Multimodal assessment of the role of intolerance of uncertainty in fear acquisition and extinction

Personality traits linked to internalizing disorders influence the way we develop fears, but also how we regain a sense of safety. In the present study, we investigated the effect of intolerance of uncertainty (IU) on defensive responses using a differential fear conditioning protocol with an extinction phase. The conditioned stimulus was associated with an aversive sound (90 dB) in 75 % of the presentations during acquisition. A final sample of 176 participants completed the experiment. We measured self-reports of associative (expectancy of the unconditioned stimulus in acquisition) and evaluative learning (arousal and valence), and both physiological (skin conductance response) and electrocortical responses (steady-state visually evoked potentials, ssVEPs; late positive potentials, LPP) to the conditioned stimuli. Our results show that IU ’ s impact is limited, with no effect in both acquisition and extinction. These findings emphasize the necessity of large samples in research on inter-individual differences and contribute to our understanding of how IU may or may not be involved in fear and safety learning processes considering multiple aspects of fear responding.


Introduction
Learning to associate sensory cues with threats (aversive learning) is critical for enabling the analysis of the environment for potential threats.The ability to learn these associations and to keep them updated plays a significant role in our survival and general well-being.Alterations in aversive learning are characterized by displaying fear responses to an element that has no threatening features (i.e., overgeneralization) or by consistently showing fear responses towards an element that is no longer associated with threat anymore.This alteration has been linked to the etiology and maintenance of anxiety disorders (Beckers et al., 2023;Carpenter et al., 2019;Craske et al., 2018;Pittig et al., 2018).
Aversive learning has been studied mainly via experimental fear conditioning paradigms (Lonsdorf et al., 2017), in which a threat association is established by presenting a neutral stimulus (CS+; threat stimulus) along with an aversive unconditioned stimulus (US; e.g., an electric shock or a loud noise).In contrast, a safety association is established by presenting another stimulus (CS-; safe stimulus) without an accompanying aversive stimulus.This phase is called fear acquisition.In addition, fear conditioning paradigms can contain an extinction phase, where a former threat stimulus is re-introduced without the accompanying US (Hermans et al., 2006;Sangha et al., 2020).These two phases of Pavlovian conditioning allow researchers to answer two different questions: during the acquisition phase, they investigate 'why do fears develop?', while during extinction they explore 'why do fears persist when no longer relevant?' (Pittig et al., 2018).
In human fear conditioning studies, conditioned fear responses (CR) are quantified using a range of self-report and physiological measures (Lonsdorf et al., 2017).These measures provide valuable insights into the various facets of fear discrimination and its underlying systems.Self-report measures allow for the assessment of conscious experiences associated with the conditioned stimuli.These measures include ratings of fear, valence, arousal, and US-expectancy.Physiological measures offer an additional level of how fear affects us, and an objective means to bypass potential self-report biases by assessing biological changes in the central and peripheral nervous systems.Some measures include steady-state visually evoked potentials (ssVEPs), which capture attention allocation to a stimulus through the oscillatory response of the visual cortex to an oscillating stimulus (Norcia et al., 2015;Vialatte et al., 2010).This signal has shown a discriminative response towards the CS+ and CS-in fear conditioning experiments (Miskovic & Keil, 2013;Wieser et al., 2014), although in some studies this discrimination and expected generalization effects are absent (e.g., Aslanidou et al., 2023;Stegmann et al., 2020).The late positive potential (LPP), an event-related potential (ERP), reflects arousal and is larger for pleasant and unpleasant stimuli compared to neutral ones (e.g., Cuthbert et al., 2000).Recently, it has been measured in differential conditioning paradigms and found enlarged LPP amplitudes for CS+ compared to CS- (Bacigalupo & Luck, 2018;Bauer et al., 2020;Bruchmann et al., 2021;Klein et al., 2023;Panitz et al., 2015Panitz et al., , 2018)).The skin conductance response (SCR) is associated with activity in the autonomic nervous system (Boucsein et al., 2012) and has been used to differentiate threat and safety responses (Dawson & Biferno, 1973;Mordkoff et al., 1967) and it is still commonly employed (for reviews using SCR see Beckers et al., 2023;Lonsdorf et al., 2017;Ojala & Bach, 2020).Combining these diverse measures in fear conditioning studies yields a comprehensive understanding of the different processes involved in fear discrimination.It is worth noting that these measures are often not aligned (Lonsdorf et al., 2017), and studies employing multiple measures provide a more holistic view of the fear system and its responses.
Analysis of the fear responses during the extinction phase can serve as a model for exposure therapy (Craske et al., 2018).A maladaptive reaction in this phase (e.g., persistent fear responses in the absence of threat) is thought to reflect resistance to exposure and has been linked to pathological anxiety (Duits et al., 2015).One possible important factor for this resistance to extinction learning may be uncertainty (Morriss, Zuj, et al., 2021;Van den Bergh et al., 2021), since in the extinction process the uncertainty related to threat and safe contingencies may maintain the conditioned response (Bouton, 2002).The degree to which the perception and emotional reaction to uncertainty impacts a person's life has been conceptualized as intolerance of uncertainty (IU).IU refers to the individual inability to endure uncertain or ambiguous situations and events (Carleton, 2012(Carleton, , 2016;;Freeston et al., 1994).It is connected to fear processes through the concept of "fear of the unknown" (Carleton, 2016;McEvoy et al., 2019) and threat anticipation (Grupe & Nitschke, 2013).People high in IU interpret uncertain events as aversive and threatening (Dugas et al., 2005).IU is closely related to a wide range of anxiety disorders and depression (Carleton, 2012), forming a transdiagnostic construct (Gentes & Ruscio, 2011;McEvoy & Mahoney, 2011;Shapiro et al., 2020).In the same vein, IU is considered a risk and a maintaining factor for internalizing disorders (McEvoy et al., 2019).Two sub-factors of IU have been proposed: Prospective IU (P-IU) pertains to the inclination for predictability and actively seeking certainty, whereas inhibitory IU (I-IU) refers to avoidance, behavioral inhibition, and feeling paralyzed in uncertain situations (Carleton, 2012).Thus, intolerance of uncertainty seems to be a candidate trait for a potential risk factor in maladaptive extinction learning and consequently, vulnerability for the maintenance of anxiety (Morriss, Zuj, et al., 2021;Morriss, Abend, et al., 2023).
The acquisition phase is thought to model the development of anxiety disorders (Mineka & Zinbarg, 2006;Pittig et al., 2018).Indeed, some evidence has shown that individuals with heightened and even clinical anxiety show stronger fear acquisition, however this evidence is inconsistent and limited to single cue acquisition paradigms (Duits et al., 2015;Lissek et al., 2005;Pittig et al., 2018).However, it has been argued that individual differences in fear acquisition may depend on the US's intensity or aversiveness.This intensity on one hand is required to be strong for a successful fear acquisition, while on the other hand a too strong intensity is thought to be responsible of blurring inter-individual differences in the participants' responses (Beckers et al., 2013(Beckers et al., , 2023;;Lissek et al., 2006).Therefore, when researchers are interested in how different people develop fear, they must find ways of adjusting from a strong to a weakened intensity (Lissek et al., 2006).One way of weakening this situation is by reducing the reinforcement rate of the CS-US association, which then will increase uncertainty.This amplifies the possible impact that an individual's level of IU can have in this phase, as previous research suggests that IU is more prominently associated with compromised fear responses in situations