Fig. 6

Evolutionary cycles of control profiles. a Examples of control profiles in reconstructed networks in three avian groups, taking into account the shared ancestry of the lineages as delineated by phylogeny (Supplementary Fig. 4), binned by 10 my (shaded areas on phylogenetic tree) over the last 45 my. Control profiles of extant species of each group are shown on the right. In all lineages, gains of dietary carotenoids and associated increase in Ns has led to greater redundancy in source controls and, correspondingly lower η_s and higher η_i or η_e in a subsequent period, followed by loss of some dietary carotenoids (Fig. 3, Supplementary Fig. 7) and repeat of the cycle. b The basis for the cycles in a is equally frequent transition between degenerative, source-dominated networks (category 3 in Fig. 1b) and networks with internal controls (category 4 in Fig. 1b). This transition was closely associated with the gain and loss of additional dietary carotenoids (Supplementary Fig. 7). Shown are the rates (line thickness) and likelihoods (shade) of evolutionary transitions between control categories. Black lines indicate highly likely transitions (zero-value rate parameters (z) in less than 1.5% of models), dark gray lines show likely transitions (1.5 < z ≤ 10% of models), and light gray shows probable transitions (10 < z < 20% of models), transitions with z > 20% are not shown. Supplementary Fig. 6 shows all results and Supplementary Fig. 1 lists sources for drawings. Cardinalis and allies include genera Cardinalis, Pheucticus, Piranga, Nesospiza, Paroaria, Coereba, Ramphocelus, and Sicalis; Carpodacus and allies include genera Carpodacus, Haematospiza, Uragus, and Euphonia; Pyrrhula and allies include genera Pyrrhula, Rhynchostruthus, Vestiaria, Mycerobas, Coccothraustes, Bucanetes, and Pinicola