDERJA-051 from the Spanish Ministerio de Economía y Competitividad and the Spanish regional government Junta de Andalucía, respectively. Tania Valle was hired as a PhD researcher on a budget from Grant UMA18-FEDERJA-051. María J. Quintero has been awarded a PhD fellowship from the Spanish Ministry of Science, Innovation, and Universities (FPU Programme, FPU18/00917).

Contributed to conception and design: PLC, FJL

Contributed to acquisition of data: TV, MJQ

Contributed to analysis and interpretation of data: FJL, TV, MJQ, PLC

Drafted and/or revised the article: PLC, FJL, TV, MJQ

Approved the submitted version for publication: PLC, FJL

Neither this manuscript nor any similar one has been published or submitted for publication elsewhere. All the authors have read the manuscript and have approved this submission. The authors report no conflicts of interest. None of them has any financial interest or activity that might be seen as influencing the research.

The raw data file has also been uploaded to the Open Science Framework repository and can be found by clicking on the link https://osf.io/ew52z/.