Integrative taxonomy resolves the cryptic and pseudo-cryptic Radula buccinifera complex (Porellales, Jungermanniopsida), including two reinstated and five new species

Abstract Molecular data from three chloroplast markers resolve individuals attributable to Radula buccinifera in six lineages belonging to two subgenera, indicating the species is polyphyletic as currently circumscribed. All lineages are morphologically diagnosable, but one pair exhibits such morphological overlap that they can be considered cryptic. Molecular and morphological data justify the re-instatement of a broadly circumscribed ecologically variable R. strangulata, of R. mittenii, and the description of five new species. Two species Radula mittenii Steph. and R. notabilis sp. nov. are endemic to the Wet Tropics Bioregion of north-east Queensland, suggesting high diversity and high endemism might characterise the bryoflora of this relatively isolated wet-tropical region. Radula demissa sp. nov. is endemic to southern temperate Australasia, and like R. strangulata occurs on both sides of the Tasman Sea. Radula imposita sp. nov. is a twig and leaf epiphyte found in association with waterways in New South Wales and Queensland. Another species, R. pugioniformis sp. nov., has been confused with Radula buccinifera but was not included in the molecular phylogeny. Morphological data suggest it may belong to subg. Odontoradula. Radula buccinifera is endemic to Australia including Western Australia and Tasmania, and to date is known from south of the Clarence River on the north coast of New South Wales. Nested within R. buccinifera is a morphologically distinct plant from Norfolk Island described as R. anisotoma sp. nov. Radula australiana is resolved as monophyletic, sister to a species occurring in east coast Australian rainforests, and nesting among the R. buccinifera lineages with strong support. The molecular phylogeny suggests several long-distance dispersal events may have occurred. These include two east-west dispersal events from New Zealand to Tasmania and south-east Australia in R. strangulata, one east-west dispersal event from Tasmania to Western Australia in R. buccinifera, and at least one west-east dispersal from Australia to New Zealand in R. australiana. Another west-east dispersal event from Australia to Norfolk Island may have led to the budding speciation of R. anisotoma. In contrast, Radula demissa is phylogeographically subdivided into strongly supported clades either side of the Tasman Sea, suggesting long distance dispersal is infrequent in this species.

