he World Bank report Assessing Aid argues that aid can have positive effects on growth and infant mortality, but only when good policies are being followed by the recipient. It follows, especially since aid is fungible, and so cannot be targeted to particular uses, that donors should focus their aid on low-income countries with good policies (i.e. apply greater selectivity). This paper explores a number of weaknesses in these arguments. The growth regressions are not robust, so that different results can be obtained with relatively minor variations in model specification. In particular, the argument that aid only works when policies are right is not supported in other studies - and even the World Bank's evidence can be interpreted as saying policies work better when supported by aid inflows. The choice of which policies are good policies is also problematic, and the analysis in the report ignores the likely presence of threshold effects and other non-linearities; others would, anyhow, propose a different set of right policies, especially if the focus is poverty reduction rather than growth. The importance of fungibility may be over-stated so that donors can in fact target poverty-reduction activities, suggesting that the selectivity rules proposed in Assessing Aid are misleading. Even if the report's proposals are to be accepted it is silent on a number of important issues-such as whether to use the level or change in the index-that face the aid manager in practice.