We initially compiled a data matrix of mitochondrial ND2 sequences for the Sericornithinae (excluding Oreoscopus and Acanthornis, following Gardner et al. (2010), and basal Smicrornis) using all available sequences present in GenBank as of September 2015. To this, we added 25 complete ND2 sequences obtained following the methods of Norman et al. (2007), and an additional three partial ND2 sequences for which DNA extraction and sequencing procedures largely followed Norman et al. (1998). These samples provided increased taxonomic and geographic coverage of species (see taxonomic list and accession numbers in Table S1, Supplementary Information) as well as comparative data for assessing sequence quality, identifying errors in taxonomy, sample contamination and potential pseudogenes. Sequences of Hylacola, Calamanthus, Pycnoptilus, Chthonicola and Pyrrholaemus were treated as outgroups to facilitate reconstruction of phylogenetic relationships amongst the scrubwrens and mouse-warblers. Cross-referencing of specimen details from the independent studies using GenBank and OZCAM online databases identified a duplicate sequence of Calamanthus fuliginosus (Vigors and Horsfield 1827) which was excluded from the dataset. The remaining 51 ND2 sequences included 22 of the 25 species recognised in the combined treatments of Christidis and Boles (2008), Dickinson and Christidis (2014) and Beehler and Pratt (2016), following Gardner et al. (2010). Species missing from the data matrix were the New Guinean perplexing scrubwren Sericornis virgatus (Reichenow 1915) variously treated as conspecific with S. beccarii, the Vogelkop scrubwren Sericornis rufescens (Salvadori 1876) an allospecies of S. perspicillatus, and from Australia, the western fieldwren, Calamanthus montanellus (Milligan 1903) an allospecies of Calamanthus campestris (Gould 1841) with which it may be conspecific. The proposed relationships of these taxa were largely confirmed by Marki et al. (2017).

The data matrix was aligned and edited in MEGA 6 (Tamura et al. 2013) with all heteroplasmic or ambiguous sites coded as missing data. Two sequences published by Gardner et al. (2010) contained a large number of ambiguities (4.5– 7.7% of sites) compared to other sequences in the dataset and their position in the tree was carefully evaluated in preliminary analyses. Visual inspection revealed alignment errors in an additional three sequences from the Gardner et al. (2010) study. These are deposited in GenBank as ‘ND2-like’ sequences and could not be translated as an open reading frame. Two contained errors in the 3′ and/or 5′ regions which were most likely due to compressions, a common sequencing artefact. Manual insertion of gaps (coded as missing data) was used to correct alignment errors and the edited sequences used in subsequent analyses.