The Winter Coast outcrops (Ust′-Pinega Formation) contain a rich ‘Ediacara-type’ biota9,10. Kimberella occurs in several facies all through the section, but most specimens have come from the undersides of fine-grained sandstone gutter casts, in member 9 of the local stratigraphic section10. Channels in a clay substrate were colonized by a rich biota; when the channels were rapidly infilled with sand, the biota was preserved in situ11. Associated with Kimberella in member 9 are large round ‘medusoids’, the distinctive fossils Tribrachidium and Dickinsonia, and some undescribed forms. Small round fecal pellets, meandering trace fossils, and small pyritized algae also occur11.

Kimberella is an oval-to-pear-shaped, bilaterally symmetrical fossil, with several zones arranged concentrically. Fossils range from 3 to 105 mm in length; the largest specimen is incomplete (Fig. 1c) and may have been 130–140 mm long in life. The size range is continuous; no discontinuities exist in size or structure that would suggest sexual dimorphism or a size-structured population. Nor is there striking ontogenetic change over the observed size range; the smallest specimens are near-miniatures of the large ones. Films of framboidal pyrite on some specimens resulted from anaerobic bacterial activity after burial.

Figure 1: Specimens of Kimberella from Ust′-Pinega Formation, Winter Coast of White Sea, all coated with NH4Cl.
figure 1

All specimens are curated at the Paleontological Institute (PIN), Russian Academy of Sciences, Moscow. All are preserved on undersides of sandstone channel casts in member 9 unless otherwise noted. All scale bars are 1 cm. a, Large, well preserved specimen, with deep central depression. Note folding of soft tissues away from shell edge, which has buckled (arrow) (PIN 3993/4003). b, Specimen partially folded over, demonstrating deformation of soft parts and bilateral symmetry of organism (PIN 3993/4026). c, Largest specimen known; note lateral displacement of shell, with crenellations extending out from under shell edge at right (PIN 3993/4004) (d). Specimen from thinly interbedded argillite and siltstone, member 1 of local section; arrow indicates position of anterior bulge (PIN 3993/4001). e, One of two specimens showing series of pits (spicules?) along proximal ridge (PIN 3993/4036). f, Small but well preserved specimen showing primary features (PIN 3993/4033). g, Specimen showing rotation and displacement of shell from the main axis (PIN 3993/4027). h, Poorly preserved specimen from talus, probably from sandstones of member 10 of local section; note continuation of zones around posterior (pointed) end, as well as possible buccal mass indicated by arrow (PIN 3993/4009).

The distal portions of the fossils demonstrate a relatively inflexible outer margin, and an inner zone that is often distorted or folded over but never cracked (for example, see Fig. 1a, b, e). Within the inner zone is a row of crenellated structures, typically about 30 in number. These usually form a U-shaped band, but extend completely around the fossil in a few specimens. A thin linear structure, the proximal ridge, separates the crenellated zone from the interior; this ridge seems to be an overprint of a higher anatomical structure onto the imprint of the ventral surface of the organism. Rare specimens show a band of resistant point-like structures associated with the proximal ridge (Fig. 1e). The central part of the organism is preserved in negative relief. Its position and depth vary, depending on how much the organism was compacted and deformed; it may extend up to 20 mm into the host rock in large specimens. It may be compacted to be almost flat (Fig. 1f) or displaced with respect to the softer parts (Fig. 1c), but is rarely found bent laterally. Therefore, its original substance was comparatively firm12. Numerous fine striae run laterally from the medial depression in many specimens; they are generally clearest in the most flattened specimens. At least three specimens show a small bulge near the broader end, on the midline of the body, inside the proximal ridge (Fig. 1d, h).

Kimberella was first formally described as a cubozoan medusa preserved on its side2. The crenellated zones were interpreted as gonads or as muscle bands, and the organism was reconstructed as having tetraradial symmetry. However, none of our specimens shows anything other than bilateral symmetry. If Kimberella were tetraradially symmetrical, we would expect to find more than two crenellated zones in at least a few specimens. Even folded specimens show no more than two crenellated zones (see Fig. 1bfor example). Nor was Kimberella triradially symmetrical13; the lateral zones are not repeated in the central zone. Kimberella, preserved in negative relief, was a ‘resistant’ organism; extant cubozoans are delicate and, if preserved at all, would form fossils in positive hyporelief, like most ‘medusoids’12. We cannot confirm the presence of gonads, tentacles and rhopalia5; the best-preserved specimens lack any trace of tentacles or rhopalia. The crenellated structures, formerly interpreted as gonads, formed a continuous band around the body, unlike the discrete gonads of cubozoans. Furthermore, Kimberella is common in member 9, which otherwise includes definitely benthic forms almost exclusively; there is no evidence of subaerial exposure or mass stranding at this level. Ediacaran fossils of planktonic organisms typically appear in different facies from benthic forms7,10. We conclude that Kimberella was a benthic bilaterian, not a medusa.

