The data was collected in the field from 15 April to 15 May (from 27 April to 15 May for the Voorstonden location). The first column shows the 7 sites (from 1 to 7) used in the models as random effects. Site 1 corresponds to a location called ‘Lebret’s Hoeve’, site 2 to ‘Voorstonden’, site 3 to ‘Radio Kootwijk’, and site 7 to ‘Hijkerveld’. Sites 4 and 5 are adjacent spatially and correspond to a location called ‘ASK Doornspijk’. Finally, site 6 consists of two sites merged together because they overlap in a location called ‘Klaterweg’. The onsets (column 7) have therefore been obtained from the mean onsets in the two former sites for each light treatment on each day. The second column shows the 3 levels of 'Treatment', which represents the transects where the data has been collected (green, red, or white light transect). The third column shows the 'Formal date', i.e., the date at which the bird was singing (in calendar days). The fourth column shows the (relative) ‘date', which is the date relative to 15 April (0 = 15 April), and which has been used in the models as a random slope. The fifth column shows the 'onset relative to control'. This variable is calculated for each species, at each site and on each day, as the difference (in min) in the onset of dawn song between each light treatment (red/green/white) and the dark control. The sixth and last column shows the 'species' for which the onset is shown (14 levels, vernacular names shown).