Extended Data Fig. 7. Spatial distribution and stability of pattern components.

a. Examples of measurement of spatial overlap between pattern components (green and red; overlap in yellow): each column stands for a correlation coefficient of spatial density using a different bin size. Upper row: components showing high spatial overlap; lower row: components showing low spatial overlap. b. For pairs of pattern components, correlation of activity and spatial overlap do not correlate (N = 435 component pairs defined in single trajectories between camouflage patterns). Purple dot: pair in upper row in a; black dot: pair in lower row in a. c. Absence of correlation between the pairwise distance between pattern components (measured as averaged physical distance between pairs of chromatophores) and the correlation of their mean activities during repeated transitions. N indicates the number of repeated trials (trajectories) included in each analysis (Same N values for d-e). High dot density near 20 mm is explained by the half-width of the animals and the left-right symmetry of the pattern component pairs. d. Measurements as in c, using Wasserstein distance as a metric for distance between pattern components. e. Relationship between the dissimilarity of two transitions (measured as pairwise distance in 200 PCs, Methods) and the proportion of chromatophores that remain in the same component across those two transitions. f. Comparison as in e, using Wallace distance as a metric for clustering similarity. Distance was normalized by the s.d. of all dataset of each animal. Pearson’s r = −0.382, p = 2.9e-100. g. Comparison as in e, using Adjusted-Rand-Index as a metric for clustering similarity. Pearson’s r = −0.436, p = 4.3e-133.