Does How You Do Exposure Therapy Matter? UCLA's First Head-to-Head Trial of Inhibitory Learning vs. Habituation for Social Anxiety (2025)

SocialAnxiety.co Research Summary | Based on research from the Anxiety and Depression Research Center (ADRC), UCLA

Research Attribution: This article summarizes findings from Treanor, M., Zbozinek, T.D., Rosenberg, B.M., Sewart, A., Sandman, C.F., Ruiz, J., & Craske, M.G. (2025), “A randomized controlled trial comparing two processes of exposure therapy: Extinction learning and habituation,” published in the Journal of Consulting and Clinical Psychology, 93(10), 705–717. The research was conducted at the UCLA Department of Psychology. SocialAnxiety.co is an independent psychoeducation platform and is not affiliated with the University of California, Los Angeles.

Why This Research Matters

Exposure therapy works for social anxiety. That much is established. But why it works — the underlying mechanism that produces lasting change in the brain — has been debated for decades.

Two competing theories have shaped clinical practice:

The habituation model says exposure works because fear naturally decreases when you stay in a feared situation long enough. The anxiety rises, peaks, and falls. Repeated cycles of this rise-and-fall pattern gradually weaken the fear response. Under this model, the therapist’s job is to keep the patient in the situation until their fear drops — and to track that drop as evidence of progress.

The inhibitory learning model, developed by Prof. Michelle Craske and colleagues at UCLA, says something fundamentally different. It says exposure works not because the original fear is erased, but because the brain forms a new, competing memory — “I did this feared thing and the catastrophe didn’t happen.” The old fear memory still exists. But the new safety memory, if strong enough, overrides it. Under this model, the therapist’s job is to maximize expectancy violation — the gap between what the patient predicted would happen and what actually happened.

These two theories lead to different clinical practices. Different instructions during sessions. Different measures of progress. Different designs for exposure exercises.

Until 2025, no randomized controlled trial had directly compared them in actual social anxiety patients. This study is that trial.

The Study Design

Participants

Eighty-nine treatment-seeking adults were enrolled. Seventy-seven had a primary diagnosis of Social Anxiety Disorder and twelve had panic disorder. All participants were recruited through the UCLA Anxiety and Depression Research Center.

The Two Treatment Arms

All participants received nine weekly sessions of individual exposure therapy. The difference was entirely in how the exposure was conducted:

Arm 1: Inhibitory Retrieval (Craske’s model)

In this condition, the therapist designed exposures to maximize expectancy violation. Before each exposure exercise, the patient was asked: “What do you predict will happen?” They rated the expected probability and severity of the feared outcome. After the exposure, they reviewed: “What actually happened?”

The therapeutic focus was not on whether fear decreased during the exposure. It was on whether the patient’s prediction was violated — whether reality contradicted the catastrophic expectation. Progress was measured by expectancy change, not by fear reduction.

Key principles from the inhibitory learning framework were applied: variability of exposure contexts (doing the same feared behavior in different settings), removal of safety behaviors, occasional reinforced extinction (sometimes the feared outcome partially occurs, and the patient learns it is survivable), and deepened extinction (combining multiple feared stimuli in a single exposure).

Arm 2: Habituation-Focused Exposure (Traditional model)

In this condition, the therapist designed exposures to maximize within-session fear reduction. Before each exposure, the patient rated their Subjective Units of Distress (SUDS). During the exposure, they continued rating SUDS at regular intervals. The exposure continued until SUDS had dropped by at least 50% from peak.

The therapeutic focus was on fear going down. Progress was measured by the magnitude and consistency of within-session habituation — the fear curve flattening over time. The logic: if fear decreases during the session, learning has occurred.

What They Measured

Three categories of outcome:

Self-reported symptoms: Standardized anxiety questionnaires completed by patients
Interviewer-rated distress and impairment: Clinical severity ratings by assessors who were blinded to treatment condition (they did not know which arm the patient was in)
Behavioral approach task: Patients attempted a feared social behavior while their distress was measured in real time

The Results: Both Worked — But One Worked Differently

The Primary Finding: No Significant Difference on Main Outcomes

On the primary outcome measures — interviewer-rated clinical severity and impairment — there were no significant differences between the two groups at post-treatment or follow-up. Both treatments produced meaningful clinical improvement. Neither was clearly superior on the measures that matter most for formal clinical evaluation.

This is an important finding in itself. It means that habituation-focused exposure is not better than inhibitory retrieval — contradicting the assumption that within-session fear reduction is necessary for therapeutic change.

