Extended Data Fig. 9: The visual stimulus conditions used in the electrophysiological and behavioural rigs.

a, Stellarium images of the austral starry night sky (and randomized sky) used during the spring and autumn migratory seasons in 2018 and 2019. Geographic North (gN) is upwards in each image. Randomized skies in 2018 were created by randomizing pixels (insets), whereas those in 2019 were improved by randomizing stars (see Methods). b,c, Brightness distributions of natural starry skies and randomized skies projected onto a screen above the moth in the behavioural (b) and electrophysiological rigs (c). The probe of a spectrometer (Ocean Optics QE65 PRO, see Methods), equipped with a collimating lens (25° field of view), was positioned at the same location as the moth and systematically pointed at the projected sky in different directions (0–360° in 30° steps) relative to geographic North. The radiance (photons/cm²/sec/sr) was measured from the screen in each direction in (5 measurements per position): mean radiance is shown as a function of measurement direction (green squares = spring projection, magenta squares = autumn projection, blue squares = randomized projection). Error bars give standard deviations. Note that the brightest areas of the autumn and spring projections were respectively located south-easterly and westerly. d. The spectra of the projectors used in the behavioural (left) and electrophysiological rigs (right), measured using the Ocean Optics QE65 PRO spectrometer. For the projector in the behavioural rig, a ring of UV LEDs was used to create a quasi-realistic spectrum (see Methods). Note that for this spectrum the noise level is relatively high since we used all the neutral density filters normally used during an experiment.