characterized by heightened uncertainty (Lonsdorf & Merz, 2017;Morriss, Zuj, et al., 2021).For instance, using a 37.5 % reinforcement rate, Kanen et al. (2021) found a worse SCR threat-safety discrimination for individuals with higher IU.In contrast, Chin et al. (2016) found that the IU level was positively related with the differential (CS+ vs. CS-) FPS response when the reinforcement rate was 50 % and no longer when it was 75 %.Surprisingly, with a similar approach Johnson et al. (2023) used a continuous (100 %) and partial (50 %) reinforcement rate and could not find an association of IU with SCR threat discrimination.However, when looking into IU's subfactors they found that, regardless of the reinforcement rate, higher I-IU was associated with diminished SCR discrimination and the opposite was the case for higher P-IU (i.e., heightened threat discrimination).In addition, a recent study by Mertens et al. (2022), analyzing 120 participants found that IU correlated positively with the fear ratings of the CS+, however this effect was no longer present after controlling for trait-anxiety and neither when it was tested on the differential responses (CS+ vs. CS-).In addition, this association was not observed in physiological responses (SCR and startle response) nor in expectancy ratings (Mertens et al., 2022).Furthermore, Klingelhöfer-Jens et al. (2022) found a positive correlation of IU (controlled for trait-anxiety) with fear ratings of CS+ (but not with SCR and startle responses).It is important to note that several studies have not found an effect of IU on fear responses during acquisition (Johnson et al., 2023;Mertens & Morriss, 2021;Morriss et al., 2015;Morriss, Christakou, et al., 2016;Morriss, 2019;Morriss, Saldarini, Chapman, et al., 2019;Morriss, Saldarini, & van Reekum, 2019;Morriss et al., 2020;Morriss & van Reekum, 2019;Sjouwerman et al., 2020;Wake et al., 2020Wake et al., , 2021;;Wroblewski et al., 2022).Thus, which role IU plays in fear acquisition is not clear.This ambiguity arises from the inconsistent influence of reinforcement rates and the varying presence of effects on ratings or physiological measures (Chin et al., 2016;Kanen et al., 2021).Considering the substantial body of research reporting a lack of a link between IU and these responses in fear acquisition, the effect of IU might be very small and dependent on certain aspects in the experimental design or IU does not exert a significant impact in this phase.This observation aligns with the concept of acquisition as a 'strong situation' (Lissek et al., 2006).Interestingly, even after adjustments are made to weaken this situation, the reviewed studies also fail to reveal a substantial role of IU in this phase.In our study we aim to contribute to the clarification of this relationship by analyzing the responses in the acquisition phase and exploring their link to the individual's level of IU.
The extinction phase of fear conditioning helps us to understand the persistence of fears, by testing the defensive responses that appear when the feared stimulus is presented without the aversive consequences.In a way, it reflects the individual's difficulties in trusting that the situation is safer than expected.The absence of US presentations triggers the establishment of a new safe association that competes with the former threatening one (Bouton, 1993;Bouton & King, 1983), progressively updating the responses until they are equivalent for both CS+ and CS-.This can be referred to as a successful extinction.A meta-analysis by Duits et al. (2015) found that anxiety patients in this phase show stronger fear responses towards the CS+ (small to medium effect size), and persistent discrimination of CS+ and CS-(small effect size), while there were no differences between groups regarding the responses to the CS-.It is relevant to mention that a more recent study (Pöhlchen et al., 2020) could not find these differences in the fear responses in extinction, while they further differentiated in their clinical sample between post-traumatic stress disorder and other anxiety disorders.
During extinction, the absence of US can be either instructed or uninstructed, the latter meaning that without being explicitly mentioned, the CS+ is no longer followed by the US and hence no longer threatening (Lonsdorf et al., 2017).Thus, in uninstructed extinction a period where the uncertainty about the CS-noUS link increases is created.This increase in uncertainty would affect individuals with higher IU particularly at the beginning of the extinction phase (Morriss, Zuj, et al., 2021).Morriss, Macdonald, et al. (2016) analyzed the time course of fear responses during extinction learning and found that lower IU showed a differentiation of CS+ and CS-in the SCR during early extinction which in late extinction disappeared (signaling successful extinction).Conversely, higher IU manifested heightened SCR values for both stimuli in early extinction, suggesting a generalization effect from the threat to the safe stimulus in this uncertain stage.Moreover, in late extinction, higher IU participants exhibited larger SCR for CS+ than CS-, signifying a resistance to extinction.Similarly, another study found also an unsuccessful extinction for participants with high IU and trait anxiety (Wroblewski et al., 2022).It is relevant to note that Wroblewski et al. (2022) used a delayed extinction design, conducting the extinction phase 24 h after the initial acquisition period.In their study, all participants exhibited higher fear responses (expectancy, arousal, and valence ratings) to the CS-at early extinction compared to the end of acquisition.This pattern resembles the effect found by Morriss, Macdonald, et al. (2016) on SCR in early extinction, but there it was limited to high IU individuals.Further, after the initial rise in responses to CSduring early extinction, the low IU/trait anxiety group decreased responses to both stimuli, achieving successful extinction at the end of the extinction phase.In contrast, the high IU/trait anxiety group continued showing a robust differentiation for CS+ vs. CS-expressed in the FPS and SCR, throughout extinction.Despite some differences in the previous studies, an effect that has been consistently found is the resistance to extinction in individuals with higher IU (Bauer et al., 2020;Morriss, 2019;Morriss, Christakou, et al., 2016;Morriss et al., 2015;Morriss, Macdonald, et al., 2016;Morriss, Saldarini, & van Reekum, 2019;Morriss & van Reekum, 2019;Wake et al., 2020Wake et al., , 2021;;Wroblewski et al., 2022).
The current study seeks to further unveil the nuanced interplay between IU and fear responses during both fear acquisition and extinction learning in a relatively large sample, aligning with recent calls for larger sample sizes in the field (Mertens et al., 2022).Fear responses were measured on multiple levels (ratings, SCR, LPP, and ssVEPs), providing a detailed understanding of how IU impacts physiological and subjective fear responses.Recognizing the dynamic nature of fear learning, each phase is divided into early and late sections, allowing a detailed exploration of the evolution of fear responses, and shedding light on critical moments characterized by heightened uncertainty (i.e., early uninstructed extinction).We expected a successful fear acquisition, characterized by higher ratings (valence, arousal, and US-expectancy) and greater amplitudes of SCR, LPP, and ssVEPs in response to the threat compared to the safety cue.Successful fear extinction should be manifested through a decrease in the discrimination of CS+ and CS-in respective variables.
We investigated the impact that IU has on fear responses and the specificity of this influence (Morriss, 2023) while controlling for the influence of other traits such as trait anxiety, worry, and depression.We expected to replicate previous results which found elevated intolerance of uncertainty to hinder extinction learning (e.g., Morriss, 2019;Morriss, Christakou, et al., 2016;Morriss, Saldarini, & van Reekum, 2019), evident through heightened fear responses to both safe and threat stimuli during early extinction (especially in SCR, Morriss, Wake, et al., 2021;and in LPP, Bauer et al., 2020; and exploratorily in the ssVEP and ratings).Moreover, we expected that threat-safety discrimination persists in all measured variables during the late extinction phase for individuals with high intolerance of uncertainty.Given aforementioned inconsistent findings about the influence of IU on fear acquisition, we had no directed hypothesis a-priori here, but rather exploratorily tested possible effects in all dependent variables as well.

