The brain and nerve cord

A nervous system, in eight movements → — scroll, and this animal's brain is found, measured, taken apart and switched on. Every number measured from the scan or the transcriptome; four of them never measured by anyone before.

We cut the nerve cord of the glass tardigrade into its four ganglia and the connectives between them, measured each one, found where each ganglion sits against its legs, and drew what the literature knows about the rest into one map you can click. The brain is 1,389 µm³, about the volume of a 14 µm bead, and the four ganglia together come to the same. The whole nervous system probably holds a few hundred neurons, and nobody has mapped how they connect yet.

What is known about each part of the nervous system — every published statement, with its page and the species it was actually seen in, which three times out of four is not the laboratory animal. Including the 79 places the papers contradict each other.

How the tardigrade eye is wired — four cells, none of them with an axon, and five neighbours that carry the signal out. A question asked in 2013 and answered in 2026, with the part that is still untraced drawn as untraced.

How a tardigrade senses its world — seven ways a small animal could sense its surroundings and what is known about each. Four of the seven are blank, including the one that should exist: nothing we hold describes how the animal detects that it is drying out.

Inside the glass tardigrade

We placed the brain from the head scan into the whole animal, and it sits where the literature puts it: dorsal, in front of the pharynx, above the buccal tube. It measures 1,389 µm³, against the 1,393 in the scan's original paper, so we found the right label and the right voxel size (how).

Behind it the whole-body scan holds the nerve cord as one unbroken strand. On 20 September 2026 we cut that strand into its parts, so the viewer now carries eleven nervous structures instead of three: the brain, four trunk ganglia, the paired connectives between them, and, from the head scan, the outer connectives that link the brain to the first ganglion. The cut is made where the cord's cross-section passes 10 µm². The connectives run at 2.5–6 µm² and the ganglia peak at 47–55, so anything in between separates them; the volumes move by under 10% for any cut from 8 to 25 µm².

Registering the head scan settles two labels the paper leaves unnamed. The 404 µm³ mass beneath the pharynx is trunk ganglion 1: 86% of it lands inside the cord's first swelling, so it is that ganglion seen at 106 nm, not a separate ganglion under the pharynx, a structure some authors report and others deny (see by question above). A pair of thin strands leaving its back is the connectives to ganglion 2 (77% lands there). Two more pairs run forward from it towards the brain: an outer pair, ±8–9 µm either side of the midline, touching both brain and ganglion, which is where the paper says it segmented the outer connectives; and a shorter, more medial pair that stops 2.7 µm short of the brain, lying where the inner connectives run — the connectives the paper says it could not identify. We mark that pair probable, not certain.

Where the ganglia sit

We measured each ganglion's position along the body from the label volume and compared it with the claw glands at the tip of its legs. Ganglion 1 sits 4.2 µm in front of its legs, ganglion 2 10.3 µm and ganglion 3 12.7 µm. (Leg IV points backwards, so its claws give no comparable mark.) That is the pattern Martin et al. 2017 drew for adults, a small shift for ganglion 1 and a larger one behind it, now measured in H. exemplaris. Embryos at 38–50 hours have their ganglia level with the limb buds (Gabriel et al. 2007, p. 554), so the shift happens later.

How big each ganglion is

Cutting the cord where its cross-section passes 10 µm² gives 385, 430, 379 and 194 µm³ from front to back, with 214 µm³ of connectives between them. Together the four ganglia come to 1,386 µm³: the same volume as the brain, to within the width of a voxel. Ganglion 4 is half the size of the rest. Ganglion 2 is the largest, by a little. Nobody has published these numbers — the paper that made the scan says in its discussion that it did not measure the trunk ganglia.

Ganglion 4 being the smallest matches the Macrobiotus nuclei counts and our own embryo. The rest does not: Zantke's counts make ganglia 2 and 3 about a quarter smaller than ganglion 1, and here they are not. The published counts already disagree with each other — one species shows four equal ganglia — and volume is not cell number. Our embryo, which was given no cell numbers, arrives at a front-heavy pattern on its own (Development).

