Numerous theoretical studies have investigated how limited dispersal may provide an explanation for the evolution of cooperation, by leading to interactions between relatives. However, despite considerable theoretical attention, there has been a lack of empirical tests. In this article, we test how patterns of dispersal influence the evolution of cooperation, using iron-scavenging in the bacterium Pseudomonas aeruginosa as our cooperative trait. We found that relatively limited dispersal does not favor cooperation. The reason for this is that although limited dispersal increases the relatedness between interacting individuals, it also leads to increased local competition for resources between relatives. This result supports Taylor's prediction that in the simplest possible scenario, the effects of increased relatedness and local competition exactly cancel out. In contrast, we show that one way for cooperation to be favored is if individuals disperse in groups (budding dispersal), because this maintains high relatedness while reducing local competition between relatives (relatively global competition).K E Y W O R D S : Experimental evolution, kin selection, local competition, microorganisms, relatedness, social evolution.Explaining cooperation is fundamental for understanding the evolutionary transitions from associations of replicator molecules to single-celled organisms to complex animal societies (Maynard Smith and Szathmary 1995; Hamilton 1996;Frank 2003;West et al. 2007a). Why should an individual carry out a costly cooperative behavior that benefits other individuals? Hamilton's (1963Hamilton's ( , 1964) theory of kin selection provides an explanation for how cooperation can be favored if interacting individuals are likely to share the gene for the cooperative behavior, for example, if they are relatives. By helping a close relative reproduce, an individual passes its genes to the next generation, although indirectly. This theory is encapsulated in a pleasingly simple form by Hamilton's (1963Hamilton's ( , 1964 rule, which states that altruism is favored when rb − c > 0; where c is the fitness cost to the altruist, b is the fitness benefit to the beneficiary and r is their genetic relatedness. Hamilton (1964Hamilton ( , 1972 originally suggested that a high relatedness could arise in two ways: (1) direct kin recognition between individuals, or (2) limited dispersal (population viscosity) can keep relatives together, and so could favor indiscriminate altruism.The possible role of limited dispersal in favoring cooperation has attracted much theoretical attention (Hamilton