The key to interpreting this fossil is the fact that the organism consisted of both firm and soft parts, and that compaction during decay ‘overprinted’ structures that in life were at different levels in the organism. In one specimen (Fig. 1g), both the central depression and the outermost ridge have been slightly rotated and moved forward together over the softer parts. Another (Fig. 1a) shows the softer ‘crenellated structures’ pulled away from the outer ridge, and another (Fig. 1c) shows them protruding beyond the outer ridge. Specimens like these show that the firmer parts formed a discrete anatomical unit, partially separable from the soft parts. We therefore interpret the central depression and outermost zone of Kimberella as parts of a thin oval shell, not mineralized but fairly stiff, at least in its central part. This shell lent the animal strength during burial and sediment compaction, at which time softer tissues were deformed and compressed by unlithified sediment entering the shell from below, and finally overprinted on the overlying sand.

The crenellated structures are often deformed and were clearly soft; we interpret them as lateral folds of the body, usually retracted under the shell but occasionally seen protruding (Fig. 1c). The frequent U-shape of the crenellated zone may be explained by the asymmetry of the anterio–posterior outline of the shell (the highest point of the shell was closer to the anterior). The crenellations may have had a respiratory function. Although their number does not increase significantly with organism size, except in the largest specimens, the folds are small or even absent in small specimens, and longer and deeper in larger specimens. Assessing the respiratory function of the crenellations is difficult, because compaction makes it hard to estimate the volume of the organisms. However, our estimates of the total surface area of the crenellations scale isometrically with the product of body length and width. Kimberella may also have hosted microbial symbionts within the crenellations, as with the cerata or parapodia of symbiont-bearing gastropods14; symbiosis has often been proposed as a lifestyle for Ediacaran organisms13,15. Some specimens show a broad flat region between the crenellated zones, which we interpret as a creeping foot. There is no sign of true segmentation of the body, but fine lines running from the midline to the foot in some specimens may be strands of dorsoventral musculature. The anterior bulge, identified in the Australian material as the ‘stomach’, may be the mouth and its associated muscles forming a buccal mass. There is no sign of a true radula, but radulae are very rarely preserved in any fossil molluscs.

Figure 2shows our reconstruction of Kimberella. The basic architecture of the fossil, showing metamerism without observable segmentation, a shell and a broad flattened foot, is closely comparable to that of primitive molluscs. The arrangement and inferred function of the crenellations is like that of the ctenidia of chitons and the ‘ventilatory flaps’ of monoplacophorans, although as similar structures have evolved independently at least three times in the Mollusca16, the question of homology is not straightforward. The anatomy of Kimberella is also comparable to the soft anatomy of the Cambrian halkieriids, which appear to be relatives of molluscs and possibly other protostomes. Halkieria has a similar shape and division of the body into central and axial zones. It also has a zone of faint striae around the sole of its creeping foot, similar to the crenellated zone of Kimberella and plausibly having a respiratory function17,18. It is even possible that the resistant pointlike structures seen in rare specimens of Kimberella (Fig. 1e) are impressions left by sclerites or spicules like those of halkieriids and some molluscs, although no actual sclerites have yet been recovered. More material will be needed to test this hypothesis.

Figure 2: Reconstructions of Kimberella.
figure 2

a, Dorsal view. c, Crenellated zone; l, lobe; s, striae; dr, distal ridge; pr, proximal ridge; m, medial depression; a, anterior knoll. b, View of living organism, with folds of ‘crenellated zone’ extended beyond margin of shell; the folds would usually have been retracted under the shell at the time of burial.

We cannot prove a molluscan affinity conclusively; some definitive molluscan synapomorphies, like the radula, cannot be seen, and others are open to other interpretations. However, no other extant metazoans resemble Kimberella as closely. Metamerism is consistent with a flatworm grade of organization, but the firm parts and complex anatomy of Kimberella tend to rule out a relationship to the Platyhelminthes. Kimberella somewhat resembles certain pelagic salps (phylum Urochordata), which have an oval shape, a tough thickened outer test, and serial muscle bands that could resemble the crenellations of Kimberella19. However, not only was Kimberella benthic, it lacks the atrial and branchial openings and the distinctive budding growth pattern of salps. The benthic habitat and absence of tentacles or obvious zooids rule out an affinity with siphonophores or chondrophorines. Kimberella superficially resembles ctenophores such as Beroë, but its resistant nature and lack of octamerous symmetry rule out a ctenophore affinity. Nor can we interpret Kimberella as a non-metazoan ‘Vendobiontan’20, protist21, or lichen22. It is always possible to stretch such hypotheses to cover almost any fossil imaginable, but we can find no definite features that would place Kimberella in any of these groups. Kimberella has nothing like the ‘quilted pneu’ morphology which has been used to argue against metazoan affinities for other Ediacaran fossils19, and its complex structural properties are not well explained by protist or lichen models.

We conclude that Kimberella is a bilaterian metazoan, more complex than a flatworm, more like a mollusc than like any other metazoan, and plausibly bearing molluscan synapomorphies such as a shell and a foot. This interpretation counters assertions that the Ediacara biota represents an extinct grade of non-metazoan life. It confirms hypotheses based on trace fossils that metazoan triploblastic lineages, including ‘molluscan-grade bilaterians’, began to diversify before the beginning of the Cambrian23. It also suggests a pre-Ediacaran origin of major metazoan clades.