Behavioral Experiment

Research

What the evidence says

A short account of where behavioral experiments come from and what has been found, written for someone deciding whether to try one. Every claim below points to a reference in the list at the end. Where the evidence is thin, it says so.

Where the method comes from

Cognitive behavioral therapy grew out of Aaron Beck's work on depression in the 1970s and was extended to anxiety in the 1980s (Beck, Rush, Shaw and Emery, 1979; Beck, Emery and Greenberg, 1985). Its central claim is that thoughts, feelings, and behavior each shape the other two, so a change in what you do can change what you believe. Behavioral experiments are the most direct expression of that claim: rather than debating a belief, you arrange to test it.

The method was set out in detail by the Oxford group in 2004. The Oxford Guide to Behavioural Experiments in Cognitive Therapy (Bennett-Levy et al., 2004) collected the format, the rationale, and hundreds of examples across anxiety, depression, low self-esteem, and other problems, and its chapter on designing experiments (Rouf et al., 2004) is still the standard description of the planning steps. A 2025 guide from the same tradition (Warnock-Parkes et al., 2025) updated the method into seven steps and drew on two decades of trials in anxiety disorders and post-traumatic stress. The worksheet on this site follows those seven steps.

Why doing beats arguing

Bennett-Levy (2003) compared two staples of cognitive therapy, the thought record and the behavioral experiment, and found that people rated experiments as more powerful for changing beliefs and behavior. His explanation drew on the distinction between knowing something intellectually and knowing it in a way that changes how you feel, and on the observation that direct experience reaches the second kind of knowledge more reliably than reasoning does.

The mechanism is not mysterious. Anxiety persists partly because feared predictions are never put to a fair test (Clark, 1999). Avoidance removes the test altogether, and safety behaviors corrupt it: when the feared outcome fails to appear, the credit goes to the precaution rather than to the fact that the prediction was wrong (Salkovskis, 1991). An experiment removes the precaution, so the result can count.

Safety behaviors

The evidence that safety behaviors maintain anxiety is among the more solid findings in this area. Wells and colleagues (1995) showed that people with social anxiety who dropped their safety behaviors during a feared situation learned more from it, with larger drops in anxiety and belief, than people who kept them. Salkovskis and colleagues (1999) found the same for panic with agoraphobia. A systematic review across anxiety disorders (Helbig-Lang and Petermann, 2010) concluded that safety behaviors generally interfere with treatment, while noting a debate about whether some carefully used precautions can help early on (Rachman, Radomsky and Shafran, 2008). A later critical analysis from an inhibitory learning perspective (Blakey and Abramowitz, 2016) reached a similar conclusion: on balance, dropping them helps. Attention matters too. Wells and Papageorgiou (1998) showed that shifting attention outward during a feared social situation reduced anxiety and negative beliefs on its own.

Experiments compared with exposure

Exposure therapy and behavioral experiments look alike from the outside, and the honest position is that they have rarely been compared head to head. McMillan and Lee (2010) reviewed the studies that existed and found some evidence favoring experiments, but the number of direct comparisons was small and the review is careful about that. What the two procedures share is that both work; what separates them is the aim. Exposure targets the anxiety itself, and the modern account of how it works stresses new learning that inhibits the old fear (Craske et al., 2008, 2014, 2022), which is also what an experiment is trying to produce. The distinction is described in more detail in the Behavioral experiments article.

Does CBT work, and does the homework matter?

Reviews of the trials are consistent that CBT is effective for anxiety disorders and depression (Butler et al., 2006; Hofmann et al., 2012; Cuijpers et al., 2016; Kaczkurkin and Foa, 2015). Within it, the amount and quality of between-session work is associated with better outcomes: a meta-analysis of homework compliance found a moderate relationship with results (Kazantzis et al., 2016). Behavioral experiments are, in practice, the homework that this evidence is about.

Doing it on your own

Self-guided CBT has been studied more than most people expect. Guided self-help, where a professional checks in even briefly, produces results comparable to face-to-face therapy for many people with anxiety and depression (Cuijpers et al., 2010; Carlbring et al., 2018; Andrews et al., 2018). Fully unguided self-help also helps, but effects are smaller and more people stop partway. The difference matters most for people with more severe symptoms: an individual patient data analysis of internet CBT for depression found that guidance made little difference for mild symptoms and a substantial difference for moderate to severe ones (Karyotaki et al., 2021), and a component analysis pointed to behavioral work as one of the active ingredients (Furukawa et al., 2021). The practical reading, given in the article on working alone or with a therapist, is that a worksheet like this one is reasonable on its own for mild difficulties and does more as a companion to therapy than as a replacement when problems are severe or long-standing.

Step ladders and distress ratings

Graded hierarchies come from Wolpe's early work (Wolpe, 1969), and the 0 to 100 distress scale he introduced is still used because it is quick and holds up reasonably well (Tanner, 2012). The Oxford group notes that when the focus is on testing a belief, a formal hierarchy is often unnecessary; situations are chosen for how well they test the belief and whether the person is willing, not for their rank on a list (Bennett-Levy et al., 2004). Where the target is out of reach, the graded approach described in standard exposure texts applies (Abramowitz, Deacon and Whiteside, 2019), with the reminder that repeating a step, and varying it, produces more durable learning than a single pass (Rowe and Craske, 1998; Craske et al., 2014).

What this does not show

None of the studies above tested this website or this app. They tested the method, mostly delivered by trained therapists, sometimes in guided self-help programs. A worksheet cannot notice what you are leaving out, and it cannot tell whether a belief is one that should be worked on with a professional. Where the evidence points to support making a difference, we say so rather than promising the worksheet will do the job alone.

References

Full references are given in APA style. DOIs link to the publisher's record. Where a source is a book, no link is given.