To assess whether alterations in the normal pattern of testosterone (T) secretion might be beneficial or detrimental, we studied a breeding population of darkeyed juncos in which we elevated T experimentally and measured its effect on potential correlates of fitness. We treated both free-living and captive males with implants that were either empty (C-males, controls) or packed with T (T-males, experimentals). Timing of implant varied and was designed to mimic natural peak breeding levels except that peaks were either prolonged or premature. We bled the birds at recapture and analyzed their plasma, and that of their female mates, for T and corticosterone (B). We also measured body mass and fat score in free-living T- and C-males. In the field, T-implants elevated T and kept it elevated for at least a month. Experimental males also had higher B than controls. In captives, the effect of the implants on plasma T was detectable within 24 hr. B in captive T-males was again higher than in captive C-males. In females, neither T nor B differed between mates of T- and C-males. T-males implanted in early spring lost more mass between implant and recapture in late spring than did controls and also had lower fat scores when recaptured. When implants were inserted in summer, treatment did not influence mass. Elevated T in early spring apparently hastened the transition from the winter to the breeding mode of fat storage. We suggest that prolonged elevation of testosterone might be selected against because of the association between T and B. Premature elevation of T might be costly because of the resultant loss of mass and fat reserves, which could lead to mortality when spring snowstorms prevent access to food.