Animals are selective about when to learn by observing others. Models predict that social information becomes less reliable in uncertain environments, and therefore animals should reduce their use of social information in these environments; however, these parameters are often difficult to manipulate and control. We investigated how information reliability and environmental uncertainty affect the use of both social and nonsocial signals. Captive blue jays, Cyanocitta cristata , were given a choice between two perches, one of which was rewarded. Jays could see either a social signal (a conspecific) or a nonsocial signal (a light) that provided some information about the rewarded perch. The nonsocial signal was yoked to the bird that generated the social signal, ensuring the two signals were of identical reliability. We manipulated signal reliability (i.e. the probability that the signal correctly indicated the rewarded perch) and environmental certainty (i.e. the probability that a given perch was rewarded). Qualitatively, jays used both social and nonsocial signals more often when the signals were reliable, and used them less often when environments were predictable. However, jays used social signals less than equally reliable nonsocial signals when environments were unpredictable. Our results suggest that signal reliability and environmental predictability interact to determine signal use, but they do not affect social and nonsocial signals in the same way. • The reliability of social information has rarely been quantified. • We presented blue jays with social and nonsocial signals of identical reliability. • Blue jays attended to social signals less than identical nonsocial signals. • Blue jays used both types of signal more when the signals were reliable. • Blue jays used both types of signal less when environments were predictable.