GUIDE · CBT

Safety behaviors in anxiety: why the things that feel safe keep the fear going

You sit near the exit, keep water in your bag, and run a sentence through your head before you say it. The moment goes fine, and your mind files the lesson: the routine got you through. Safety behaviors are the things you do to head off a feared outcome or to make anxiety easier to bear. They feel like common sense. They can also keep a fear going without anyone noticing. This page covers what they look like, why they backfire, and a gradual way to test what happens without them.

What safety behaviors are

A safety behavior is anything you do, outwardly or in your head, to prevent a feared outcome or to get through a situation with less distress. Paul Salkovskis described the pattern in a 1991 paper, using the term safety-seeking behavior. He wrote that it arises from the sense that a serious threat is near, and that it can come before the feared moment (avoidance) or during it (escape).

It makes sense that you do them. They work in the moment: you use the behavior, the feared thing doesn't happen, and you feel relief.

Overt and subtle examples

Overt ones are easy to spot, like the bottle in your hand or the seat by the door. Subtle ones hide inside routines, like rehearsing a sentence in your head or stiffening your legs. Common examples fall into a few groups:

• What you carry: water, medication, a paper bag, or a phone kept in hand.
• Where you position yourself: near the exit, on the aisle, at the edge of a group.
• Who comes along: a person who can do the talking, drive, or get help.
• What you say and do with your eyes: rehearsed sentences, avoiding eye contact, running your words through your head before you speak.
• What your body does: gripping a cup, stiffening your legs, breathing deeply on purpose.
• Checking and asking: taking your pulse, checking your phone, searching symptoms, asking whether you're really okay.
• Preparing: extra planning and backup plans for a low-stakes event.

The same action can be a safety behavior for one person and ordinary for another. A water bottle on a hot hike is just water. The same bottle, carried because you're afraid a panic attack will leave you choking, is a safety behavior. A 2022 review of safety behaviors in chronic illness says telling a sensible precaution from a safety behavior can be hard, and proposes judging by function and context. The Oxford panic manual suggests a question: did I do this to stop something from happening, and what did I think would happen if I hadn't?

Why they keep anxiety going

The loop has four beats. You fear an outcome, you use the behavior, the outcome doesn't happen, and your mind gives the behavior the credit. The Oxford panic manual, by David Clark and Paul Salkovskis, describes it this way: the catastrophe's failure to occur gets attributed to the safety behavior. The belief survives because it never gets a fair test.

Relief adds a second pull. In a 2007 experiment, 27 patients with what the paper calls hypochondriasis met a personally relevant health trigger. Fourteen then performed safety behaviors such as checking, and 13 were told not to. Anxiety and urges fell in those who checked, and fell more gradually in those who didn't, over the hour they were monitored.

Some behaviors even produce the sensations they're meant to prevent. The Oxford manual notes that walking with stiff legs makes people feel more unsteady, that breathing fast and deep makes some people feel more short of breath, and that pushing distressing thoughts away can make them intrude more. In a 2021 experiment with 57 adolescents, conversations with self-focus and safety behaviors left people feeling and looking more anxious, and performing worse, than conversations with an outward focus and no safety behaviors.

Three examples (invented)

These examples are made up to show the shape of an experiment. The people aren't real, and nothing here is data.

Social anxiety: Tomas rehearses every sentence in his head before speaking in meetings. He predicts that if he speaks unrehearsed, he'll stumble and his coworkers will decide he's not competent. In the next meeting he makes one unplanned comment. He trips over a word, someone nods, and the discussion moves on. The prediction didn't come true.

Panic: Ana holds a water bottle on the train, sits by the door, and keeps her phone out. She predicts that without them a panic attack will end with her fainting. Her doctor has already checked her heart. Her first experiment drops only the bottle. Her heart pounds and her hands tingle, and she doesn't faint. She writes that the sensations came, the fainting didn't, and the bottle wasn't the reason. The seat and the phone stay for now.

Health anxiety: Jordan takes his pulse several times a day and searches his symptoms whenever his chest feels odd. His doctor has found nothing wrong. He predicts that without the search he'll miss a warning sign. After the next flutter he waits an hour before any checking or searching. The urge is strong, then eases without a search, and nothing happens that a search would have caught.

More on the social version: fear of judgment.

When dropping them backfires

Dropping everything at once: Going without every hedge in your hardest situation can overwhelm you, and one rough day can convince you that you need them. One NHS trust's self-help guide goes gradually: list the behaviors, build a ladder of situations from easiest to hardest, and work upward.

White-knuckling: If you clench, grip, distract yourself, or keep tight control to get through, you may have swapped one safety behavior for another. The Oxford manual counts tight control of thoughts and behavior as safety behaviors, and the NHS trust guide says not to take on new ones. The question is whether the feared outcome happens, not how calm you felt.

Real precautions: Some measures are medical or fit a real risk, and those aren't the target. If a clinician told you to carry an inhaler or an epinephrine pen, or to take a medication, keep doing it. Don't stop prescribed medication or any medically advised precaution because of this page, and bring any question about changing a prescription to the person who prescribed it. The 2022 review proposes adapting treatment for chronic illness, which takes a clinician.

How this fits with avoidance and reassurance seeking

Avoidance means staying out of the situation. Safety behaviors let you in, with conditions. The NHS trust guide lists the two separately, and the 2022 review calls safety behaviors a more subtle kind of avoidance. You can stop avoiding and still learn nothing: you go to the party but stay by the door with your phone out. Most experiments need both moves, going in and going in without the hedge.

