Compulsive avoidance in youths and adults with OCD: an aversive Pavlovian-to-instrumental transfer study

The study investigates how conditioned (Pavlovian) cues influence instrumental avoidance behavior in obsessive-compulsive disorder (OCD) using an aversive Pavlovian-to-Instrumental Transfer (PIT) paradigm. PIT comprises specific transfer (stimulus selectively biases the congruent instrumental response linked to the same outcome) and general transfer (stimulus with aversive properties generally increases avoidance responding). Prior PIT work in clinical OCD primarily used appetitive paradigms and suggested reduced specific PIT with preserved general PIT, aligning with reported goal-directed control dysfunction in OCD. Because compulsions in OCD often function to prevent harm, the authors hypothesized that an aversive avoidance-based PIT might either reproduce weakened specific PIT with intact general PIT in OCD or alternatively reveal enhanced PIT due to heightened avoidance tendencies. They also explored whether learning performance, metacognitive confidence, and avoidance urges modulate PIT strength differently in OCD versus controls. By including both youths and adults, the study aimed to probe developmental aspects of compulsive behavior, given OCD often onsets in childhood/adolescence and emerging evidence points to age-related differences in learning and cognition.

PIT has been used to demonstrate how goal-directed actions can become more habitual under Pavlovian influence, with specific transfer linked to model-based control and higher-order cognition, and general transfer reflecting model-free motivational properties. In OCD, PIT research is sparse but recent appetitive studies report reduced specific PIT and intact general PIT, consistent with impaired goal-directed control. Aversive contexts are especially relevant to OCD due to harm-avoidance compulsions and documented deficits in safety signaling and persistence of avoidance despite devaluation. Cognitive research in OCD highlights atypical associative learning and a dissociation between action and confidence (metacognition), whereby knowledge may not appropriately guide behavior. These literatures motivate testing aversive PIT in clinical OCD and examining modulatory roles of learning, confidence, and motivational factors across development.

Design: Cross-sectional case-control study with two age cohorts (adults and youths). Participants: 41 OCD (21 adults; 20 youths aged 12–19) and 44 healthy controls (21 adults; 23 youths). Inclusion for OCD: DSM-5-TR diagnosis with OCD as primary disorder and CY-BOCS/Y-BOCS > 12; other major Axis I disorders excluded; control participants free of neurological/psychiatric history. Medication: 18/21 adult OCD on medication (10 SSRI only; 5 SSRI+other; 3 non-SSRI); 10/20 youth OCD on SSRIs. Ethics: Approved by relevant UK REC committees; informed consent obtained (parental consent <16). Clinical assessments: OCI-R for all; symptom severity by Y-BOCS (adults) and CY-BOCS (youths); depression/anxiety by MADRS and STAI for adults, BDI-Y and BAI-Y for youths; IQ estimated by NART (adults) and WASI-II FSIQ-2 (youths); working memory via WASI-II Digit Span Backwards. Task: Aversive PIT adapted from Garofalo & Robbins (2017) with three phases. Phase 1 Instrumental Conditioning: Learn response-outcome contingencies to cancel two aversive outcomes (O1, O2—unpleasant noises paired with bombs/missiles images) by moving joystick left/right. Each trial: “Defend yourself” (2 s), upcoming outcome image (1 s), 30 s response window with attacks every 1.5–3 s; correct direction cancels outcomes (80% contingency, undisclosed). Eight trials (4 O1, 4 O2), ~8 min. Post-phase: explicit outcome-response pairing test, confidence ratings (1–9), and urge-to-avoid ratings (1–9). Training with 4 blocks without noises preceded testing. Phase 2 Pavlovian Conditioning: Learn stimulus-outcome associations for 4 galaxy images: S+1→O1, S+2→O2, S+3→new O3 (dynamite), S−→safe (no noise). 68 trials (17 per stimulus), each: stimulus (4.5 s) then outcome/no-outcome display (1 s); ITI 7–9 s; S+ partial reinforcement at 60–40; pairings counterbalanced. Pre/post ratings: subjective liking (1–9) for scenes; post-phase explicit stimulus-outcome mapping and confidence (1–9). Phase 3 PIT (extinction): Stimuli from Phase 2 presented for 30 s with ITI 1–3 s; joystick re-enabled but no outcomes delivered. Six blocks per stimulus (24 trials total), ~10 min. Measures: Two PIT indices per construct: proportion of responses and mean maximum grip force (kg) via isometric hand dynamometer integrated with joystick. Specific PIT: more congruent vs incongruent responses/force. General PIT: more responses/force to S+3 vs S−. Statistical analysis: Conducted in R (4.0.4) and Jamovi (1.6.23). Frequentist tests with alpha 0.05 and Bayes Factors (BF10) to quantify evidence for alternative hypotheses; multiple-comparisons correction applied to frequentist correlations. Bayesian linear regressions examined modulators (learning indices, confidence, avoidance urges, subjective evaluation changes). Data/code: OSF https://osf.io/pxfhz/.

