You have probably won this argument a hundred times. Your therapist asks what the evidence is for the thought they'll think I'm boring, and you list the counterevidence: friends who like you, people who laugh at your jokes. You nod. You leave the session feeling lighter. Then on Thursday you walk into a room, your chest tightens, and the old prediction takes the wheel as if the argument never happened.
That isn't a failure of willpower. It points to something about how fear is actually unlearned.
Why winning the argument doesn't end the fear
A lot of early cognitive therapy leaned on verbal disputation: spot the distorted thought, weigh the evidence, build a more balanced one. It helps, and it's part of many good treatments. But many people find that the balanced thought lives in one part of the mind while the feared prediction lives in another.
Cognitive behavioral therapy has a tool built for that gap: the behavioral experiment. The idea is simple. Instead of debating a prediction, you treat it as a hypothesis and go collect data. It was developed and described in detail by clinicians such as Freeman Bennett-Levy and colleagues, whose Oxford Guide to Behavioural Experiments in Cognitive Therapy is a standard reference. In practice, it means doing the thing you expect to go badly, carefully and on purpose, and watching what happens.
Notice what changes. You're no longer asking Is this thought rational? You're asking What actually happens if I do this? The second question has an answer your nervous system can believe.
What the learning research says is happening
The best-supported explanation comes from fear-extinction research. Michelle Craske and colleagues have described an inhibitory learning model: when a feared outcome doesn't occur, you don't erase the original fear memory. You build a new, competing memory, something like this situation can be safe. Both memories stay. Which one wins in the moment depends partly on how well the new one was learned and how easily it comes to mind.
In this view, the engine of change is expectancy violation. The gap between what you predicted and what happened is what drives new learning. If you expect to be humiliated and you aren't, that mismatch is the lesson. The bigger and more clearly you stated the prediction beforehand, the more the mismatch teaches.
It also explains why talking yourself through a worry rarely lands the same way. A thought you've merely argued against hasn't produced a surprise. An experiment can.
This fits with something we already know about nonerasure: old patterns can resurface after progress, and nothing was lost when they did. The new learning sits beside the old, which is why you want to build it with real experiences and not only with reasoning.
The quiet saboteur: safety behaviors
Here's where well-meaning experiments quietly fail. Paul Salkovskis described safety behaviors, the subtle things you do to prevent the feared outcome. Rehearsing every sentence before you speak. Gripping a drink so no one sees your hands shake. Keeping your phone out as an exit. Standing near the door.
The trouble is attribution. If you survive the party while doing all those things, your mind concludes I was safe because of the safety behaviors, not the situation was safe. The prediction survives intact, and you collect no real surprise.
So a good experiment names the safety behaviors in advance and drops them, or at least reduces them, for the test. This is often the difference between an experiment that shifts something and one that leaves you saying "it went fine, but I still dread it."
A word on pacing: the aim isn't to white-knuckle through the worst scenario you can imagine. A well-designed experiment is a step you can actually take, usually one your therapist has helped you size. If anxiety is severe, or the fear is tied to trauma, build these with a professional rather than alone.
Anatomy of one experiment
Take a person who believes If I disagree with a coworker, they'll be angry at me for days. The argument approach produces a nice sentence. The experiment looks like this:
- State the prediction specifically. Not "it'll go badly" but "If I say I'd approach the deadline differently, Dana will go cold and stop talking to me this week."
- Rate your belief. How strongly do you believe it, 0 to 100? This is your baseline, and it's yours, so any number is fine.
- Name the safety behaviors. Over-apologizing, softening the point until it's meaningless, immediately agreeing if there's pushback.
- Design the test. Say it once, plainly, in the Tuesday meeting. No apology preface.
- Run it and observe like a scientist. What did Dana actually do? What did you notice in your body, and how long did the discomfort last?
- Re-rate the belief and write the result. What happened versus what you predicted. What would you now predict next time?
Step 6 is where people skip, and it's the one that makes the learning stick. An experience nobody writes down is subject to the same memory editing as everything else, and a feared outcome that didn't happen has a way of being quietly rewritten as "I got lucky."
When the experiment "fails"
Sometimes the feared thing partly happens. Dana is curt for an afternoon. That's not a wasted experiment. It's accurate information, and it usually lands in the middle: the outcome was unpleasant and survivable, much smaller than the prediction, or a different problem than the one you feared. Your therapist can help you refine the prediction and run another one.
The only truly useless version is the one you avoid, because avoidance protects the belief from ever being tested.
Your next moves
- Pick one prediction you actually make. Write it as an If I do X, then Y will happen sentence with a specific person, place and time. Vague predictions can't be proven wrong.
- Rate your belief from 0 to 100 before you do anything. Write the number down. You'll compare it afterward.
- List your top two safety behaviors for that situation. Decide which one you'll drop for the test. Choose a step small enough that you'll really do it this week.
- Do it once, then write the outcome within an hour. Record what you predicted, what happened, and your new 0–100 rating, while the details are fresh.
- Bring the result, not the interpretation, to your next session. Say "here's what I predicted and here's what happened," and let your therapist help you design the next, slightly bigger test.
Where this meets the space between sessions
Behavioral experiments are a good example of why change happens in the days between appointments. The session is where you design the test. The week is where the surprise happens, and the record you keep is what lets the surprise outlast the afternoon.
That's the gap Sesh was built for: a private place to capture what came up in therapy, including the predictions you made and what actually happened when you tested them, so the next session starts from your evidence and not your recollection. If that sounds useful, you can take a look at Sesh. The experiment works just as well with a notes app and a pen.