Participants
In total, 213 psychology undergraduate students from Erasmus University Rotterdam were invited via an online recruitment system to participate in the lab session.Thirty-seven participants had to be removed (13 had missing data in at least one dependent variable; and twenty-four stopped the experiment), resulting in 176 participants for the analyses.The age range was 17-46 years (M = 21.3,SD = 3.30), 81 % were females, and 53 % were Dutch.A breakdown of the ethnic selfidentification was 71 % for White/Caucasian, 10 % for Asian, 5 % for Middle Eastern, 4 % for Black, 4 % for mixed, 2 % for Latin-American, and 3 % for other.Participants received course credit and a monetary compensation (€5) related to a decision-making task performed in the same session, but not reported in this article (see Section 2.4).A-priori exclusion criteria were family history of photic epilepsy, related to the flickering images in this experiment.Informed consent was obtained before participation, and the research protocol was approved by the ethical board of the Erasmus School of Behavioural and Social Sciences in accordance with the declaration of Helsinki.The lab sessions took place between September 2020 and July 2021.

Questionnaires
After finishing the experimental task, participants completed the following questionnaires (either in Dutch or English, for Dutch or international students, respectively): The Intolerance of Uncertainty Scale (Buhr & Dugas, 2002;Freeston et al., 1994;Dutch translation: de Bruin et al., 2006).It is a 27-item questionnaire that assesses the degree to which individuals find uncertain scenarios distressing and undesirable, as well as their reactions to them.The items are rated on a five-point Likert scale ranging from 1 ("not at all characteristic of me") to 5 ("entirely characteristic of me").Internal consistency was excellent for the total scale IU (α = 0.94).Total IU scores in the current sample ranged from 28 to 110 (M = 65.47;SD = 17.97).In addition to the total score, two subfactors can be extracted The Penn State Worry Questionnaire (PSWQ; Meyer et al., 1990;Dutch translation: van Rijsoort et al., 1999) is a 16-item self-report questionnaire that measures the trait of worry across the dimensions of excessiveness, generality, and uncontrollability.Each item is rated on a five-point Likert scale from 1 ("not at all typical of me") to 5 ("very typical of me").Internal consistency of this scale was good (α = 0.83).Scores in the current sample ranged from 25 to 72 (M = 44.80;SD = 10.63).
The trait subscale of the State Trait Anxiety Inventory, Form Y (STAIT; Spielberger et al., 1983;Dutch translation: Van der Ploeg, 2000) is a self-report questionnaire consisting of 20 items that measure someone's general propensity for anxiety.Respondents' rate items based on how they generally feel, using a four-point Likert scale from 1 ("almost never") to 4 ("almost always"), with higher scores indicating greater anxiety (range: 20 to 80).Internal consistency of this scale was good (α = 0.80).Scores in the current sample ranged from 24 to 66 (M = 44.78;SD = 8.20).

