Trust Test: What You Extend and What You Return

The classic experiment by Berg, Dickhaut and McCabe (1995). You hold a sum and decide how much to send a stranger. Whatever you send triples, and then they decide how much to return. Sending a lot only pays if they turn out decent — which is why the decision measures trust rather than arithmetic. The second player's answer measures their trustworthiness.

Because that is the point of the test. Asking "are you a reliable person?" is useless — almost everybody says yes. How much you return out of a tripled pot, when keeping all of it costs you nothing, is the real answer. The two halves give two independent numbers, and the interesting thing is the gap.

No, and that is the confusion this test works hardest to avoid. Yamagishi showed that generalised trust and caution are separable: high-trust people are not naive — they are better at spotting the untrustworthy, because they deal with more people. High caution alongside high trust is the healthiest combination, and the result says so.

Bohnet and Zeckhauser (2004) posed the same risk twice: once resolved by chance, once by another person who could let you down. On average people accepted the lottery at 37% odds but demanded 54% when a person created the risk. The difference is the price not of risk but of the possibility of being betrayed.

The effect itself has replicated many times, but its interpretation is contested: Fetchenhauer and Dunning (2020) doubt that fear of betrayal is what actually drives trust toward strangers. We show your gap and say so plainly rather than selling the effect as settled.

The full version is 32 screens, roughly 6 minutes. The short one is 14 screens and about 3 minutes with two items per scale, so the result page marks it as a rougher read. Answers are saved on this device, so you can stop and come back.

No. It is a self-knowledge instrument built on published scales, not a clinical assessment. Use it on yourself, not on other people.

Not from a share link — that carries the scores for each scale and nothing more, so it shows where you came out without showing what you said to any single statement. What is worth knowing is the other half: the answers and the history of your runs are stored in this browser, so someone using the same browser profile on the same device can open the tool and find them. If that matters, take the test in a private window, or press "Clear history" when you are done. On a shared computer, that one habit is the whole of the protection.