In humans, short-wavelength light evokes larger circadian responses than longer wavelengths [1-3]. This reflects the fact that melanopsin, a key contributor to circadian assessments of light intensity, most efficiently captures photons around 480nm [4-8] and gives rise to the popular view that ‘blue’ light exerts the strongest effects on the clock. However, in the natural world, there is often no direct correlation between perceived colour (as reported by the cone-based visual system) and melanopsin excitation. Accordingly, while the mammalian clock does receive cone-based chromatic signals , the influence of colour on circadian responses to light remains unclear. Here, we define the nature and functional significance of colour influences the mouse circadian system. Using polychromatic lighting and mice with altered cone spectral sensitivity (Opn1mwR), we generate conditions that differ in colour (i.e. ratio of L- to S-cone opsin activation) whilst providing identical melanopsin and rod activation. When biased towards S-opsin activation (appearing ‘blue’), these stimuli reliably produce weaker circadian behavioural responses than those favouring L-opsin (‘yellow’). This influence of colour (which is absent in animals lacking cone phototransduction; Cnga3-/-) aligns with natural changes in spectral composition over twilight, where decreasing solar angle is accompanied by a strong blue-shift [9-11]. Accordingly, we find that naturalistic colour changes support circadian alignment when environmental conditions render diurnal variations in light intensity weak/ambiguous sources of timing information. Our data thus establish how colour contributes to circadian entrainment in mammals and provide important new insight to inform the design of lighting environments that benefit health.