Answers are grounded in this paper and cite their sources.
Space-filling models of tissue — Voronoi tessellations, power (Laguerre) diagrams, the space-filling limit of vertex models — all assume that the interface between two cells is flat and perpendicular to the line joining the cells' generators. The assumption is made for convenience and, as far as we can establish, has never been measured against a developing embryo. It is exactly testable, and the test already exists: in an unrelated literature on mechanism design, where the question of whether a partition corresponds to a power diagram decides whether a contract can truthfully elicit a belief. We bring that characterisation across, adapt it from an exact decision problem to a least-squares estimator that tolerates noisy segmentation — a case that literature never posed — calibrate it against a synthetic floor and a permutation null, and apply it to segmented Caenorhabditis elegans embryos and to a second animal, the ascidian Phallusia mammillata. Both are power-diagram-consistent through gastrulation (92–100% of contacts and 97–100% of contact area, except the 64-cell ascidian at 83% and 94%). In the ascidian neurula the model fails in specific tissues: the notochord loses three quarters of its contact area and the neural plate almost half, while the epidermis stays within about 10%, against a floor that misses nothing. The normality theorem explains why early embryos fit; the tissues that depart are where they stop being convex, normal partitions. Generators at centroids outperform generators at nuclei. Code (the estimator, both animals' pipelines and the tessellation engine) and every result file behind the tables and figures are in power-diagram-code-data.zip.
No takes yet. Share an insight, caveat, or question.
Meow-Ludo Meow-Meow (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: