From adaptive to maladaptive fear: Heterogeneity in threat and safety learning across response systems in a representative sample

Abstract

Individual differences in fear learning are a crucial prerequisite for the translational value of the fearconditioning model. In a representative sample (N = 936), we used latent class growth models to detect individual differences in associative fear learning. For a series of subsequent test phases varying in ambiguity (i.e., acquisition, extinction, generalization, reinstatement, and re-extinction), conditioned responding was assessed on three response domains (i.e., subjective distress, startle responding, and skin conductance). We also associated fear learning across the different test phases and response domains with selected personality traits related to risk and resilience for anxiety, namely Harm Avoidance, Stress Reaction, and Wellbeing (MPQ; Tellegen and Waller, 2008). Heterogeneity in fear learning was evident, with fit indices suggesting subgroups for each outcome measure. Identified subgroups showed adaptive, maladaptive, or limited-responding patterns. For subjective distress, fear and safety learning was more maladaptive in the subgroups high on Harm Avoidance, while more adaptive learning was observed in subgroups with medium Harm Avoidance and the limited-or non-responders were lowest in Harm Avoidance. Distress subgroups did not differ in Stress Reaction or Wellbeing. Startle and SCR subgroups did not differ on selected personality traits. The heterogeneity in fear-learning patterns resembled risk and resilient anxiety development observed in real life, which supports the associative fear-learning paradigm as a useful translational model for pathological fear development.

Increases and decreases in fear potentiated startle during fear acquisition: A latent class growth analysis.

“…Though Gazendam et al (2020) did not report any subgroups with an overall decrease in differential responding during acquisition, analysis of mean responding during each trial revealed that danger cue responding was beginning to decrease by the end of acquisition for three of five subgroups. Of note, the fear learning paradigm used by Gazendam et al (2020) included a total of six danger cue trials during acquisition. A longer acquisition period may have led to an overall decrease in differential fear responding during acquisition in some participants.…”

“…However, mean comparisons of LCGA-derived extinction subgroups revealed substantial differences in the rate of FPS increase during acquisition, leading to a more than fivefold difference in FPS between Modal FPS Responders and the two other classes at the beginning of extinction (Galatzer-Levy et al, 2017). Though Gazendam et al (2020) did not report any subgroups with an overall decrease in differential responding during acquisition, analysis of mean responding during each trial revealed that danger cue responding was beginning to decrease by the end of acquisition for three of five subgroups. Of note, the fear learning paradigm used by Gazendam et al (2020) included a total of six danger cue trials during acquisition.…”

“…Based on evidence showing that partial US reinforcement is less predictable than complete US reinforcement (e.g., Douglas & Parry, 1994;Lissek et al, 2006), we paired the US with the danger cue on 75% of danger cue trials during the first two blocks of acquisition. Thus, our reinforcement rate was lower than previous LCGA studies, which used 100% reinforcement throughout acquisition (Galatzer-Levy et al, 2017) or 100% reinforcement after the first trial (Gazendam et al, 2020). Additionally, we did not include contingency instructions or contingency ratings, the absence of which is theorized to decrease predictability and increase interindividual differences in fear learning trajectories during acquisition Mertens et al, 2021;Sjouwerman et al, 2016;Weidemann et al, 2016).…”

“…Though FPS was not calculated, LCGA analysis was performed on reflexive startle during danger cue presentations (CS+), reflexive startle during safety presentations (CS−), and reflexive startle during intertrial intervals (NA; Gazendam et al, 2020). Subgroups differed with respect to the degree of change in startle during acquisition, but differential responding was reported in all groups and decreases in differential responding during acquisition were not reported for any group (Gazendam et al, 2020). While heterogeneity may be present during fear acquisition, no extant studies of which we are aware have identified subgroups which differ with respect to the direction of change in FPS during fear acquisition using LCGA.…”

“…Additionally, we did not include contingency instructions or contingency ratings, the absence of which is theorized to decrease predictability and increase interindividual differences in fear learning trajectories during acquisition Mertens et al, 2021;Sjouwerman et al, 2016;Weidemann et al, 2016). By contrast, previous LCGA studies included either partial contingency instructions (Gazendam et al, 2020) or contingency ratings (Galatzer-Levy et al, 2017). To reduce the potency of the US, we used a relatively mild US that has been shown to be less aversive than shock (Gazendam et al, 2020;Glenn et al, 2012).…”