Dropping Safety Behaviors Within Experiments

Remove the subtle avoidance strategies (safety behaviors) that prevent experiments from genuinely testing the belief.

Why it works

Safety behaviors are covert avoidance strategies: speaking quickly so people don’t notice anxiety, sitting near an exit, over-preparing to reduce perceived risk. They are maintained because they appear to prevent the feared outcome — "nothing went wrong, but only because I was prepared." This interpretation prevents belief disconfirmation: the person concludes the safety behavior (not the experiment) caused the success, and the belief stays intact. Dropping safety behaviors is therefore essential for genuine experimental tests — it must be the situation itself, not the avoidance, that produces the positive outcome.

How to do it

1. Before designing an experiment, list every safety behavior you might use in that situation.
2. For each one, evaluate: "Does this behavior prevent the experiment from genuinely testing my prediction?"
3. Design the experiment to drop the most belief-critical safety behavior.
4. Run the experiment without that behavior and note what happened.
5. Compare: "Did the absence of [safety behavior] confirm the fear, or disconfirm it?"

Evidence

Safety behavior removal is a well-supported CBT technique with RCT evidence for social anxiety and panic disorder. Wells et al. found that attention retraining and safety behavior removal outperformed exposure alone. (rct)

Evidence is strongest for social anxiety and panic; application across all anxiety presentations is clinical extrapolation from a well-evidenced foundation.

Sources

• Wells et al. (1995), "Social phobia: the role of in-situation safety behaviors in maintaining anxiety and negative beliefs", Behavior Therapy
• Salkovskis (1991), "The importance of behaviour in the maintenance of anxiety and panic", Behavioural Psychotherapy

Common mistake

Partially dropping safety behaviors — stopping one (not speaking quickly) while maintaining another (sitting near the exit) — means the experiment still cannot attribute success to the situation rather than the remaining avoidance.