Th e genus Radula is named for a morphologically distinctive lineage belonging to the Porellales (Davis 2004;Forrest et al. 2006) which comprises the groups Porellinae and Radulinae sensu Schuster (1980). Morphological features characterising Porellinae include incubous leaf insertion, conduplicately-bilobed leaves, lack of ventral branching, and rhizoids in fascicles. Th e Porellales has a global distribution with peak diversity in tropical regions where high rates of endemism are reached (Schuster 1980;Vanderpoorten et al. 2010). Morphologically, Radula is highly divergent within Porellales (see Schuster 1980) and is characterised by several potentially synapomorphic characters including loss of the ventral merophyte row and production of rhizoids from the dorsal surface of the leaf at the apex of the lobule carinal region. So unusual are these features that Schuster (1958) attributed Radula to its own suborder. Radula has been resolved as a monophyletic genus and there is currently little doubt regarding its integrity (Devos et al. 2011a, b). Th is contrasts with the infrageneric classifi cation, where all four subgenera proved poly-or paraphyletic (Devos et al. 2011b), and species were resolved on seven fully supported main lineages (Fig. 1). Identifi cation of the genus in the fi eld is straightforward, the incubously inserted, conduplicately-bilobed leaves, the lack of underleaves, the production of rhizoids in clusters from the surface of the ventral leaf lobe, and the dorso-ventrally compressed perianths (from which the generic name is derived), combined with the above mentioned synapomorphies make the genus distinctive. However, due to morphological plasticity and variation in character states (So 2005), the circumscription of many species is debated (e.g. So 2005So , 2006Renner , 2006, and estimates of global diversity range from around 200 (Gradstein and Costa 2003) to 350 (Yamada 1979). Reported variation in species estimates for both geographical regions and lineages refl ects a contemporary debate about the utility of typological species concepts within bryophytes (Heinrichs et al. 2009a) that has signifi cant consequences for estimates of global species diversity (von Konrat et al. 2010a).
Th irty one and seventeen species of Radula are currently accepted in Australia and New Zealand respectively following studies on Queensland, Tasmanian and New Zealand species (Yamada 1984;So 2005;Renner 2006;Renner et al. 2010a). Th e most common and widespread Australasian species is Radula buccinifera (Hook.f. & Taylor) Taylor ex Gottsche, Lindenb. & Nees, which is distributed from south-west Western Australia through Victoria, Tasmania, eastern New South Wales, and across the Tasman to New Zealand and the Chatham Islands 900 km east of Christchurch (Fig. 2). Radula buccinifera has a latitudinal range from the subantarctic Campbell and Auckland islands in the south (52°30'S), to Moa Island in Torres Strait at the tip of Cape York Peninsula in the tropics (10°10'S). Across this geographic range, R. buccinifera can be found in a wide variety of habitat types, including tropical and subtropical lowland forest, tropical montane rainforest, warm and cool temperate rainforest including those dominated by Nothofagus , wet sclerophyll forest, sub-alpine shrubland, and even alpine grassland and herbfi eld. Within this remarkable diversity of habitats R. buccinifera can be found on an equally remarkable variety of substrates, including on leaves, twigs, branches, naked bark and bryophyte mats on tree trunks, rotting logs, clay soil banks and on rocks including those almost permanently submerged beneath running water. Coupled with the exceptional geographical and ecological variation exhibited by R. buccinifera is considerable morphological diversity (Fig. 3). Radula buccinifera was last revised in 2005, in a study that expanded the number of synonyms from four to six, and circuitously concluded that ' R. buccinifera has been described under several diff erent names, indicating the variability exhibited by the various forms' (So 2005). Within Australasia Radula buccinifera is therefore a typical example of a widespread variable liverwort species.
Morphologically circumscribed bryophyte species tend to have broader geographical distributions than angiosperm species (Shaw 2001;Vanderpoorten et al. 2010), and within an Australasian context R. buccinifera does not appear to be an exception.  Th is may be a real pattern attributable to spore-based reproduction and the relative ease with which spores are transported and new populations established (Muñoz et al. 2004;Pohjamo et al. 2006, Cassie et al. 2008Hutsemékers et al. 2008), or an artefact of failing to detect cryptic or semi-cryptic species (Heinrichs et al. 2009b;Medina et al. 2012). In the Australasian region 290 species and 200 genera of seed plants (c. 9% of New Zealand's fl ora) are indigenous to both Australia and New Zealand (Wilton and Breitwieser 2000;McGlone et al. 2001;de Lange et al. 2007), 90 of New Zealand's c. 200 fern species are shared with Australia (Brownsey 2001), and 125 of 500 moss and 284 of New Zealand's 595 liverwort species also occur in Australia (Engel and Glenny 2008a;Perrie et al. 2010) suggesting widespread exchange of all fl oristic components via long-distance dispersal (LDD). Indeed, LDD in bryophytes has frequently been identifi ed by molecular phylogenetic studies (e.g. McDaniel and Shaw 2003;Shaw 2001;Rycroft et al. 2002;Stech and Dohrmann 2004;Heinrichs et al. 2006Heinrichs et al. , 2009aFeldberg et al. 2007;Shaw et al. 2008;Hentschel et al. 2009;Dong et al. 2012;Patiño et al. 2013a), and trans-Tasman (McDaniel and Shaw 2003;Perrie et al. 2010) and trans-Antarctic (Pokorny et al. 2011) dispersal events have been inferred.
Although cryptic haplotype diversity was detected within Radula lindenbergiana (Laenen et al. 2011), no cryptic species complexes have yet been identifi ed within Radula . Th e only molecular phylogenetic study at species level, also conducted on the Macaronesian fl ora, found agreement between morphological and molecular species circumscription (Stech et al. 2010). However, a recent global phylogeny for the genus included three representatives of R. buccinifera that formed a clade within which R . australiana, a morphologically distinctive species , was nested (Devos et al. 2011a, b). In this study we assess evidence for or against a broad circumscription (i.e. a widespread, ecotypically and morphologically variable species) of R. buccinifera and whether R. australiana should be subsumed within R. buccinifera . We sequence multiple accessions of R. australiana and R. buccinifera encompassing the geographical, ecological and morphological variation exhibited by the species as currently circumscribed. Our results provide the fi rst demonstration of cryptic species in Radula , and one of the most extreme cases yet documented in liverworts in terms of neglected (i.e. not previously recognised at any level) diversity uncovered by phylogenetic data. Th ough subtle morphological diff erences between most entities mean the complex is dominated by semi-cryptic species, one pair of species can, by virtue of overlapping morphological variation, be considered genuinely cryptic. Th is study demonstrates the generality but also the potential severity of the phenomenon of cryptic and semi-cryptic species in liverworts.

Materials and methods
Taxon sampling and molecular protocols.
Sampling for DNA was based on material collected haphazardly throughout the geographical range reported for R. australiana and R. buccinifera and including New

DNA sequence alignmnt and phylogenetic analyses
For each DNA region, forward (5'-3') and reverse (3'-5') sequences were assembled and checked for inaccurate base calling using Geneious (Drummond et al. 2012). Consensus sequences were aligned by MUSCLE (Edgar 2004) on the CIPRES portal (Miller et al. 2010) and manually edited in BioEdit 5.0.9 (Hall 1999) following alignment rules and principals of homology outlined in Kelchner (2000) and Morrison (2006). Table 1. Voucher information and GenBank accession numbers for specimens included in the molecular dataset. All 'MR' isolates were newly generated for this study. Maximum parsimony (MP) analyses were performed on individual markers, which revealed no signifi cant (≥70% bootstrap support) incongruence among markers, so markers were concatenated for subsequent analyses. MP analyses were run using the parsimony ratchet (Nixon 1999) performed by PAUP* v4.0 (Swoff ord 2002) on the CIPRES portal. Branch swapping was performed using TBR and 500 random addition sequence replicates, and the strict consensus of most parsimonious trees computed. Autapomorphic and constant characters were excluded prior analysis. Data consistency was assessed by conducting 1000 bootstrap replicates with 10 random addition sequence replicates per bootstrap replicate, in PAUP* v4.0. Clades were considered supported if present in ≥ 70% and strongly supported if present in ≥ 90 % of trees from each replicate.

