Intense Fear Resists Immediate Extinction: Two Experiments on Intervention Timing

Exposure works, when it works, by building a new prediction that competes with the old one at the moment of retrieval. Almost everything we know about that process comes from laboratory extinction using mild aversive stimuli in people who are not patients – and the field has quietly assumed those findings scale up to clinical fear. A two-experiment study from South China Normal University tests that assumption directly, and finds the standard intuition about timing running backwards.

Where immediate extinction breaks down

Feng and colleagues used vicarious extinction, the observational analogue of exposure: rather than facing the conditioned cue themselves, participants watch a model meet it without consequence. Experiment 1 (N = 54) crossed this with fear intensity, high versus moderate, and delivered the intervention immediately after acquisition. The moderate-intensity condition behaved as the existing literature predicts – fear dropped, and it stayed down when the aversive outcome was later re-presented to trigger reinstatement. The high-intensity condition did not. Those participants showed significantly higher overall skin conductance responses at reinstatement testing and rated the aversive outcome as more expected in retrospect. The same procedure, applied to a stronger fear, largely stopped working.

Experiment 2 (N = 56) asked whether timing could recover the loss, comparing a 1-day delay and a 14-day delay against the immediate baseline from Experiment 1. One day was not enough: that group remained suboptimal. Fourteen days was. The 14-day group showed lower overall reinstatement than the immediate group, and both weaker CS+ reactivity and weaker differential responding (CS+ minus CS-) than the 1-day group – landing at outcomes comparable to the moderate-intensity group from the first experiment. Intensity and delay interacted; neither factor told the story alone.

This pattern is what the animal and human literature calls the immediate extinction deficit: extinction training run too soon after acquisition yields a weaker, less durable inhibitory memory. What this study adds is that the deficit survives the shift to observational learning, and that it scales with fear intensity – which is precisely the direction that makes it clinically inconvenient.

What this does and does not license

The tempting misreading is "wait two weeks before treating trauma." Nothing here supports that. A cue-outcome association consolidated over 14 days in a laboratory is not a traumatic memory, reinstatement is not relapse, and these were non-clinical volunteers, not patients. Read literally, the study is about conditioning, not about care.

Read carefully, it converges with something our field already paid for in failed trials. Single-session psychological debriefing delivered immediately after trauma does not prevent PTSD and can worsen outcomes, which is why NICE recommends watchful waiting instead. The immediate extinction deficit is a plausible mechanism for why "get in early" underperformed – and this study shows the same shape in a controlled setting, at high fear intensity, in humans.

The second implication is closer to the consulting room. Vicarious and modelling components – therapist demonstration, video modelling, watching a group member go first – are cheap and are often treated as interchangeable with participation. This result suggests they weaken exactly where the fear is strongest. For an intensely feared stimulus, watching is likely not enough; the patient has to do it.