How close to the skin, and what the scan cannot show

The paper says every major neural structure lies next to the epidermis. We put a number on it: measured from the traced body surface, the median depth is 2.4 µm for the connectives, 3.2–3.8 µm for the four ganglia and 3.6 µm for the brain, against 5.2 µm for the storage cells, 9.1 µm for the ovary and 11.6 µm for the midgut. No neural voxel lies more than 8.2 µm in. The nervous system of this animal is pressed against its own skin.

What the scan will not give is the inside of the brain. The label is one compact mass, 24.4 µm across and 18.3 µm along the body, and the largest sphere that fits inside it has a radius of 2.6 µm. Eroding it never splits it into lobes, only into two symmetrical tips 10 µm either side of the midline. Its X-ray density is flat (a 15% spread, with no trend from surface to core), so the neuropil and the layer of cell bodies around it cannot be told apart. The brain is at least properly two-sided: about its best mirror plane the halves measure 688 and 700 µm³, 1.7% apart. Everything in the map at the top of this page — lobes, commissures, named nerves — comes from confocal and histological work on other species, and none of it can be placed in this animal from these data. That is the gap an electron-microscopy volume would close.

More in the figure library

How many neurons

Nobody has counted the neurons of H. exemplaris: "the exact number of neurons in any tardigrade life-history stage remains unknown" (Lyons & Kato 2025, p. 3). The counts that exist come from related species, and they add up to a consistent picture.

Brain plus ganglia in the two Macrobiotus-group species comes to roughly 350–400 nuclei, some of them glia. That fits "a few hundred". TardiMap's larger figure may include neurons outside the brain and ganglia (in the legs, the gut and the body wall); the site doesn't say.

A check from our volume. Single tardigrade cell bodies measure 3–5 µm across in confocal images (Martin et al. 2017, p. 574). Two hundred 3 µm cell bodies would fill 2,800 µm³, twice the 1,389 µm³ brain we measured, before any room for fibres. Either the H. exemplaris brain holds fewer cells than the Mesobiotus count, or its cell bodies are smaller: its nuclei are about 2 µm in the ganglia of related species (Mayer et al. 2013, p. 3). Our animal was also dried for scanning, which shrinks tissue. A direct count from TardiMap's electron microscopy will settle it.

Is the number fixed? After hatching, dividing cells take up the label EdU only in storage cells, germ cells and the gut, never in the brain or ganglia (Quiroga-Artigas & Moriel-Carretero 2024, Results). A hatchling seems to carry the neurons it will keep. Whether every animal gets the same number, as a worm does, is still open (Lyons & Kato 2025).

Eyes and light

H. exemplaris has two eyespots, one on each outer brain lobe. Each is an inverse pigment cup, a simple ocellus (Martin et al. 2017, pp. 582–583, with a picture of this species' eyes). The eyes are easy to see in live animals and dissolve in the usual mounting medium, so preserved specimens often lack them (Gąsiorek et al. 2018, p. 55). They appear in the embryo between 40 and 50 hours, with the claws and stylets (Gabriel et al. 2007, p. 554).

The genome carries one visual rhabdomeric opsin, as velvet worms do (Hering & Mayer 2014, via Martin et al. 2017, p. 583). The velvet worm's single visual opsin sits only in its photoreceptors, and its eye sees one colour, peaking at 474 nm (Beckmann et al. 2015, p. 916). From that we infer, with low confidence, that the tardigrade eye is also monochromatic. Its peak wavelength can't be carried over. A 2026 preprint finds an extensive expansion of opsin genes and, as its title says, a "non-cerebral origin of the minimalist eye"; the spectral properties of the eye's ciliary opsins "remain unknown" (Dutta et al. 2026, preprint, p. 10). Opsins are also switched on in eggs, where their job is unknown (Fleming et al. 2021, p. 8).

How light is turned into steering, the tardigrade's phototaxis, "remain[s] unknown" (Lyons & Kato 2025, p. 13). The old experiments disagree: some animals moved towards light (von Erlanger, with a "H. dujardini" that predates the lab strain), others away from it (Marcus, with Dactylobiotus dispar), and others depending on how bright it was. Young Hypsibius convergens were drawn to light only until their second moult (Ramazzotti & Maucci 1983, p. 63). That last one is easy to test in H. exemplaris hatchlings.

