Prior conceptual structure in shaping fear generalization: The impact of co-occurrence frequency and retrieval...

Conceptual fear generalization refers to the transfer of acquired fear to stimuli that are conceptually related to a conditioned threat, beyond perceptual similarity. Although previous studies have demonstrated that fear can generalize along semantic or associative relations, it remains unclear how the internal structure of conceptual knowledge shapes this process and under what conditions such structure is expressed. Across two experiments, we examined conceptual fear generalization within experimentally constructed associative structures. Prior to fear conditioning, participants learned artificial categories composed of meaningless symbols, within which associative strength was systematically manipulated via co-occurrence frequency (COF). Fear generalization was then assessed to category members that were never directly paired with the unconditioned stimulus, under varying levels of COF (Experiments 1 and 2) and retrieval context congruency (Experiment 2). Results revealed that fear generalization, indexed primarily by US expectancy, varied as a function of COF, with stronger prior co-occurrence with the conditioned threat predicting higher threat expectancy for threat-related generalization stimuli. Critically, these COF-based effects were amplified when the retrieval context matched the context in which the associative structure was originally learned. Skin conductance responses showed broader sensitivity to threat and context but provided less specific support for structure-dependent effects. Together, these findings demonstrate that conceptual fear generalization is shaped by graded associative structure and is contingent on retrieval conditions. The present study highlights the importance of considering both conceptual organization and retrieval context in understanding variability in fear generalization, with implications for models of learned fear and individual differences in threat processing.