Many economists believe we behave solely in our own best interests. Credit: © Corbis

Get one student to give another money to invest on their behalf and you're bound to uncover something fundamental about human behaviour.

Behavioural scientist Ernst Fehr did. He is now the first to prove that we reward those who seem to trust us but think nothing of swindling those who doubt our honour - even when we stand to gain nothing by doing so1.

"If you trust people, you make them more trustworthy," says Fehr, who works at the University of Zurich, Switzerland. "Sanctions designed to deter people from cheating actually make them cheat."

Until now, scientists and economists have assumed that we help or punish others only when it benefits us. But in the real world there's a wealth of evidence to the contrary. We regularly reward people whom we may never see again, in tipping waiters, say; and we risk sacrificing ourselves to punish others - fighting those who verbally abuse us, for instance.

This mismatch between hypothesis and reality arose because customs - such as legally binding contracts, handshakes and marriage vows - make it difficult to study how and why people trust and betray each other. To get past these trappings, Fehr went to a university canteen with his colleague Bettina Rockenbach of the University of Erfurt, Germany.

Using real money, the duo set up experiments in which one student invested money with an anonymous other who was guaranteed to make three times the original investment.

You don't threaten your workforce with the sack unless you want trouble Truman Bewley Yale University

In one scenario, investors were asked to threaten their trustees with a financial penalty if they failed to return a specified sum. In another, they were not to mention a punishment that the trustee knew existed. Finally, they were asked to invest the money without any notion of punishment on either side.

Where punishment was threatened, investors got the least cash back. In the absence of punishment, investors saw a favourable return on their money. But where punishment was an option but was not used, trustees rewarded investors most handsomely.

What Fehr identified among students applies to many human interactions, says Truman Bewley, an economist at Yale University in New Haven, Connecticut. Labour markets are one example: "It's now becoming accepted that you don't threaten your workforce with the sack unless you want trouble," Bewley says.

Defining the way we relate to each other scientifically should help social scientists and economists to update their models. Economics is dominated by the idea that everyone operates in his or her own best interests, explains Bewley. "If you can find experimental evidence to the contrary, people are forced to take notice."