How it develops

In our simulated embryo four ganglia gather under the belly from 40 hours, the fourth a few hours after the others. That delay is our assumption. The published timeline for H. exemplaris, in hours after laying:

• 24 hThe head-patterning gene otd is on across the whole head (Smith et al. 2016, p. 225). Brain-patterning genes (six3, otd, pax6, unplugged, pax2/5/8, and the neuron marker elav) switch on in a staggered front-to-back order at 24, 35 and 45 h (Smith et al. 2018).
• 25–33 h, another speciesIn Thulinius stephaniae each trunk ganglion is founded by one neural progenitor cell, made by two divisions whose sister cells become skin. Brain progenitors leave the surface first (Hejnol & Schnabel 2005, p. 1353).
• 29–71 hNotch and Delta, the signalling pair that picks neurons out of a sheet of cells, are co-expressed and peak across the window of neurogenesis (Li et al. 2024, p. 8, on their time scale).
• 35 hexd3 marks the developing central nervous system, and Sp6-9A the developing ganglia (Game & Smith 2020, p. 3; Game & Smith 2025, p. 6).
• ~40 hAxons start growing, in the brain first and then in the trunk ganglia from front to back, and finish 5–10 hours later (Gross & Mayer 2015). Ganglia 1–3 appear together and ganglion 4 in the next panel of their sequence (Martin et al. 2017, Fig. 9).
• 38–50 hFour clusters of cells are visible along the belly, level with the limb buds (Gabriel et al. 2007, p. 554). The whole embryo's gene expression is at its most evolutionarily conserved across this stage (Li et al. 2024, p. 3).
• hatching onwardNo new cells are added to the brain or ganglia (Quiroga-Artigas & Moriel-Carretero 2024).

Some brain cells steer the whole animal's growth: a few cells at the tips of the inner lobes make the moult-inhibiting hormone Chh/mih and are a candidate control centre for moulting (Yamakawa & Hejnol 2024, p. 5807). See the life cycle.

What it tells us about evolution

Tardigrades sit beside velvet worms and arthropods, and their nervous system is the best evidence for how the three are related. Toggle the head hypotheses under by question in the map to see each reading coloured.

How many segments make the head?

Even the three-segment side disagrees with itself: Persson et al. 2012's table places the older labels protocerebrum and deutocerebrum on the inner and outer lobes, the reverse of its text. Hox expression adds a surprise either way: aligned by their Hox genes, the tardigrade's whole body corresponds mainly to the arthropod head (Smith et al. 2016, p. 225).

Were ganglia invented once or twice?

Velvet worms have no ganglia, just two even cords with nerves to the legs (Martin et al. 2017, p. 570). If tardigrades are the arthropods' sister group, ganglia arose once in their shared ancestor; if tardigrades branched off first, twice (Martin et al. 2017, pp. 573–574). Tardigrade ganglia sit in front of their legs as arthropod ganglia do, which Mayer & Whitington 2009 (p. 272) take as a sign they are the same invention. Two leg nerves per leg, found in tardigrades and velvet worms but not arthropods, are probably the ancestral state (Martin et al. 2017, p. 572; Mayer & Harzsch 2008, p. 1205).

Scale

A 3 cm velvet worm's brain may hold hundreds of thousands of neurons (Strausfeld et al. 2006, p. 170). A whole tardigrade brain would fit inside the cell body of a single 50 µm locust neuron (Martin et al. 2017, p. 574). Our volume puts a number on it: 47 tardigrade brains to one locust cell body.

Mapping every connection

A connectome is the full wiring diagram: every neuron and every synapse. The worm C. elegans has one for both sexes; the hermaphrodite's is 302 neurons with 4,887 chemical and 1,447 electrical connections (Cook et al. 2019, p. 63). A tardigrade has about the same number of neurons and legs to coordinate, which is why it is proposed as the next small animal to map (Lyons & Kato 2025). The TardiMap consortium is imaging whole H. exemplaris by electron microscopy at 8 nm: one animal ("Riker") is imaged and being reconstructed, a second ("Quark") imaged, with the data to be released openly.