Materials and stimuli
The conditioned stimuli (CS) were two pictures of female faces with a neutral expression from the NimStim Set of Facial Expressions (03F NE C, 10F NE C; Tottenham et al., 2009).The CSs were presented in the center of a 22-in.iiyama HM204DT-A computer screen and their size was 8.6 × 11.6 cm, the participants were seated at 1.60 m from the screen resulting in 3.08 × 4.15 • visual angle size.The images were converted to grey scale and luminance and brightness were kept constant.The background was constant throughout the experiment and was grey (R: 133, G: 133, B: 133).Each CSs was presented flickering at a driving frequency of 15 Hz. 1 The US consisted of a female scream taken from the International Digitized Affective Sounds database (Bradley & Lang, 2007) combined with white noise played through the room speakers for 1 s at 90 dB upon termination of the face stimulus (CS+).

Procedure
Upon the arrival at the lab, participants received a verbal description of the experimental procedure and signed the written informed consent.After this, they were seated in an isolated dimly lit room where the EEG and EDA electrodes were attached.Afterwards, they completed the Doors task (not reported here) with a duration of approximately 40 min.Then they had a break (10 min) and after this the Fear conditioning experiment took place (with an approximate duration of 25 min).Lastly, they completed a Flanker task (duration of around 10 min).This task is not part of this manuscript and is reported in Malbec et al. (2022).
The fear conditioning protocol consisted of three phases: habituation, acquisition, and extinction.During habituation, both CSs were presented 10 times each, while no US was presented.Participants were then instructed to pay attention to the faces presented and that one of them will be followed sometimes by a loud noise.During the acquisition phase, the CS+ was followed by the US in 75 % of the times, starting at CS+ offset, while the CS-was never associated with the US.Both CS-and CS+ were presented 20 times each.Finally, an uninstructed extinction phase followed, where CS+ and CS-were presented 20 times each and the US was never delivered.In all phases, the duration of the CSs was 5 s, while the inter-trial-interval (ITI), defined as the time between the offset of one stimulus and the onset of a following stimulus, had a random duration between 4.5 and 5 s, during which a white fixation cross was presented.The order of presentation was fully randomized in habituation and extinction phases, whereas in acquisition phase it was pseudorandomized meaning that the same stimulus was not presented more than twice in a row, and the first trial of the phase was always a CS+ that was linked with the US.The associations of each picture as CS+ or CSwas counterbalanced across participants.
After each phase, participants' arousal and valence for each CS was assessed by using the Self-Assessment Manikin scale (SAM; Bradley & Lang, 1994).The scales ranged from 1 as "calm" or "happy" to 9 "excited" or "unhappy", for arousal and valence, respectively.The scales were presented after the respective CS for 1 s.To explore contingency learning, we measured the US-expectancy using an analogous slider during the acquisition phase (after ten trials) and at its end (after 40 trials).It consisted of the question "How likely is to hear a noise after this image?"and 0 % and 100 % were displayed in both ends of the slider, the corresponding CS image was presented on top of the slider until response.

Psychophysiological signals recording and processing
Continuous EEG was recorded using Ag/AgCl 32 electrodes arranged according to the 10/20 configuration.The electro-oculogram (EOG) was recorded using 4 electrodes, two above and below the left eye, and two at the outer canthus of each eye.Data were recorded using the Active-Two BioSemi system (BioSemi, Amsterdam, Netherlands).This system includes two additional electrodes, common mode sense (CMS) and driven right leg (DRL), which serve as online reference and ground.The signal was digitized at a 512 Hz sampling rate and 24-bit A/D conversion.All electrode impedances were kept below 20 kΩ.Additionally, EDA was recorded with two auxiliary channels, two 8 mm flat electrodes with a 0.05 M NaCl electrolyte gel.These electrodes were attached to the second phalanx of the middle and ring finger of the non-dominant hand after lightly cleaned with water.
For the ssVEP analysis, the signal was then transformed into the frequency domain using a Fast-Fourier-Transformation for the last 2000 ms of stimulus presentation to eliminate initial non-stationary ssVEP components and to highlight ssVEP power which has shown to be more sensitive to conditioning effects in the second half of CS presentation (Miskovic & Keil, 2013;Moratti & Keil, 2005;Moratti et al., 2006).The ssVEP power was averaged across O1, Oz, and O2 electrodes, where ssVEP amplitudes were maximal (see Fig. 3) and in line with previous studies (Aslanidou et al., 2023;Stegmann et al., 2020;Wieser et al., 2014).
The LPP was measured as the mean amplitude at the Pz electrode in the time window between 400 ms to 1000 ms after CS onset (based on Bacigalupo & Luck, 2018;Bauer et al., 2020;Klein et al., 2023).For the LPP analysis we applied a band-rejection filter to diminish the effect of the flickering images at 15 Hz (12.5 Hz for the participants that were exposed to that frequency) with 0.1 Hz bandwidth (4th order).Then, a baseline correction was applied using the 200 ms before CS onset.
Second, to calculate SCR, the EDA was filtered offline with a high cut-off filter of 1 Hz and a notch filter of 50 Hz.The signal was segmented in epochs of − 1000 ms to 9000 ms after CS onset.The signal was quantified using a manual Foot-to-Peak method by identifying the foot in a time-window of 800-4000 ms after stimulus onset and the Peak as the first following response peak (Boucsein et al., 2012).Responses below 0.02 μS were scored as zero. 2   For both EEG signals and the SCR, we calculated averages based on the phases and stimulus type, resulting in one for CS+ and one for CS-in the habituation phase.To investigate the temporality of fear acquisition and extinction, we divided these phases into early (first 10 trials per condition) and late (last 10 trials per condition) sections (Morriss, Christakou, et al., 2016).This results in four averages in acquisition (CS+ early, CS-early, CS+ late, CS-late) and four for extinction (CS+ early, CS-early, CS+ late, CS-late).

