We examine a simple theory of altruism in which players' payoffs are linear in their own monetary income and their opponents. The weight on the opponent's income is private information and varies in the population, depending, moreover, on what the opponent's coefficient is believed to be. Using results of ultimatum experiments and the final round of a centipede experiment, we are able to pin down relatively accurately what the distribution of altruism (and spite) in the population is. This distribution is then used with a reasonable degree of success to explain the results of the earlier rounds of centipede and the results of some public goods contribution games. In addition, we show that in a market game where the theory of selfish players does quite well, the theory of altruism makes exactly the same predictions as the theory of selfish players. (Copyright: Elsevier) (This abstract was borrowed from another version of this item.)