Eoandromeda octobrachiata is a poorly understood Ediacaran organism, with spiral-octoradial arms, found in South Australia and South China. The informal Nilpena member of the Rawnsley Quartzite, Flinders Ranges in South Australia preserves more than 200 specimens of Eoandromeda. Here we use the novel application of rotational geometric morphometrics together with palaeoenvironmental information to provide a better insight into their palaeobiology and ecology and to address conflicting hypotheses regarding mode of life and taxonomic affinity. We find that Eoandromeda likely had a radially symmetrical shape in life, was cone-shaped and had a high relief off the microbial mat. Analysis of the symmetric and asymmetric shape components revealed they deform strongly in the direction of palaeocurrent, therefore are thought to be made of a flexible material. Almost all specimens are compressed flat. Specimens that appear to have not fully collapsed support the idea that Eoandromeda was likely cone-shaped and further suggest that they possibly collapsed spirally. Our shape analysis along with observed morphological features support the benthic mode of life hypothesis rather than pelagic. Morphological and ecological inconsistencies do not fully support the hypothesis of a Ctenophora taxonomic affinity.

R and excel required to view and process the files. Funding provided by: Australian Research CouncilCrossref Funder Registry ID: https://ror.org/05mmh0f86Award Number: DP220102772 Funding provided by: NASA Exoplanet Science InstituteCrossref Funder Registry ID: https://ror.org/01trfvq12Award Number: 18-EX018-0096 Funding provided by: NASA Exoplanet Science InstituteCrossref Funder Registry ID: https://ror.org/01trfvq12Award Number: 20-EXO20-0009

Three-dimensional (3D) surfaces of these specimens were captured using the HDI Compact C506 and 3Shape TROIS 4 Intraoral 3D laser scanners (accuracies reported to 12µm and c.5µm respectively). Scans were prepared using the FlexScan3D and TRIOS Design Studio software (correspondingly) and processed (landmarked/measured) in the MeshLab software (Cignoni et al. 2008).