Data analysis
The analyses were conducted using IBM SPSS (Version 29.0, IBM, Armonk, NY).For all the analyses, we considered statistical significance at the alpha level of p = .05.We report the Cohen's d or partial eta squared as estimations of the effect sizes.
In habituation, we compared the average responses to CS+ and CSthrough paired t-tests (two-tailed).This was applied to both ratings and physiological responses.
In the acquisition and extinction phases, we compared the affective ratings of CS+ and CS-after each phase using paired t-tests (two-tailed).To analyze the US-expectancy ratings, we used an ANOVA with stimulus (CS+, CS-) and time (early [after 10 first trials] and late [after 40 trials, end of phase]) as within-subject factors.We ran ANOVAs with stimulus (CS+, CS-) and time (early, late) as within-subject factors for SCR, ssVEPs, and LPP, separately.Significant effects and interactions were followed up with post-hoc paired t-tests (Bonferroni corrected).Further, to investigate the effect of IU on these responses we used ANCOVAs with the same factors and including total IU as a covariate. 3 To assess the specificity of the influence of IU on conditioned fear responses during fear acquisition and fear extinction, we performed a bivariate correlational analysis between the IUS and its subscales with the outcome measures of conditioned fear responding (i.e., valence and arousal ratings, expectancy ratings in acquisition, SCR, ssVEPs, and LPPs) for each phase.To control for the influence of the other trait measures (STAI, BDI, and PSWQ), a partial correlational analysis was performed between the IUS and the respective outcome measures.These correlations were tested for acquisition and extinction phases and divided by time (i.e., early and late sections).For each outcome measure, we examined the association with IUS by looking at average responses to the CS+ only, the CS− only, and the difference score (i.e., CS discrimination), which was calculated by subtracting the CS− response from the CS+ response.The Holm-Bonferroni multiple testing correction method was used to control for the fact that several different outcome measures were included (Holm, 1979).To this end, the p-values from the tests were ordered from lowest to highest.We began by comparing the lowest p-value to a Holm-Bonferroni-adjusted alpha, which is the original alpha level divided by the total number of tests.The next smallest p-value was then compared with an adjusted alpha calculated for one fewer test than the previous step (Holm, 1979).Once a p-value exceeded the adjusted significance level, the procedure was stopped, and all remaining hypotheses were considered non-significant.As reported in Tables 2 and 4, the alpha level was divided by 4 (early acquisition), 6 (late acquisition), 3 (early extinction), and 5 (late extinction), based on the number of measures at each time.

Results
The dataset of the final sample is publicly available at https://osf.io/vupj5/.

Habituation
The analysis of the ratings and physiological responses to CS+ and CS-after and during the habituation phase showed no significant differences (all p values > .205).

Psychophysiology and EEG
For SCR, the main effect of stimulus type was significant, F(1,175) 2 Additionally, we applied a square root transformation and z-transformation to the SCR signal before rerunning the same analyses.Notably, the results closely mirror those obtained using the raw SCR scores; further details can be found in the Supplementary materials.
3 Following a suggestion from an anonymous reviewer, we explored the impact of IU scores on SCR by grouping participants, either by extreme high or low IU scores (M ± 1 SD) or by dividing the sample into quartiles.However, neither grouping method revealed an effect of IU on SCR.The full results and additional details are provided in the Supplementary materials.

Influence of intolerance of uncertainty
To analyze the effect of IU on the ratings we ran ANCOVAs.For valence, the main effect stimulus had a significant role still after accounting for IU, F(1,174) = 5.80, p = .017,η p 2 = .032.However, the interaction of Stimulus with IU revealed no influence of IU on the valence ratings after acquisition, F(1,174) = 0.10, p = .754,η p 2 = .001.
The results of the ANCOVAs for SCR revealed no significant effect of IU.The interactions of IU with the factors (stimulus, time, and Stimulus X Time) were all not significant (all p values > .122).The only factor that remains significant after accounting for IU is stimulus (F(1,174) = 6.40, p = .012,η p 2 = .035).For ssVEPs, also the results show no significant role of IU in acquisition.The interaction of IU with the factors as well as the single factors are not significant (all p values > .126).The same pattern was found for LPP, neither the interactions with IU nor the individual factors showed significant effects (all p values > .339).
We in addition ran bi-variate correlations and partial correlations (BDI, PSWQ, STAI-T) of IU and the subscales (I-IU and P-IU) separately with all dependent variables for CS+ and CS-and the differences scores (CS+ -CS-) for early and late acquisition.The results are presented in Tables 2a-2c.For CS+, there was an association found between total IU and arousal ratings after the acquisition phase (r = .19,p = .014).For CS-, total IU and inhibitory IU, when controlled for other traits, were correlated with expectancy ratings in late acquisition (r = − .18,p = .018,and r = − .15,p = .049,respectively).Additionally, during the same period, there was a link between arousal ratings and prospective IU (r = .17,p = .024).When exploring the difference score in this phase, there was an association between total IU (when controlled)  and expectancy ratings (r = .16,p = .034).However, none of the correlations remain significant after Holm-Bonferroni corrections for multiple testing.
CS-in early and late extinction.The effect of time was also significant, F (1,175) = 6.48, p = .012,η p 2 = .036,showing a decrease in the ssVEP from early to late extinction, whereas the interaction between Stimulus and Time was not significant (p = .843)(Fig. 3; Table 3).
For the LPP, the analysis revealed a significant effect of stimulus, F(1, 175) = 13.30,p < .001,η p 2 = .071,with a larger LPP for CS+ than for CS-.Additionally, there was a significant effect of time, F(1, 175) = 4.46, p = .036,η p 2 = .025,reflecting a decrease in the amplitude of this signal from early to late extinction.However, the interaction between Stimulus and Time was not significant (p = .274,see Fig. 4 C and D; Table 3).

