Extended Data Fig. 8: Transitions and correlations between camouflage and blanching patterns.
From: The dynamics of pattern matching in camouflaging cuttlefish

a. Trajectories corresponding to all blanching trials (evoked by threatening stimulus) in three animals (sepia218: n = 22, sepia219: n = 19, sepia221: n = 6), projected in the space defined by their first three principal components. Trajectories are coloured by trial number, and the rare trials showing different start and end patterns are highlighted in magenta. Solid lines: outward trajectories towards blanched state; dashed lines: return to camouflage patterns. b. Same as in a but shown in only their two first PCs and split at their peak blanching point to separate blanching and return trajectories. Colour represents the instantaneous speed in space defined by the first 200 PCs (scale as in Fig. 5b). c. The correlation between the starting (or ending) pattern of a trial and the blanched pattern reached in another trial predicts the correlation between the starting (or ending) patterns of both trials (both pattern correlations are positively correlated). This suggests that blanched patterns carry information about the starting (and ending) camouflage pattern of a same trial. In addition (data not shown), the mean correlation coefficient (z-scored) between starting and blanching patterns of the same trial is significantly higher than the mean correlation coefficient between starting and blanching patterns of different trials, suggesting that blanching patterns depend on the camouflage pattern preceding blanching (mean ± s.e.m.: 1.29 ± 0.16 vs. −0.00 ± 0.06, P = 0.013, two-sided paired t-test, N = 3 animals). This is also true for blanching and end patterns (0.91 ± 0.19 vs −0.51 ± 0.06, P = 0.0081, two sided paired t-test, N = 3 animals). d. Aligned tanglegrams to visualize hierarchical clustering performed on start, blanching and end patterns at chromatophore resolution (mean of 10 frames per chunk) for two animals, showing that similarities that exist between patterns during camouflage are conserved during blanching (left; sepia218, start-to-blanch cophenetic corr. = 0.63, P = 0.0046, Mantel test, blanch-to-end cophenetic corr. = 0.80, P = 2e-04, Mantel test, right; sepia219, start-to-blanch cophenetic corr. = 0.26, P = 0.0047, Mantel test; blanch-to-end cophenetic corr. = 0.26, P = 0.015, Mantel test). Colours denote common subtrees in each dendrogram pair; numerical leaf labels denote trial ID. e. Heatmap of pairwise correlation coefficient between all frames of all strong blanching trials for sepia219 (n = 17 trials). Trials are sorted by pairwise correlation coefficient between their respective start and end patterns.