We sized the job from our CT:

• Data. The whole 137,068 µm³ body at 8 nm voxels is about 270 billion voxels, a quarter of a teravoxel. The brain and nerve cord alone, about 3,050 µm³, are 6 billion. For comparison, eight C. elegans brains through development came to 5 teravoxels (Witvliet et al. 2021, Methods). A living animal is larger than our dried one, so these are lower bounds.
• Tracing. In the browser tool webKnossos, trained annotators trace axons at 1.5 mm an hour (Boergens et al. 2017, p. 692). If each of 600 neurons had 150 µm of fibre, about one body length (our guess, not a measurement), one pass would take 60 hours, and the usual three- to six-fold redundancy 180–360 hours, before synapses. Proofreading the worm brains took more than 1,120 person-hours (Witvliet et al. 2021).
• Animals. One map is not the map. Across genetically identical worms about 43% of connections vary (Witvliet et al. 2021, p. 258), and about ten worms would be needed to separate the fixed core from the variable rest (Brittin et al. 2021, p. 107). TardiMap's two animals, with a third planned, give the first comparison.
• Checks. Neurons identified by staining give fixed points to find in the EM: 14 serotonin-like and about 26 RFamide-like brain neurons, and two RFamide swellings at the core of every trunk ganglion (Mayer et al. 2013, p. 3). Our walking data give behaviour for the circuit to explain.

Where the open data are

Open questions

We collect every question a source leaves open. There are 49 about the nervous system and senses, each quoted with its page: how many neurons, what each ganglion's commissures look like, how the eyes are wired, how the brain coordinates eight legs. Read them on Questions.

Sources

Numbers from other species are labelled with the species. Hypsibius dujardini in papers before 2018 is the same laboratory strain as H. exemplaris. Page numbers point to the passage each statement comes from.