Influence of intolerance of uncertainty
To analyze the effect of IU we used ANCOVAs.There was no influence of IU neither on valence (F(1,174) = 0.02, p = .899,η p 2 < .001)nor on arousal ratings after the extinction phase, F(1,174) = 0.31, p = .581,η p 2 = .002.None of the effects and their interactions (all ps > .225)or the interactions with IU scores (all ps > .556)were significant for SCR.The same was found for ssVEPs, with all the main effects ps > .132and their interactions with IU ps > .106.Also, for the LPP, no main factor (all ps > .070)nor interactions with IU (all ps > .191)were significant.
Overall, this analysis shows that IU did not influence the physiological responses to CS+ and CS-during extinction.
Like for acquisition, we also ran bi-variate correlations of total IU and subscales I-IU and P-IU with all dependent variables separately for early and late extinction.The results are presented in Tables 4a-4c.Regarding CS+, there was an association found between prospective IU (when controlled) and ssVEP during late extinction (r = − .17,p = .030).For CS-, during early extinction there was an association found between prospective IU (both controlled and alone) with the ssVEP (r = − .16,p = .035,and r = − .15,p = .049,respectively.In late extinction, valence ratings were also associated with prospective IU (r = .15,p = .050).No associations were found for the difference score during both sections of extinction.Resembling the results from acquisition, no significant correlations survived the Holm-Bonferroni corrections.

Discussion
This study aimed to investigate the role of intolerance of uncertainty (IU) during both fear and safety learning in a differential conditioning paradigm.In contrast to our expectations, we could not replicate earlier  findings and we did not find an association between IU and conditioned defensive responses during both fear acquisition and extinction.Overall, we observed the anticipated indicators of fear acquisition, characterized by heightened defensive responses to the CS+ compared to the CS-.Extinction learning was not evident in the affective ratings as verbal fear responses towards the CS+ remained elevated compared to the CS-.This pattern corroborates with the results of the LPP and ssVEP signals during both the early and late phases of extinction.In contrast, the SCR signal gradually extinguished meaning that discriminative responses were visible during the early trials but not during later trials.

Table 2a
Correlation coefficients and p-values reported for the relation between intolerance of uncertainty and the threat cue (CS+) during early and late acquisition.Note.P-values reported are uncorrected.None of the p-values are significant after Bonferroni-Holm corrections (p = .013for early acquisition and p = .008for late acquisition).IU = Total IU; P-IU = Prospective IU; I-IU = Inhibitory IU; r = Pearson correlations; Partial r = Correlations were controlled for STAI-T, BDI-II, and PSWQ using partial correlations.a Valence and arousal ratings were only obtained one time after the acquisition phase.
For fear acquisition, our findings align with the majority of studies that have not observed an effect of IU on fear learning (Johnson et al., 2023;Mertens & Morriss, 2021;Morriss, 2019;Morriss, Christakou, et al., 2016;Morriss et al., 2015Morriss et al., , 2020;;Morriss, Saldarini & Chapman, Pollard, et al., 2019;Morriss, Saldarini, & van Reekum, 2019;Morriss & van Reekum, 2019;Sjouwerman et al., 2020;Wake et al., 2020Wake et al., , 2021;;Wroblewski et al., 2022).However, some studies manipulating the reinforcement rate between the CS+ and US to increase uncertainty during learning have shown contradictory results.For example related with high IU levels, a 37.5 % reinforcement rate was associated with worse threat-safety discrimination (Kanen et al., 2021), while a 50 % rate showed enhanced discrimination compared to 75 % (Chin et al., 2016).In contrast, Mertens et al. (2022) using a 75 % reinforcement rate and Klingelhöfer-Jens et al. ( 2022) using 100 % reinforcement rate demonstrated more pronounced discrimination in the fear ratings for individuals with higher IU.In our study, we did not collect fear ratings, which might be the reason why we did not observe effects for the IU even when using a 75 % contingency rate.The US in our study was

