We experimentally test whether intentional and observable discriminatory pay of symmetric agents in the Winter (2004) game causes low paid agents to reduce effiort. We control for intentionality of wages by either allowing a principal to determine wages or by implementing a random process. Our main observations are that discrimination has no negative effiect on effiorts and principals do not shy away from using discriminatory pay if it is observable. Rather, with experience discrimination enhances efficiency as it facilitates coordination among agents. The only evidence for reciprocity is that subjects receiving a low payment from a principal (discriminatory or not) exert signiïcantly less effort.