• Beckmann H et al. 2015. Spectral sensitivity in Onychophora (velvet worms). J Exp Biol 218:915. doi:10.1242/jeb.116780
• Boergens KM et al. 2017. webKnossos: efficient online 3D data annotation for connectomics. Nat Methods. doi:10.1038/nmeth.4331
• Brittin CA et al. 2021. A multi-scale brain map derived from whole-brain volumetric reconstructions. Nature. doi:10.1038/s41586-021-03284-x
• Cook SJ et al. 2019. Whole-animal connectomes of both Caenorhabditis elegans sexes. Nature. doi:10.1038/s41586-019-1352-7
• Dutta S et al. 2025. Pigment-dispersing factor neuropeptides act as multifunctional hormones and modulators in tardigrades. Open Biol. doi:10.1098/rsob.240242
• Dutta S et al. 2026. Extensive opsin gene expansion and non-cerebral origin of the minimalist eye in a model tardigrade. bioRxiv preprint. doi:10.64898/2026.04.15.718617
• Fleming JF, Pisani D, Arakawa K 2021. New tardigrade opsins and differential expression analyses show ontogenic variation in light perception. Genome Biol Evol. doi:10.1093/gbe/evab164
• Gabriel WN et al. 2007. The tardigrade Hypsibius dujardini, a new model for studying the evolution of development. Dev Biol. doi:10.1016/j.ydbio.2007.09.055
• Game M, Smith FW 2020. Loss of intermediate regions of perpendicular body axes contributed to miniaturization of tardigrades. Proc R Soc B. doi:10.1098/rspb.2020.1135
• Game M, Smith FW 2025. An Sp6-9 ortholog is expressed in developing legs in the tardigrade Hypsibius exemplaris. Arthropod Struct Dev. doi:10.1016/j.asd.2025.101471
• Gąsiorek P et al. 2018. An integrative redescription of Hypsibius dujardini. Zootaxa 4415. doi:10.11646/zootaxa.4415.1.2
• Gross V et al. 2019. X-ray imaging of a water bear offers a new look at tardigrade internal anatomy. Zool Lett 5:14. doi:10.1186/s40851-019-0130-6 (CC BY 4.0; the CT behind our measurements)
• Gross V, Mayer G 2015. Neural development in the tardigrade Hypsibius dujardini based on anti-acetylated α-tubulin immunolabeling. EvoDevo. doi:10.1186/s13227-015-0008-4
• Hejnol A, Schnabel R 2005. The eutardigrade Thulinia stephaniae has an indeterminate development and the potential to regulate early blastomere ablations. Development. doi:10.1242/dev.01701
• Hering L, Mayer G 2014. Analysis of the opsin repertoire in the tardigrade Hypsibius dujardini. Genome Biol Evol. doi:10.1093/gbe/evu193
• Li C et al. 2024. Conserved and specific gene expression patterns in the embryonic development of tardigrades. Evol Dev. doi:10.1111/ede.12476
• Lyons AM, Kato S 2025. The tardigrade as an emerging model organism for systems neuroscience. arXiv:2501.06606
• Martin C et al. 2017. The nervous and visual systems of onychophorans and tardigrades: learning about arthropod evolution from their closest relatives. J Comp Physiol A. doi:10.1007/s00359-017-1186-4
• Mayer G et al. 2013. Neural markers reveal a one-segmented head in tardigrades (water bears). PLoS ONE. doi:10.1371/journal.pone.0059090
• Mayer G, Harzsch S 2008. Distribution of serotonin in the trunk of Metaperipatus blainvillei (Onychophora). J Comp Neurol. doi:10.1002/cne.21603
• Mayer G, Whitington PM 2009. Neural development in Onychophora (velvet worms) suggests a step-wise evolution of segmentation in the nervous system of Panarthropoda. Dev Biol. doi:10.1016/j.ydbio.2009.08.011
• Møbjerg N et al. 2018. Morphology and functional anatomy. In Water Bears: The Biology of Tardigrades. doi:10.1007/978-3-319-95702-9_2
• Persson DK et al. 2012. Neuroanatomy of Halobiotus crispae: tardigrade brain structure supports the clade Panarthropoda. J Morphol. doi:10.1002/jmor.20054
• Persson DK et al. 2014. Brain anatomy of the marine tardigrade Actinarctus doryphorus. J Morphol 275:173. doi:10.1002/jmor.20207
• Quiroga-Artigas G, Moriel-Carretero M 2024. Storage cell proliferation during somatic growth establishes that tardigrades are not eutelic organisms. Biol Open. doi:10.1242/bio.060299
• Smith FW et al. 2016. The compact body plan of tardigrades evolved by the loss of a large body region. Curr Biol. doi:10.1016/j.cub.2015.11.059
• Smith FW, Cumming M, Goldstein B 2018. Analyses of nervous system patterning genes in the tardigrade Hypsibius exemplaris illuminate the evolution of panarthropod brains. EvoDevo. doi:10.1186/s13227-018-0106-1
• Smith FW, Goldstein B 2017. Segmentation in Tardigrada and diversification of segmental patterns in Panarthropoda. Arthropod Struct Dev. doi:10.1016/j.asd.2016.10.005
• Strausfeld NJ et al. 2006. The organization and evolutionary implications of neuropils and their neurons in the brain of the onychophoran Euperipatoides rowelli. Arthropod Struct Dev. doi:10.1016/j.asd.2006.06.002
• TardiMap, Tardigrade Connectome Consortium. tardimap.org, read 17 September 2026.
• Witvliet D et al. 2021. Connectomes across development reveal principles of brain maturation. Nature. doi:10.1038/s41586-021-03778-8
• Yamakawa S, Hejnol A 2024. Ecdysteroid-dependent molting in tardigrades. Curr Biol. doi:10.1016/j.cub.2024.10.054
• Zantke J, Wolff C, Scholtz G 2008. Three-dimensional reconstruction of the central nervous system of Macrobiotus hufelandi. Zoomorphology. doi:10.1007/s00435-007-0045-1

Cited through the Martin et al. 2017 review, not read by us directly: Marcus 1929, Kristensen 1983, Greven 2007. The map is our own drawing, not traced from any figure; the measurements and the embryo are ours (CC BY 4.0).