Reassurance seeking is a safety behavior aimed at other people or search engines: asking whether you're really okay, rechecking, searching again. ADAA clinicians Sally Winston and Martin Seif describe how the brief drop in anxiety reinforces the worry that came before it, so the urge for certainty returns. For more on that pattern, see intolerance of uncertainty and health anxiety.

What the research does and doesn't say

Salkovskis's 1991 paper proposed that safety-seeking behavior can maintain anxiety disorders by preventing disconfirmation of the threat. In a 1999 experiment, patients with panic disorder and agoraphobia did 15 minutes of exposure while either stopping or keeping their safety behaviors. When a behavior test was repeated within two days, the group that stopped showed a significantly greater decrease in catastrophic beliefs and anxiety.

The wider evidence is less tidy. A 2010 systematic review found experimental evidence for harm less consistent than the cognitive models emphasize, yet still recommended identifying safety behaviors and dropping them in therapy. A 2016 review called the findings on safety behaviors during exposure limited, mixed, and controversial. A 2016 meta-analysis of 20 studies found no statistically significant difference in fear from removing or adding them during exposure. Rachman, Radomsky, and Shafran argued in 2008 that judicious use, especially early in treatment, can help. A 2019 randomized trial of 60 adults with clinically significant spider fear found both groups improved, with no significant group differences in outcome or acceptability between exposure that removed safety behaviors and exposure that used them judiciously.

The studies we read tested short exposure sessions, some for a single fear such as spiders, so they can't say what happens in your own life. As far as we could find, no study tested the worksheet below. It's assembled from clinical guides. The pattern is plausible and backed by several experiments, but it isn't proven that every safety behavior does harm or that dropping each one helps. That's why the method tests your own prediction and goes gradually.

Drop one safety behavior and test the prediction

You'll need a notes app or a sheet of paper. Work on one safety behavior at a time.

1. List what you do to feel safer. Think back over two or three recent anxious situations and write what you did to stop the feared thing or get through it, including things in your head. The Oxford manual's question helps: what did I think might happen if I hadn't? Mark anything a clinician told you to do with an M for medical, and leave it alone.
2. Rank them. Rate each one from 0 to 10 for how anxious you'd feel going without it, then sort from lowest to highest to make your ladder.
3. Pick one near the bottom. Choose one you'd rate around 3 or 4, and a situation where you can test it soon. Leave the rest alone, and don't add a substitute.
4. Write your prediction. Finish this sentence: if I go into [situation] without [behavior], then [specific outcome]. Rate how strongly you believe it from 0 to 100. Choose an outcome you could see, like fainting or a coworker's reaction, instead of a feeling like "it'll be awful."
5. Run the experiment. Go in without that one behavior and keep everything else the same. Stay long enough for the prediction to have a chance to come true. If you catch yourself clenching, gripping, or distracting, note it and loosen up.
6. Compare prediction and outcome. Afterward, write what happened as plain facts and re-rate your belief from 0 to 100. If the prediction came true, write how you coped. The test is whether the feared outcome happened, not how calm you felt.
7. Repeat, then move up. Repeat it in similar situations. Once your belief has dropped and stays down, take the next item on the list. Go gradually, one behavior at a time. If one feels like too much, go back a rung or ask a therapist to plan it with you.

For the general format of testing a belief, see behavioral experiments. If panic is the main fear, fear of panic attacks and CBT for panic attacks cover the cycle around it.

Plan your first experiment with Helpy

The chat below runs with Helpy, the site's AI assistant. Tell it one thing you do to feel safer in a situation, and it can help you word a prediction and plan a first experiment. It can't tell you what's medically safe to drop.

Evidence and sources

• Salkovskis (1991), Behavioural Psychotherapy: Uses the term safety-seeking behavior; argues it blocks disconfirmation. Abstract only.
• Clark and Salkovskis (2009), panic manual for therapists: Examples, amplified sensations, experiments, cautions. Not a trial.
• NHS trust social anxiety guide (Cumbria, Northumberland, Tyne and Wear): Examples and a gradual ladder. Self-help, not a trial.
• Salkovskis and colleagues (1999): Stopping safety behaviors during exposure lowered beliefs and anxiety more. Abstract only.
• Abramowitz and Moore (2007): 27 patients; checking brought relief, not checking a slower drop. Abstract only.
• Leigh, Chiu and Clark (2021): 57 adolescents; self-focus and safety behaviors, varied together, raised anxiety. Abstract only.
• Winston and Seif, ADAA: Brief relief reinforces worry. Expert commentary, not data.
• Helbig-Lang and Petermann (2010): Systematic review; harm evidence less consistent than models predict. Abstract only.
• Rachman, Radomsky and Shafran (2008): Argues judicious use can help. Abstract only.
• Blakey and Abramowitz (2016): Findings limited, mixed, controversial. Abstract only.
• Meulders and colleagues (2016): Meta-analysis of 20 studies; no significant difference. Abstract only.
• Blakey and colleagues (2019): 60 adults with spider fear; no group differences. Abstract only.
• Sharpe and colleagues (2022): Precautions versus safety behaviors in chronic illness. Abstract only.

Read next

Behavioral experiments: how to test an anxious belief in real lifeThe tool for dropping a safety behaviorFear of a panic attack: how anticipatory anxiety keeps you trappedHow the fear of fear keeps panic goingIntolerance of uncertainty: why your brain demands guarantees and how to let goWhy checking and reassurance feel necessaryReassurance seeking: why asking "are you sure?" keeps the doubt coming backWhy checking and asking keep the doubt going

← All CBT guides