What is known about the tardigrade nervous system, and in which animal it was seen: a per-structure concordance, with the first measurements of the trunk ganglia

Two groups are now building connectomes of Hypsibius exemplaris. They will interpret their data against a literature that is mostly about other animals, and nobody has counted how much. We assembled every published statement we could find about each named part of the tardigrade nervous system, with its locator and the species it was actually observed in: 324 claims from 40 sources across 104 structures. 133 of the 317 species-tagged claims (42%) are from H. exemplaris; the rest are from Mesobiotus cf. harmsworthi (55), Halobiotus crispae (44), Macrobiotus hufelandi (36), Echiniscus testudo (25), Actinarctus doryphorus (19) and others spanning both tardigrade classes, so some of the comparisons routinely made cross more phylogenetic distance than mouse to human. 38 of the 104 structures have no H. exemplaris statement at all — including the median ganglion of the brain, the sensory cirri and clavae, and the stomatogastric ganglion — and 60 pairs of claims directly contradict one another. That the literature is skewed is not new: Persson et al. stated it in 2014 at class level. What is new is the grain, which is what makes it actionable: per structure, per species, per claim, with a locator on each, and therefore a gap list rather than a headline. Alongside the concordance we report measurements that fill some of its own gaps, all from openly licensed deposits: the four trunk ganglia of H. exemplaris measured for the first time (385, 430, 379 and 194 µm³, a total equal to the brain's), their positions relative to their own legs, the depth of nervous tissue beneath the cuticle, the measured limit that a 106 nm scan cannot resolve brain lobes, three unnamed labels in a published scan identified by registration, and the neural gene set placed on the chromosomes. We also report, and correct, the two errors we made assembling this: a published percentage that was wrong by a counting artefact, and a reading-order bias we had to measure on ourselves before we could report the headline at all. Files: the preprint; the full per-structure concordance as Markdown (human-readable) and JSON (machine-readable), built from the project knowledge graph on 6 October 2026.