Under what conditions are we justified in believing what people tell us? Traditionally, the answer has either been reductionist or anti-reductionist. Either our justification reduces to non-testimonial reasons, or we have a presumptive, though defeasible, right to believe what we are told. However, some philosophers have argued that we need to look beyond this dichotomy. Following this line of thought, this paper examines and criticizes a recent proposal by John Greco. According to Greco, testimony has two functions: (1) passing on (...) information within an epistemic community; (2) receiving information from an outside source. (1) calls for anti-reductionism, while (2) requires reductive reasons. In response to Greco, several cases are presented, showing that his position leads to counterintuitive consequences. In the light of this result, an alternative, risk-dependent, account is put forward. It is argued that the practical risks connected to receiving false information determine which epistemic treatment is appropriate. (shrink)