Sequence data were generated with the Big Dye Terminator Cycle Sequencing Ready Reactions DNA sequencing kit (Applied Biosystems, Foster

City, CA, USA) and bidirectional reads assembled (excluding the 5′ and 3′ primer regions) for all three markers using Sequencher™ 4.10 (Gene Codes Corporation, Ann Arbor, MI, USA). The resulting sequences were aligned in MacClade 4 (v. 4.08) for OSX with additional data sourced from GenBank, only if necessary (Table 1). For specimens loosely field-identified as Meredithia sp., Psaromenia sp. or Kallymeniaceae sp., a neighbor joining analysis of a COI-5P barcode alignment (42 specimens, 664 sites) with HKY-corrected distances (default setting) was completed using Tree Builder in Geneious Pro on a Mac Pro [OS X version 10.6.8] (Drummond GDC 0068 et al. 2009). This tree was used to identify genetic species groups. For phylogenetic

analyses, the best models for the individual gene regions LSU rDNA (54 taxa, 2,831 sites; 135 variably aligned sites removed prior to analyses), rbcL (53 taxa, 1,358 sites), and COI-5P (49 taxa, 664 sites; removing replication within species as identified in the previous analysis) were first estimated (AIC) in Model test (v 3.06; Posada and Crandall Wnt assay 1998) as implemented in PAUP* (Swofford 2003) through Geneious. Each alignment was then subjected to maximum likelihood (ML) analysis with the best model of evolution using PHYML in Geneious. Branch support was estimated using the Shimodaira-Hasegawa-like (SH) approximate likelihood ratio test (aLRT) (in our experience SH values are typically similar to bootstrap support values). A multigene alignment was then constructed (54 taxa, 4,718 sites, Glycogen branching enzyme 98% complete: LSU for all taxa; rbcL data missing for one taxon, ~98% complete; COI-5P data missing for five taxa, ~91% complete) and also analyzed (a GTR+I+G model

in all analyses) in PHYML as outlined previously, as well as subjected to Bayesian analyses in MrBayes (v. 3.1.2; Huelsenbeck and Ronquist 2001) with two independent trials (each with parallel runs). Parallel runs of four Markov chains were completed with one million generations and sampling each 200 generations. Data were partitioned by gene, and then by codon for rbcL and COI-5P, and the parameters were unlinked with the overall rate allowed to vary across partitions. The burn-in for each run was determined by plotting overall likelihood scores against generation, which established the stationary phase of each run for estimating the posterior probability distribution. The final estimate was based on pooled samples from two independent runs. Our DNA barcode analyses for a geographically diverse assemblage of specimens loosely assignable to Meredithia sp. or Psaromenia sp. resolved as 14 genetic species groups (Fig. 1). The various Meredithia spp. were typically 2.35%–6.98% divergent with the exception of M. nana and M. pseudopeltata sp. nov., which were only 1.