Extended Data Fig. 10: Return from blanching to camouflage: identification of pattern components.

a-c. Heatmaps of chromatophore size (normalized min-to-max expansion in colour scale) during blanching trials (looming stimulus) in three animals (a-c). Chromatophore size was min-max normalized for each chromatophore across all trials, and chromatophores are ranked by time of threshold-crossing during the return from blanching. Only chromatophores whose size change was significant (Methods) are displayed. Horizontal scale bars: 2 sec. d-f. Mantle of cuttlefish in a-c, with chromatophores colour-coded by rank of recruitment time during the return phase in the trials in a-c. g-i. Components identified from Leiden community detection plotted on the mean pattern of each animal. g: sepia218; 19,313 chromatophores; 6 components. h: sepia219; 15,468 chromatophores; 6 components. i: sepia221; 37,238 chromatophores; 7 components. j-l. Chromatophore rank distribution by component shown in g-i (for the two trials in a-c). m-o. Chromatophore mean-rank distribution by component from g-i (from all trials). Shading: bin s.d. (see Methods). m: sepia218; n = 11 trials. n: sepia219; n = 17 trials. o: sepia221; n = 4 trials. Kruskal-Wallis test on the component chromatophore-averaged mean rank indicates that at least one distribution is significantly different (sepia218: H = 38.1, P = 3.5x10⁻⁷, sepia219: H = 67.3, P = 3.7x10⁻¹³,  sepia221: H = 18.5, P = 0.005). Post-hoc multiple hierarchical permutation tests on the full nested datasets for each pair of components indicate that most component distributions are significantly different (sepia218: all pairs P < 0.005; sepia219: all pairs P = 0.001; sepia221: P = 0.12, 0.16 for components 1 vs 6, 2 vs 3 respectively, other pairs P ≤ 0.005).