Test-retest reliability of human threat conditioning and generalization across a 1-to-2-week interval

Given the increasing use of threat conditioning and generalization for clinical-translational research efforts, establishing test-retest reliability of these paradigms is necessary. Specifically, it is an empirical question whether the same participant evinces a similar generalization gradient of conditioned responses across two sessions with the identical contingencies and stimuli. Here, 46 human volunteers participated in an identical auditory threat acquisition and generalization protocol at two sessions separated by 1-to-2 weeks. Skin conductance responses (SCR) and trial-by-trial shock risk ratings served as primary measures. We used linear mixed effects modeling to test differential threat responses and generalization gradients, and Generalizability (G) theory coefficients as our primary formal assessment of test-retest reliability of intraindividual stability and change across time. Results showed largely invariant differential conditioning and generalization gradients across time. G coefficients indicated fair reliability for acquisition and generalization SCR. In contrast, risk rating reliabilities were mixed, and reliability was particularly low for acquisition risk ratings. Our findings generally support reliability of the threat conditioning and generalization paradigm for shorter test-retest intervals and highlight their utility for assessments of behavioral interventions in mental health research, but challenges remain and further work is needed. Threat conditioning and generalization tasks are increasingly used for translational efforts to improve behavioral interventions, and thus test-retest reliability for these tasks needs to be established. Our results support the test-retest reliability of threat conditioning and generalization over a relatively short (1-to-2 week) interval, but this depends on the measure used (physiological vs. self-report). Overall, these tasks could be appropriate for repeated testing over the course of a short-duration intervention study, but more research is needed, particularly in regard to longer-duration studies.

Automated summary

This study investigates the test-retest reliability of threat conditioning and generalization tasks, and finds that they have good reliability over a relatively short time interval, making them useful for evaluating behavioral interventions. However, further research is needed.