The hypothesis is advanced that the circadian pacemaker in the mammalian suprachiasmatic nucleus (SCN) is composed at the molecular level of a nonredundant double complex of circadian genes (per1, cry1, and per2, cry2). Each one of these sets would be sufficient for the maintenance of endogenous rhythmicity and thus constitute an oscillator. Each would have slightly different temporal dynamics and light responses. The per1/cry1 oscillator is accelerated by light and decelerated by darkness and thereby tracks dawn when day length changes. The per2 /cry2 oscillator is decelerated by light and accelerated by darkness and thereby tracks dusk. These M (morning) and E (evening) oscillators would give rise to the SCN's neuronal activity in an M and an E component. Suppression of behavioral activity by SCN activity in nocturnal mammals would give rise to adaptive tuning of the endogenous behavioral program to day length. The proposition-which is a specification of Pittendrigh and Daan's E-M oscillator model-yields specific nonintuitive predictions amenable to experimental testing in animals with mutations of circadian genes.