Numerous studies have shown positive associations between ornaments and condition, as predicted by indicator models of sexual selection. However, this idea is continuously challenged by opposite results, which reveal our lack of full understanding of how sexual selection works. Environmental heterogeneity may explain such inconsistencies, but valid field tests of this idea are currently lacking. We first analysed the relationship between condition and ornament expression from nine populations over 7 years in a wild bird, the red grouse Lagopus lagopus scoticus. We then manipulated male aggressiveness at the population level by means of testosterone implants in a replicated field experiment. We found that the relationship between condition and ornamentation varied greatly between environments and became stronger when environmental conditions (ECs) were worse or when aggressiveness in the population was experimentally increased. Some ornaments may therefore reliably advertise a better condition only in adverse ECs. Considering environmental heterogeneity can help reconcile conflicting findings regarding the reliability of ornaments as indicators of condition and will help our understanding of sexual selection processes.