The Secondary Finding: Inhibitory Retrieval Showed Steeper Improvement

When the researchers looked beyond the primary outcomes, a pattern emerged that consistently favored inhibitory retrieval:

Self-reported anxiety decreased faster in the inhibitory retrieval group. The trajectory of improvement from baseline to post-treatment was significantly steeper — patients reported feeling better more quickly.

Session-by-session self-reported anxiety also decreased more rapidly in the inhibitory retrieval condition. Week by week, patients in this group were experiencing greater gains.

Distress during the behavioral approach task was significantly lower in the inhibitory retrieval group at post-treatment. When asked to actually perform a feared social behavior in a controlled setting, these patients showed less distress.

The Most Striking Finding: Clinically Significant Change

This is the number that matters for real patients:

43% of participants in the inhibitory retrieval condition achieved clinically significant change on self-reported anxiety at post-treatment. In the habituation condition, that number was 13%.

Clinically significant change means the patient’s score moved from the clinical range to the non-clinical range — from “has social anxiety disorder” to “does not meet threshold.” This is not a statistical abstraction. It is the difference between a person who still qualifies for a diagnosis and a person who no longer does.

The 43% vs. 13% difference was not significant on interviewer-rated measures (where both groups showed similar rates of clinical improvement). But on the patient’s own experience of their anxiety — the measure that reflects how they actually feel — inhibitory retrieval produced clinically significant change at more than three times the rate of habituation.

What This Means for How Exposure Therapy Should Be Practiced

The “Wait Until Your Fear Drops” Instruction May Be Incomplete

For decades, the standard clinical instruction during exposure has been some version of: “Stay in the situation until your anxiety comes down.” This instruction is derived from the habituation model — the assumption that fear reduction during the session is the mechanism of change.

This trial suggests that within-session fear reduction is not the critical ingredient. Patients in the inhibitory retrieval condition were not instructed to wait for fear to drop. They were instructed to notice what happened — and to compare it with what they predicted. The mechanism was cognitive (expectancy violation), not physiological (habituation).

For clinicians, this means the measure of a successful exposure is not “Did the patient’s SUDS go down?” but “Was the patient’s catastrophic prediction violated?”

Expectancy Violation Should Be Made Explicit

In the inhibitory retrieval condition, every exposure began with an explicit prediction and ended with an explicit review. This structure — predict → expose → review — makes the learning explicit. The patient does not just feel less afraid. They know why they feel less afraid: because their prediction was wrong.

This explicit structure may explain the steeper self-reported improvement. When patients understand the mechanism of their own recovery, they carry that understanding into future social situations outside therapy.

Safety Behaviors Block the Mechanism

Both models agree that safety behaviors should be eliminated during exposure. But the inhibitory learning model explains why with greater precision: safety behaviors prevent expectancy violation. If the patient uses a safety behavior during the exposure, they attribute their survival to the behavior — not to the situation being safe. The catastrophic prediction is never tested. The new competing memory is never formed.

Variability Matters More Than Repetition

The inhibitory learning model emphasizes varying exposure conditions — different locations, different people, different times of day — because variability strengthens the generalizability of the new safety memory. The habituation model emphasizes repeated exposure to the same stimulus until fear extinguishes.

This trial supports the inhibitory retrieval approach, suggesting that variability-based exposure design may contribute to stronger outcomes than pure repetition.

About the Researchers

Michelle G. Craske, Ph.D. is Distinguished Professor of Psychology and Professor of Psychiatry and Biobehavioral Sciences at UCLA. She directs the Anxiety and Depression Research Center (ADRC) and holds the Joanne and George Miller and Family Endowed Chair. She is the originator of the Inhibitory Learning model of exposure therapy, which has reshaped how exposure-based treatments are understood and delivered worldwide. She chaired the DSM-5 Anxiety Disorders Subwork Group and has published hundreds of articles, books, and clinical guides on anxiety, fear learning, and evidence-based treatment.

Michael Treanor, Ph.D. is a researcher at the UCLA Department of Psychology and a primary collaborator on the inhibitory learning research programme. He is co-author of the OptEx Nexus — a clinical tool for implementing inhibitory retrieval exposure therapy.

Tomislav D. Zbozinek, Ph.D. is at the California Institute of Technology, Division of Humanities and Social Sciences, and a long-term collaborator with the Craske lab on exposure optimization research.

Benjamin M. Rosenberg, Amy Sewart, Christina F. Sandman, and Julian Ruiz are researchers at the UCLA Department of Psychology who contributed to the design, implementation, and analysis of the clinical trial.

Does How You Do Exposure Therapy Matter? UCLA’s First Head-to-Head Trial of Inhibitory Learning vs. Habituation for Social Anxiety (2025)

Table of Contents

Why This Research Matters