To test how real the tainted altruism phenomenon is and how it might be corrected, Newman and Cain devised four experiments. In the first, subjects were offered two scenarios in which a man hoping to impress a woman offered to do volunteer work at her place of business. In one version, she worked at a coffee shop; in the other, at a homeless shelter. Some subjects read one version, some read the other and a third group read both. All of them were asked to rate, on a scale of 1 to 9, how much they liked the man and how moral and ethical they thought he was. They were also asked how much his actions benefited society.

Remarkably, the people who read the homeless shelter scenario rated the man as less moral—4.75 out of 9.00—than the people who read the coffee shop scenario, who gave him a 5.62. What’s more, they rated the benefit of his work to society as pretty much equal in both cases. Only the group that had read both scenarios—and  thus had a chance to compare the kinds of good that were being done—had a more balanced view of things, with the man being seen as equally moral in both cases and doing much greater good, with a 6.46 score, in the homeless shelter than in the coffee shop (4.67).

“To some extent,” Newman and Cain wrote, “participants recognized the inconsistency of doing some good as worse than doing no good at all.”