• Clinical measures: OCD showed higher OCI-R than controls (TWJ ≈ 97.42, p < 0.001). Adults with OCD had elevated trait (t36 = 7.52, p < 0.001) and state anxiety (t36 = 4.21, p < 0.001) and depression (t ≈ 4.70, p < 0.001); youth OCD also had higher anxiety (t41 = 5.91, p < 0.001) and depression (t41 = 7.91, p < 0.001). - Learning: Groups (OCD vs controls) learned instrumental (Phase 1) and Pavlovian (Phase 2) contingencies equivalently. - Overall PIT: No significant group differences across age in specific PIT (proportion p = 0.32, BF10 = 0.36; grip force p = 0.57, BF10 = 0.25) or general PIT (proportion p = 0.957, BF10 = 0.23; grip force p = 0.18, BF10 = 0.33). - Modulators of PIT (Bayesian regressions): • Specific PIT (proportion) was less positively associated with confidence in outcome-response matching in OCD vs controls (Group × Confidence Estimate ≈ −2.68, 95% CI −3.90 to −0.79, BF10 = 6.05, p = 0.03). • In OCD, higher urge to avoid aversive noises predicted stronger specific PIT (grip force) relative to controls (Group × Urge-to-Avoid Estimate ≈ 1.56, 95% CI 0.38–2.75, BF10 = 1.51, p = 0.011). • General PIT (proportion) in OCD increased with a greater positive change in preference for S− over S+3 from pre- to post-Pavlovian phase (Group × Change in Subjective Evaluation Estimate ≈ 2.56, 95% CI 0.49–4.61, BF10 = 1.86, p = 0.017); this effect was also stronger in youths vs adults (Age × Change Estimate ≈ 2.61, 95% CI 0.25–4.98, BF10 = 1.25, p = 0.031). • Confidence in stimulus-outcome mappings more strongly increased general PIT (grip force) in controls vs OCD (Group × Confidence Estimate ≈ −2.23, 95% CI −4.30 to −0.15, BF10 = 1.01, p = 0.025). • Adult-only analyses suggested Group × Urge-to-Avoid (Estimate ≈ 1.21, 95% CI 0.06–2.37, BF10 = 2.39, p = 0.04) and Group × Change in Subjective Evaluation (Estimate ≈ 2.37, 95% CI 1.01–3.73, BF10 ≈ 35.13, p = 0.001) effects drove these patterns; youth-only interactions were not significant. - Age-by-group and age effects: • A Group × Age interaction for specific PIT (proportion) showed differences in youths (BF10 ≈ 27.94; p = 0.022) but not adults. Youths with OCD made more incongruent than congruent responses (t19 = −2.63, p = 0.016); control youths showed no congruent vs incongruent difference. • Adults showed stronger specific and general PIT than youths overall: Specific PIT (proportion) Age × Congruence p < 0.001, η² ≈ 0.316, BF10 ≈ 2.93e+11; Specific PIT (grip force) p = 0.020; General PIT (proportion) p = 0.036, BF10 ≈ 10.41. Both ages responded more to S+3 than S−, but adults responded more to S+3 and less to S− than youths. General PIT grip force was not significantly age-modulated (p = 0.077). • Adults exerted greater grip force overall than youths in PIT trials (specific: F1,81 = 11.18, p = 0.001, BF10 ≈ 107.92; general: F1,81 = 8.55, p = 0.004, BF10 ≈ 57.47). - Correlations: Across all participants, specific PIT strength (proportion) correlated positively with age (R = 0.42, BH p < 0.001). In youths with OCD, specific PIT strength (grip force) correlated negatively with CY-BOCS total (R = −0.66, BH p = 0.015), obsessions (R = −0.62, BH p = 0.039), and compulsions (R = −0.63, BH p = 0.028).

Findings indicate that, overall, clinical OCD does not show impaired aversive PIT compared to controls. However, PIT in OCD appears to be driven more by implicit motivational factors (e.g., urge to avoid aversive outcomes and increased preference for safety cues) than by explicit metacognitive confidence in learned associations, which more strongly guided PIT in controls. This supports models of OCD emphasizing harm avoidance, model-free biases, and an action–confidence dissociation where knowledge does not appropriately inform behavior. Developmentally, youths with OCD exhibited weaker specific PIT—favoring incongruent responses—and the severity of OCD symptoms was associated with reduced specific PIT vigor. This suggests adolescent impairments in integrating Pavlovian and instrumental information (model-based control) despite intact basic conditioning, aligning with prior reports of reduced goal-directed control and atypical OFC function. Across diagnoses, PIT strength was lower in youths than adults, consistent with the maturation of model-based control during adolescence and the development of cortical systems important for threat discrimination and flexible control. Together, the results argue that compulsive avoidance behaviors in OCD may be fueled by automatic, implicit motivational processes rather than explicit learning, and that adolescent-specific deficits in specific PIT could serve as early behavioral markers of the disorder.

This study shows that while overall aversive PIT performance is comparable between adults with OCD and controls, the mechanisms underlying PIT differ: OCD participants rely more on implicit motivational factors, whereas controls’ PIT is better predicted by metacognitive confidence in learned associations. Youths with OCD display impaired specific PIT and symptom-linked reductions in response vigor, highlighting developmental vulnerability in cue integration and goal-directed control. Clinically, targeting implicit motivational processes (e.g., harm-avoidance drives and safety cue valuation) may enhance treatments for OCD. Future work should standardize PIT paradigms, extend to longitudinal designs to chart developmental trajectories, examine medication and symptom-dimension effects, and integrate neuroimaging to identify neural substrates of PIT alterations in OCD across development.