Verke Editorial

Behavioral experiments: testing what you're afraid of

Verke Editorial · May 5, 2025

What would happen if you asked a question in the meeting instead of staying silent? What would happen if you said "no" to the extra task? What would happen if you told someone you disagreed with them?

You already have answers to these questions. Your anxiety wrote them. "They'll think I'm stupid." "They'll stop liking me." "It'll be awkward and I'll regret it."

But here's the thing: you've never actually checked. A behavioral experiment is how you check. Not by thinking harder about whether the fear is rational — but by doing the thing and finding out what actually happens. A thought record asks "is this thought accurate?" A behavioral experiment asks "let's find out." Your brain trusts evidence it generated through action more than conclusions it reached through reflection.

Every behavioral experiment follows four steps. Here they are — then we'll see them in action.

1. Predict. Write down exactly what you think will happen. "If I [do X], then [Y will happen]." Rate your confidence 0–100.
2. Test. Do X. Make it manageable — not your worst fear, but not trivially easy either.
3. Compare. What actually happened vs. what you predicted? Side by side. Be specific.
4. Learn. What does this tell you about your original belief? Update your confidence. Design the next experiment.

A thought record works through reasoning: you list the evidence for and against a belief, then write a balanced thought. That's powerful — but it stays inside your head. A behavioral experiment works through experience: you go into the world and generate real data. Think of it this way: a thought record is a closing argument. A behavioral experiment is presenting a witness. Your brain trusts the witness more.

McMillan and Lee (2010) found that behavioral experiments produced both cognitive and emotional change, while thought challenging alone often shifted the cognitive layer and left the emotional layer stuck. When thought records plateau — "I know the balanced thought but I still feel anxious" — behavioral experiments break through, because the evidence is experiential, not just intellectual.

The prediction

"If I start a conversation with a stranger at this networking event, they'll look uncomfortable and try to end the conversation quickly. Confidence: 80%."

This prediction is specific enough to test. Not "something awkward will happen" — that's unfalsifiable. The test is: did the person look uncomfortable? Did they try to end the conversation quickly? Those are observable. If you can't tell whether your prediction came true or not, it wasn't specific enough.

Designing the test (the hardest part)

The experiment has to be fair — not rigged to confirm or deny the fear. This is where most people stumble.

Too easy: "I'll say hi to someone who already smiled at me." That tests nothing — you already know that works. Too hard: "I'll approach the most intimidating person in the room and tell a joke." That's flooding, not testing.

Fair test: "I'll make a contextual comment to someone standing near me who I haven't met. Something about the event, the speaker, the venue. One comment. See what happens."

One more distinction that matters: knowing you can leave is fine — that's safety planning. Rehearsing exit lines in advance is a safety behavior that undermines the experiment. The difference is between having a parachute and refusing to look out the window.

What actually happened

"I commented on the speaker's point to the woman next to me. She smiled, agreed, and added her own take. We talked for about 3 minutes about the topic. She introduced herself. I introduced myself. We both moved on to get coffee. She didn't look uncomfortable. She didn't try to end the conversation. She seemed… normal."

Write the result down within minutes. Memory warps toward the prediction. An hour later, anxiety will rewrite this as "it was fine, but…" or "she was just being polite." The written record protects the evidence.

Prediction vs. reality

Predicted: uncomfortable look, quick exit. Confidence: 80%. Actual: smile, 3-minute conversation, mutual introduction. Updated confidence in the original prediction: 20%.

One experiment doesn't prove "everyone will always be friendly." It proves one thing: the prediction was wrong this time. After 3–5 experiments testing similar predictions, a pattern emerges. Your brain starts updating the default from "they'll reject me" to "most people are neutral-to-friendly."

Fear of imperfection at work

Prediction: "If I share my idea before it's fully polished, my boss will think I'm incompetent. Confidence: 75%."

Test: Share an idea at 80% readiness in a low-stakes meeting.

Result: Boss asked two clarifying questions. She could answer one and acknowledged she hadn't thought through the other. Boss said "interesting, let's develop this." No negative reaction.

Updated confidence: 15%. The key discovery: questions aren't attacks — they're engagement. Many anxious people misread questions as criticism. The experiment surfaces the misinterpretation.

Fear of saying no

Prediction: "If I decline this extra task, my colleague will think I'm lazy and be cold to me. Confidence: 70%."

Test: Politely decline one non-essential request this week.

Result: "No worries" response. Normal friendliness at lunch. No consequences detected.

Updated confidence: 10%. The prediction was based on guilt, not evidence. That's the pattern behavioral experiments surface again and again: the feeling was real, but the prediction it generated was wrong.

If either of these examples resonates, you might also find these useful: why you're afraid of being judged and what to do when you're dreading social events.

List 5–10 situations you currently avoid. Rate each for difficulty on a 0–10 scale. Start at 3–4, not 8–9.

Each successful experiment builds evidence for the next — you're not just building courage, you're building a database of prediction errors. Repeat 2–3 experiments at the same difficulty level before stepping up. Your brain needs multiple data points before it updates a deeply held belief. One data point is a fluke. Five is a pattern.

The goal isn't fearlessness. It's accurate prediction. Some situations genuinely are difficult. The experiment helps you distinguish "difficult but manageable" from "catastrophic" — which is the distinction your anxiety can't make on its own.

It happens. Sometimes the feared thing occurs — partially or fully. A "failed" experiment is still data. Ask yourself three things:

• Was it as catastrophic as predicted? Usually the magnitude was wrong even if the direction was right.
• Did you cope with it? Usually yes — which is itself a discovery.
• Was the aftermath as bad as feared? Usually no. The feared consequence rarely materializes at the scale anxiety predicted.

When the result genuinely is bad: drop down the stepladder and try a smaller test. Not "I failed" — "I need a smaller experiment." The methodology isn't wrong; the difficulty calibration was. Adjust and try again. That's what experimenters do.

If you want a thinking partner who can help you design experiments that are genuinely fair — not too safe, not too terrifying — Judith is built for this. Her approach uses Cognitive Behavioral Therapy, the modality this article draws from, to help you turn fears into testable predictions and then help you make sense of the results. She remembers what you've been working on across sessions, so each experiment builds on the last.

Chat with Judith about this — no account needed