The Devonian record presents an opportunity to test and validate an existing marine bioregionalisation. This study is the first to use comparative biogeography and phylogenetic data to test Devonian bioregionalisation. Proposed in the 1960’s the Old World, Eastern Americas, and Malvinokaffric realms have been the functional standard within marine Devonian Biogeography. Data from 32 published phylogenies of Devonian marine taxa and a database of c.800 occurrences were analysed using phylogenetic software to test for area monophyly. The taxic occurrences within the current database were then tested against total fauna Devonian occurrences with the Palaeobiology Database to indicate differences in sampling. Results indicate that the current Devonian bioregionalisation is not representative of natural areas and requires revision. The result highlights areas that are most robust from which the study makes recommendations to improve the process and diagnosis of Devonian biogeographic areas. We found that legacy issues within palaeontology are evident within the results and their interpretation. The validation of bioregionalisation and process is critical to the advancement of biogeography and palaeontology. The sensitivity of bioregionalisation shows biotic and geographical relationships, how life and earth evolved together, and how geographic bias is evident in scientific process.

Database of 475 taxa from 32 published phylogenies was created from the primary literature (Table 1). Distributional data was taken from the published database of Dowding and Ebach (2019) and supplemented by the Palaeobiology database(PBDB; www.paleobiodb.org). Trees were selected based on the Devonian occurrence of the taxa and then polytomies were removed following the Paralogy-free Subtree Method (PFS; Nelson and Ladiges, 1996; Murphy et al, 2019).The PFS resulted in ten reduced phylogenies which were then converted to areagrams using the units, both spatial and temporal, from the bioregionalisation of Dowding and Ebach (2018; Figure 1). Vertebrate fauna (labelled ‘fish’) were separated from invertebrate trees and each (fish, invertebrates, and both) were converted into a paralogy free, uniformly weighted three-taxon statement (3TS) matrices (reviewed by Kitching et al., 1998; Williams and Ebach, 2008) using the program TAXODIUM version 1.2 (Mavrodiev and Madorsky, 2012) (command: inputfilename –ib –ob –og –phy). As a next step, the 3TS matrices were established as an arrays of the minimal trees using script FORESTER version 1.0 (Mavrodiev et al., 2017) (command: ruby ruby seedlings.rb –inputfilename).

A 3TS matrix retains a hierarchical structure, thereby rooting all 3TSs as individual trees. Subsequently, the arrays of the rooted minimal trees were imported to Clann version 4.1.5 (Creevey, 2004; Creevey and McInerney, 2009) to calculate the average consensuses. All average consensus analyses were performed in PAUP* 4.0a (Swofford, 2002) as described by Mavrodiev et al. (2017, 2019). In all cases, the optimality criterion of the best tree is defined in the simplest way – as a ‘distance with non-weighted least squares’ (Lapointe and Cucumel, 1997;  Lapointe and Levasseur, 2004; Creevey, 2004; Swofford, 2002).

Proportions of vertebrate and invertebrate data within the study were tested against the proportions within the Palaeobiology Database. The data were downloaded from the Paleobiology Database using the group name 'marine', type Occurrences’ and parameters: time interval = Devonian, paleoenvironment = marine, taxa = chordata, trilobita, brachiopoda, phyllocarida, and chelicertata. Results were sorted into chordata and invertebrata.  A Chi Squared statistic (X²) and Yates continuity correction (X²Yates)  was applied to the data using R (R Core Team 2020).

The data should be substituted by newer phylogenies. There is a bias toward OWR and EAR units over Malvinokaffric/Malvinoxhosan and Gondwanan areas.