Some fears can't be talked out of.
You can examine the evidence, weigh a thought, and still not believe it in your body. "If I speak up in the meeting, I'll be humiliated" can survive any amount of reasoning, because deep down you've never actually found out. That's where thinking reaches its limit and doing takes over. A behavioural experiment is a way of testing a prediction in the real world rather than in your head.
A prediction and a test.
The structure is simple. You name the specific prediction ("if I ask a question, my voice will shake and people will laugh"), rate how strongly you believe it, then design a small action that would put it to the test. You carry it out, watch what actually happens, and compare the outcome with what you predicted. It's the scientific method applied to your own fears — a hypothesis, an experiment, and an honest look at the result.
Designing one that's fair.
A good experiment is specific and survivable. The prediction has to be concrete enough to be proved or disproved — "it'll go badly" is too vague; "three people will visibly cringe" can actually be checked. The test should stretch you without overwhelming you, and it works best when you drop the safety behaviours (over-preparing, gripping the table, rushing) that usually let you explain away a good result. Then you note what really happened, in detail, against the prediction.
Why doing beats discussing.
Direct experience tends to shift a belief in a way that argument can't reach. Reviews of cognitive therapy have found that behavioural experiments — actually testing the feared prediction — are among its most powerful ingredients, often more convincing than discussion alone. The reason is simple: a fear that has never been tested can always claim it was right to keep you away. Once you've seen the prediction fail with your own eyes, that claim loses its grip.
Start small.
The point isn't to fling yourself at the scariest thing first — it's to gather evidence, one manageable test at a time, and let it accumulate. Sometimes the prediction even comes partly true (your voice does shake) and you discover the more important truth: it was bearable, and the catastrophe didn't follow. Each experiment, whatever the result, gives you something reasoning never could — real-world data about what actually happens when you stop running.