Daylength, or photoperiod, is perceived as a seasonal signal for the control of flowering of many plants. The measurement of daylength is thought to be mediated through the interaction of phototransduction pathways with a circadian rhythm, so that flowering is induced (in long-day plants) or repressed (in short-day plants) when light coincides with a sensitive phase of the circadian cycle. To test this hypothesis in the facultative long-day plant, Arabidopsis thaliana, we used varying, non-24-hr light/dark cycles to alter the timing of circadian rhythms of gene expression relative to dawn and dusk. Effects on circadian rhythms were correlated with those on flowering times. We show that conditions that displaced subjective night events, such as expression of the flowering time regulator CONSTANS into the light portion of the cycle, were perceived as longer days. This work demonstrates that the perception of daylength in Arabidopsis relies on adjustments of the phase angle of circadian rhythms relative to the light/dark cycle, rather than on the measurement of the absolute duration of light and darkness.