'Incentive pay' can backfire, according to some new research. Margaret Lee and Ye Li at Columbia University got people to find as many words as they could in four minutes from the letters 'a d e r s t w'. Some solvers were not paid at all. Others were paid per word they found. And others were told that their performance would determine the pay of the next participant.

They found that when people were paid per word, they actually found fewer words than people who weren't paid: an average of 25.6 compared to 29. This confirms Daniel Pink's argument that extrinsic rewards can actually crowd out intrinsic motivation.

However, people whose performance determined the pay of the next participant actually found more words than either - an average of 30. Prosocial motivations, then, can be more powerful than selfish ones.

In another experiment, Li and Lee discovered why. It is not because people are purely altruistic, but rather because they are conditional reciprocators. They'll work hard for others as long as they expect others to work hard in turn. If this expectation is not fulfilled, prosocial motivation doesn't work any better than flat-rate pay.

In a third experiment, Li and Lee found that such reciprocity was especially powerful if people expected to be accountable for their actions. They got random pairs of people to type the letters F and J alternatively for 15 minutes, with payment if more than 400 pairs were typed. Some were paid for their own performance, and some for their partners. They found that people working for their partner typed almost 10% more letters if they knew they would meet their partner afterwards than if they didn't. Such people also typed slightly more letters than those paid for their own performance.

This result might generalize. Most of us see our co-workers every day. If we know their pay depends on our efforts and vice versa, we might therefore be pressurized into working harder. And such pro-social pay gives workers an incentive to monitor each other and so reduce shirking.

Now, it doesn't follow from this that all firms should adopt 'payment for others' - type incentives. I suspect that the power of prosocial motivations depends a lot upon detailed context.

What this research does do, though, is further undermine the notion that high-powered selfish incentives are the best way of motivating people. This belief probably rests more upon a desire to justify inequality than it does upon a basis of empirical evidence.