Rules are thought to persist to the extent that the direct benefits of having them (e.g. reduced transactions costs) exceed the costs of enforcement and of occasional misapplications. We argue that a second crucial role of rules is as screening mechanisms for identifying cooperative types. Thus we underestimate the social value of rules when we consider only their instrumental value in solving a particular problem. We demonstrate experimentally that costly rule-following can be used to screen for conditional cooperators. Subjects participate in a rule-following task in which they may incur costs to follow an arbitrary written rule in an individual choice setting. Without their knowledge, we sort them into groups according to their willingness to follow the rule. These groups then play repeated public goods or trust games. Rule-following groups sustain high public goods contributions over time, but in rule-breaking groups cooperation decays. Rulefollowers also reciprocate more in trust games. However, when individuals are not sorted by type, we observe no differences in the behavior of rule-followers and rule-breakers.