Many species of birds conspicuously call or sing early in the morning, thereby creating an avian dawn chorus. While these vocalizations probably function to advertise territory occupancy, when species should start singing is not well understood. A common explanation is that birds sing at dawn to maximize signal transmission due to low atmospheric turbulence during the early morning (acoustic transmission hypothesis); however, this idea does not explain why species in the chorus often start singing at different times. Here we test a version of this hypothesis: interspecific differences in call start times during the dawn chorus are a result of noise avoidance at the song's frequency. To test this hypothesis we quantified the spectral and temporal properties of the dawn chorus and the acoustic landscape at Barro Colorado Island, Panama. Inspection of 36 dawn recordings at 12 sites identified 27 species of birds, with most species having consistent start times across sites and days. In contrast to expectations, we found that birds singing within the same frequency range did not temporally partition acoustic space during the chorus. Discriminant function analysis revealed that nearly all species in the chorus produced distinctive vocalizations. Surprisingly, song start time was positively related to peak frequency. Birds with high-frequency vocalizations did not begin singing until after orthopteran insects stopped producing loud sounds at the same frequencies. These results suggest that birds that sing at frequency bands shared by nocturnal insects avoid acoustic masking by delaying song start times. Playbacks of emerald cicada, Zammara smaragdina, buzzes conducted during the dawn chorus produced an inhibitory effect on the call activity of birds singing within the bandwidth of cicada calls. Thus, insect noise appears to create an important frequency-dependent constraint on the calling activity of birds. This constraint is overcome by temporal partitioning of acoustic space.