Table 2b
Correlation coefficients and p-values reported for the relation between intolerance of uncertainty and the safety cue (CS-) during early and late acquisition.Note.P-values reported are uncorrected.None of the p-values are significant after Bonferroni-Holm corrections (p = .013for early acquisition and p = .008for late acquisition.IU = Total IU; P-IU = Prospective IU; I-IU = Inhibitory IU; r = Pearson correlations; Partial r = Correlations were controlled for STAI-T, BDI-II, and PSWQ using partial correlations .a Valence and arousal ratings were only obtained one time after the acquisition phase.probably strongly aversive because of its loudness (90 dB) and its nature (white noise mixed with a female scream).It is therefore conceivable that these experimental settings created a so-called 'strong situation' (Lissek et al., 2006), which in turn may have overridden most of the inter-individual differences in acquisition phase.
In fear extinction, contrary to our hypothesis, we found no significant effect of IU, which aligns with some (e.g., Morriss et al., 2020Morriss et al., , 2024;;Wendt & Morriss, 2022) but not with other studies (Morriss, 2019;Morriss et al., 2015;Morriss, Christakou, et al., 2016;Morriss, Macdonald, et al., 2016;Morriss, Saldarini, & van Reekum, 2019;Morriss & van Reekum, 2019;Wake et al., 2020Wake et al., , 2021;;Wroblewski et al., 2022).Our findings also contrast Morriss, Zuj, et al.' (2021) and Morriss, Wake, et al.' (2021) meta-analysis, which analyzed 18 experiments and found that IU and both subfactors influenced SCR threat-safety discrimination in extinction.Interestingly, our results are in line with recent findings from clinical patients where no significant effects of self-reported IU were observed for differential SCR during acquisition and extinction (Morriss et al., 2024).Our findings are well in line with the only higher-powered study as of now.Morriss et al. (2020) using a well     powered sample size (N = 144) investigated the effect of extended extinction, both directly after acquisition (i.e., immediate extinction) and next day (i.e., delayed extinction).They did not find an effect of the total IU score on both types of extinction.However, higher I-IU was associated with poorer SCR threat-safety discrimination (i.e.better extinction) in delayed extinction compared to immediate extinction.Considering that our and another well-powered study (Morriss et al., 2020) reported zero-findings, one might cautiously conclude that the influence of IU on extinction learning is not so robust as previously thought.However, this may also be dependent on the type of quantification of the individual differences of interest, as a recent multiverse meta-analysis showed (Morriss et al., 2024).This underscores the importance of well-powered replications to establish the conceptual basis of IU's role in fear extinction.
Our zero-findings may also be explained partly by the relative low uncertainty in our fear conditioning paradigm.With an initial reinforcement rate of 75 % and an immediate extinction the situation is likely not uncertain or ambiguous enough.Indeed, an influence of IU is more often observed in situations where uncertainty is maximal.For example, Chen and Lovibond (2016) found that for high IU individuals, ambiguity, rather than uncertainty per se, leads to biased threat appraisal and negative affect.Other studies employing NPU threat paradigms and affective cuing also showed that high IU seems especially prone to different processing of these ambiguous cues, where uncertainty is maximized such that, the chance is 50/50 that an aversive stimulus can follow (Del Popolo Cristaldi et al., 2022;Gole et al., 2012;Nelson & Shankman, 2011).
Our results also point to a more general issue: The assessment of interindividual differences in psychophysiology faces some challenges, particularly in bridging different types of responses (e.g., questionnairebased trait variables and neurobiological variables) (Bernoster et al., 2019;Brandt & Mueller, 2022).Personality traits like IU can represent higher abstracted levels, of variables that represent experience and behavior, thus posing challenges in their association with specific neurobiological variables such as LPP or ssVEP (Brandt & Mueller, 2022).Similarly, associations between self-report measures and EEG tend to show correlations within the same type of variable rather than between types (Bernoster et al., 2019).Within this broader context, there is also a lack of consensus on how to measure and interpret IU.While some researchers advocate for the use of total IU scores (Bottesi et al., 2019;Hale et al., 2016;Lauriola et al., 2016;Morriss, Wake, et al., 2021;Shihata et al., 2018;Yao et al., 2021), others have found differential associations between IU subfactors and measured variables (e.g., Boelen et al., 2016;Jackson et al., 2016;Johnson et al., 2023;McEvoy & Mahoney, 2011;Penney et al., 2020;Saulnier et al., 2019).Further, the two IU subfactors may underline opposite mechanisms, which is challenging to merge into one construct such as the total IU.The prospective component (P-IU) is in fact thought to reflect cognitive engagement, while the inhibitory component (I-IU) represents the inhibition of action in the face of uncertainty (Birrell et al., 2011;McEvoy & Mahoney, 2011).A better clarification of IU measurement practices is essential to establish a robust conceptual and psychometric framework.
Altogether, fear conditioning was observed as predicted and we successfully induced fear acquisition.This was evidenced by significant changes in ratings (arousal, valence, and US-expectancy) and the physiological responses (SCR and ssVEP), albeit less clear for the LPP.This signal showed an enhanced amplitude for CS-compared to CS+ in early acquisition, which is contrary to our expectations and previous research (Bacigalupo & Luck, 2018).However, this effect was reversed and in line with aforementioned studies in the second half of the acquisition phase in our experiment.Other studies have also found no discrimination during acquisition (Bauer et al., 2020;Kastner et al., 2015), which distinguishes this signal from other measures of fear learning (as SCR or FPS) mainly by showing sometimes a delayed discrimination.Future research should consider these results when further validating the LPP as a reliable readout measure for both fear acquisition and extinction.
We found less consistent extinction of the conditioned responses.The arousal and valence ratings as well as the ssVEPs and LPPs still showed a difference between CS+ and CS-, while in the SCR successful extinction was observed.Previous studies have shown that ratings may be more resistant to extinction than physiological responses (Díaz et al., 2005;Hermans et al., 2002;Luck & Lipp, 2015).The resistance to extinction observed in the ssVEP and LPP might be linked to top-down attentional processes (Antov et al., 2020), which might also be involved in verbal reports, in contrast with measures that closely reflect associative learning, such as the SCR (Constantinou et al., 2021), meaning that this signal reflected more accurately the change in CS-US contingencies characteristic of the extinction phase.The discrepancy between these measures (i.e., affective ratings, LPP and ssVEP vs. SCR) may indicate that these measures index different aspects of fear as proposed by the dual-process model of fear learning, where the subjective experience of Note.P-values reported are uncorrected.None of the p-values are significant after Bonferroni-Holm corrections (p = .017for early extinction and p = .010for late extinction).IU = Total IU; P-IU = Prospective IU; I-IU = Inhibitory IU; r = Pearson correlations; Partial r = Correlations were controlled for STAI-T, BDI-II, and PSWQ using partial correlations.a Valence and arousal ratings were only obtained one time after the extinction phase.
fear can be separated from other physiological and behavioral responses (LeDoux & Brown, 2017;LeDoux & Pine, 2016;Öhman & Mineka, 2001), but can also be due to differences in habituation rates, measurements errors, or the noisiness of physiological measures (Sjouwerman et al., 2020).Some limitations of our study should be considered.First, the data collection was carried out during the COVID-19 pandemic, this context could have influenced the reactions to threat and safety as well as beliefs around uncertainty (e.g., Korte et al., 2022;Segovia et al., 2024;Tanir et al., 2020).Second, the lack of fear ratings in our design limited the direct comparison of findings with other studies.Although our affective ratings are an indicator of evaluative learning, future research can benefit from also using fear ratings as they broaden the scope of the self-report.Third, the results of the LPP analysis must be interpreted cautiously, since the design of our experiment with flickering stimuli, relatively few trials, and the lack of mastoid references are suboptimal for LPP scoring and therefore these results cannot easily be compared to other findings (Bacigalupo & Luck, 2018;Bauer et al., 2020;Bruchmann et al., 2021;Klein et al., 2023;Nelson et al., 2015).Fourth, our sample mostly consist of female participants (81 %), who have previously been shown to demonstrate slower extinction, seemingly caused by the hormonal fluctuations in their menstrual cycle (for a review see Merz et al., 2018).Even though women more often suffer from anxiety disorders, a more gender-balanced sample would allow for stronger conclusions in the general population.
In conclusion, our study utilized a fear acquisition and extinction paradigm with a large sample size to investigate the role of IU in conditioned fear responses.Our findings suggest that IU does not significantly impact fear acquisition and extinction.Altogether, these findings shed further light on a delimited role of this trait in fear acquisition and extinction in differential conditioning paradigms.Future studies with adequately powered samples are warranted to replicate our findings, alongside investigations into IU's influence on related paradigms such as fear generalization, renewal, and avoidance.

