The Middle Triassic was a time of major changes in tetrapod faunas worldwide, but the fossil record for this interval is largely obscure for terrestrial faunas. This poses a severe limitation to our understanding on the earliest stages of the diversification of lineages representing some of the most diverse faunas in the world today, such as lepidosauromorphs (e.g., lizards and tuataras). Here, we report a tiny new lepidosauromorph from the Middle Triassic from Vellberg (Germany), which combines a mosaic of features from both early evolving squamates and rhynchocephalians, such as the simultaneous occurrence of a splenial bone and partial development of acrodonty. Phylogenetic analyses applying different optimality criteria, and combined morphological and molecular data, consistently recover the new taxon as a stem-lepidosauromorph, implying stem-lepidosauromorph species coinhabited areas comprising today’s central Europe at the same time as the earliest known rhynchocephalians and squamates. It further demonstrates a more complex evolutionary scenario for dental evolution in early lepidosauromorphs, with independent acquisitions of acrodonty early in their evolutionary history. The small size of most terrestrial vertebrates from Vellberg is conspicuous, contrasting to younger Triassic deposits worldwide, but comparable to Early Triassic faunas, suggesting a potential long-lasting Lilliput effect this fauna.

Specimen availability. The holotype and only specimen of the new taxon is housed in the Staatliches Museum für Naturkunde Stuttgart, Germany under the number SMNS 91590. The anatomical analysis was made with the aid of computed-tomography (CT) scans performed with a Metrotom 800 Generation 1 scanner (S. Tomaschko Zeiss Computertomographie Dienstleistung, Essingen, Germany) using 110 kV and 265 μA at 500 ms and a voxel size of 17,94 μm. Segmentations and measurements were made in the software VG Studio Max 2.0 (Volume Graphics, Heidelberg, Germany). A complete anatomical description is provided in the Supplementary Material.

Morphological and molecular data sets. In order to assess the phylogenetic placement of Vellbergia bartholomaei among diapsid reptiles, we included it in the recently published phylogenetic data matrix of Simões et al. 2018. This data set includes the largest taxonomic sampling available for early diapsid reptiles, and also includes considerable revisions on the construction of morphological characters based on discussions provided by [51,52]. This data set contains both a morphological and a molecular partition sampled for all of the extant taxa included in the data set. Owing to the beneficial effect of removing rogue taxa in phylogenetic analysis, especially in the gain in resolution in support, we removed two taxa that operate as wildcards in the previous version of this data set (Paliguana and Pamelina). Not all taxa identified as wildcards were removed (e.g. Palaeagama and Sophineta), in order to find a balance between increase in tree resolution and support versus keeping a taxonomic sampling relevant to the questions addressed in the present study. Our results consist of morphology only, as well as morphological and molecular (combined evidence) analyses.

Parsimony analysis. Analyses are conducted in TNT v. 1.155 using the New Technology Search (NTS) algorithms. Tree searches are conducted using 1,000 initial trees by random addition sequences (RAS) with 100 iterations/round for each of the four NTS algorithms: Sectorial Search, Ratchet, Drift and Tree Fusing. The output trees are used as the starting trees for subsequent runs, using 1,000 iterations/rounds of each of the NTS algorithms. The latter step is repeated once again, and the final output trees are filtered for all the most parsimonious trees (MPTs).

Bayesian inference analyses. Analyses are conducted using Mr. Bayes v. 3.2.656. As there are no changes to the molecular data set we used, molecular partitions and models of evolution are the same as that study. The morphological partition is analysed with the Mkv model (given that autapomorphies are included in the data set, but there are no invariable characters). Rate variation across characters is sampled from a gamma distribution, and analyses used 4 independent runs with 6 Markov chains each, sampling at every 1000 generations, for a total of 50 million generations. Convergence of independent runs is assessed using: average standard deviation of split frequencies (ASDSF ~ 0.01), potential scale reduction factors (PSRF ≈ 1 for all parameters) and effective sample size (ESS) for each parameter is greater than 200. Leaf stability analysis. Leaf stability was assessed using RogueNaRok, which allows assessing the difference between the highest and the second highest support values for alternative resolutions of each taxon quartet/triplet in the data set (LSdif). We applied this method to the posterior trees from the Bayesian inference analysis including both morphological and molecular data. Because of the large number of taxa and large number of trees, it was necessary to downsample the total number of posterior trees from each analysis (100,000 trees after discarding burn-in). The final sample consisted of 10,000 trees (selecting one at every 10 trees) using the Burntrees script for Perl.

When using the CT scans provided here, please refer to the Staatliches Museum für Naturkunde Stuttgart, as well as to the original publication.