Previous theory has suggested that altruism in other people's utility does not affect the optimal level ofpublic good provision if preferences can be represented by generalized quasilinear utility. In this paper, it is shown that the conditions under which optimal provision is unaffected by altruism are narrower than is commonly recognized. Second, even when altruism does not affect optimal provision, it might still affect the ability of public good mechanisms to achieve the optimum. Altruism's effects on three public good mechanisms are considered: the voluntary contribution mechanism, the provision point mechanism, and the Groves-Ledyard tax. It is found for agents with quasi-linear utility that altruism of a "reasonable" magnitude may help, and does not hinder the efficiency of each mechanism. Implications for the experimental evaluation of public good mechanisms are also considered.