Declaration of Generative AI and AI-assisted technologies in the writing process
During the preparation of this manuscript the authors did not use generative AI technologies.

Fig. 1 .
Fig. 1.US-Expectancy ratings for CS+ (red dots) and CS-(blue dots) during early (after 10 trials) and end (after 40 trials) of acquisition.The dots depict individual participants, while the black horizontal lines indicate the group means.

Fig. 2 .
Fig. 2. SCR responses to CS+ (red dots) and CS-(blue dots) during early and late acquisition (A) and during extinction (B) phases.The dots depict individual participants, while the black horizontal lines indicate the group means.Asterisks represent post-hoc significant differences at p < .025threshold.

Fig. 3 .
Fig.3.Grand-averaged topographic maps of the ssVEPs (steady-state visually evoked potentials), depicting the power at 15 Hz.Asterisks represent post-hoc significant differences at p < .025threshold, only calculated for acquisition where the interaction Time x Stimulus was significant.

Fig. 4 .
Fig. 4. Grand-averaged LPP waveforms at Pz electrode, time-locked to the onset of the CS.The asterisks show significant post-hoc differences between stimuli responses (CS + vs. CS -) at p < .025,which were only carried out for acquisition since the interaction of CS type and Time was significant.The grey rectangle shows the time-window (400 -1000 ms) used for calculating the LPP as mean amplitude.
Note.P-values reported are uncorrected.None of the p-values are significant after Bonferroni-Holm corrections (p = .013for early acquisition and p = .008for late acquisition.IU = Total IU; P-IU = Prospective IU; I-IU = Inhibitory IU; r = Pearson correlations; Partial r = Correlations were controlled for STAI-T, BDI-II, and PSWQ using partial correlations.aValence and arousal ratings were only obtained one time after the acquisition phase. Note.P-values reported are uncorrected.None of the p-values are significant after Bonferroni-Holm corrections (p = .017for early extinction and p = .010for late extinction).IU = Total IU; P-IU = Prospective IU; I-IU = Inhibitory IU; r = Pearson correlations; Partial r = Correlations were controlled for STAI-T, BDI-II, and PSWQ using partial correlations.aValence and arousal ratings were only obtained one time after the extinction phase.
017 for early extinction and p = .010for late extinction).IU = Total IU; P-IU = Prospective IU; I-IU = Inhibitory IU; r = Pearson correlations; Partial r = Correlations were controlled for STAI-T, BDI-II, and PSWQ using partial correlations.aValence and arousal ratings were only obtained one time after the extinction phase.

Table 1
Mean and standard deviations of LPP and ssVEP amplitudes for CSs during Acquisition.

Table 2c
Correlation coefficients and p-values reported for the relation between intolerance of uncertainty and the difference score (CS+ minus CS-) during early and late acquisition.

Table 3
Mean and standard deviations of LPP and ssVEP for CSs during early and late extinction.

Table 4a
Correlation coefficients and p-values reported for the relation between intolerance of uncertainty and the threat cue (CS+) during early and late extinction.

Table 4b
Correlation coefficients and p-values reported for the relation between intolerance of uncertainty and the threat cue (CS-) during early and late extinction.
Note.P-values reported are uncorrected.None of the p-values are significant after Bonferroni-Holm corrections (p = .

Table 4c
Correlation coefficients and p-values reported for the relation between intolerance of uncertainty and the difference score (CS+ minus CS-) during early and late extinction.