This paper describes the results of a detailed analysis of 52 song bouts, recorded from 22 great tits, Parus major, during the dawn chorus. A song bout consists of a number of song bursts (called strophes) separated by periods of silence. High quality males, as measured by average strophe length, sang their bouts with a higher percentage performance time (i.e. the percentage of time spent singing in a bout), but the average number of strophes per bout was not related to male quality. In 31 of 52 bouts there was a systematic decrease in the percentage performance time throughout the bout. This was mainly caused by a prolongation of the pauses between the strophes, and sometimes by a shortening of the strophes. Both high and low quality males sang bouts with and without this decrease in the percentage performance time. Bouts that started with longer strophes and/or shorter inter-strophe pauses showed on average a more rapid decrease in the percentage performance time, and contained fewer strophes, than bouts that started with shorter strophes and/or longer inter-strophe pauses. After switching to another song type the males again used longer strophes and/or shorter inter-strophe pauses. An ‘anti-exhaustion’ hypothesis is proposed and discussed. This hypothesis gives a mainly causal explanation for the existence of song switching and song repertoires in passerine birds.