Species
Maximum likelihood (ML) analyses were implemented in GARLI (Zwickl 2006). A single GTR (six state model) was applied to the whole dataset, with parameters estimated from the data. Rough initial optimization was performed from a random seed, and the analysis run until 15,000 generations had passed without improvement in tree topology. Ten independent runs having diff erent random seeds yielded the same tree topology. Clade support under likelihood criterion was estimated using ML bootstrap in RAxML v7.2.6 (Stamatakis 2006) provided on the CIPRES portal (Miller et al. 2010). Analyses used the GTRGAMMA model and replication was terminated at the discretion of the program.
Bayesian analysis was performed with a hybrid version of MrBayes (Huelsenbeck and Ronquist 2001;Pratas et al. 2009) on the CIPRES portal. For each partition the best-fi t substitution model was selected according to the Akaike Information Criterion (AIC) calculated by jModelTest 0.1.1 (Posada 2008). Substitution models for each partition were unlinked. Default program settings specifi ed a priori probabilites for other parameters. Phylogenies were estimated using two independent Metropolis coupled MCMC runs comprising one cold and three heated chains run simultaneously for 7.9 million generations (the number completed in a 24 hour period) and sampled every 1000 th generation. Convergence between the four runs was assessed by comparing the trace fi les for each using Tracer 1.5 (Rambaut and Drummond 2009) and after plotting log likelihood values the fi rst 500 samples were discarded as burnin. Table 2. Primers used in this study for amplifi cation and sequencing of three chloroplast DNA regions.

Region Primer Sequence 5'-3' Direction References
atp Brbc L atp B-1 ACATCKARTACKGGACCAATAA Forward Chiang et al. (1998) rbc L-1 AACACCAGCTTTRAATCCAA Reverse Chiang et al. (1998) trn Ltrn F A50272 ATTTGAACTGGTGACACGAG Forward Taberlet et al. (1991) B49317 CGAAATCGGTAGACGCTACG Reverse Taberlet et al. (1991) trn G trn GF ACCCGCATCGTTAGCTTG Forward Pacak and Szweykowska-Kulinska (2000) trn GR GCGGGTATAGTTTAGTGG Reverse Pacak and Szweykowska-Kulinska (2000) Morphology Radula buccinifera from AK, BM, BRI, CANB, CHR, FH, G, MEL,  MPN, NSW, NY, PERTH, S, and WELT were examined. Morphology was assessed with the aid of dissecting and compound microscopes. In the descriptions, measurements are given for the lowest and highest observed values for a given structure, so are indicative of range only. Leaf lobe length is perpendicular to the stem axis, width is parallel to the axis, and does not include the lobule. Lobule length and width are measured parallel and perpendicular respectively to the line between the apex and base of the keel (the lobelobule junction and the postical extremity of the stem insertion line). Lobule size and shape were assessed by rehydrating and mounting material on a slide for investigation with a compound microscope. Th is is essential for any meaningful comparison, as deformation associated with dehydration obscures subtle shape diff erences. For identifi cation, lobule shape should always be assessed on the basis of hydrated material, and is best assessed from slide-mounted shoots with a compound microscope, however differences can, with practice, be observed in the fi eld with a 20× handlens in good light. Stem transverse sections were prepared by hand from primary shoots, with sections taken from three diff erent shoots for each individual, and slide mounted in water for observation. Dissections of female bracts, gynoecia, and archegonia were by hand with the aid of a pair of Inox #5 'Biologie' tweezers and slide mounted in water. Perianth longitudinal sections were also prepared by hand, with two or three perianths from a selection of individuals examined for each species depending on availability, and slide mounted in water for examination.

Specimens of
Observations of species ecology were made during fi eldwork for various purposes in New Zealand and Australia from 2000 to 2013. Geographic data was drawn from digitised collections, in particular AVH, and from geo-referenced specimens.
Capsule and perianth lengths for three specimens of R. strangulata were measured with an eyepiece micrometer on a compound microscope from perianths with dehisced sporophytes in situ. Capsule length is really capsule valve length. Correlation statistics were calculated using the stats package in R 2.12.0 (Th e R Foundation for Statistical Computing 2010, http://www.R-project.org)

Species concept
Species described here are formal placeholders for hypotheses explaining the distribution of character data from multiple sources, including morphology, ecology, geography, and molecular sequence data among individuals (see Fitzhugh 2005). Th ese hypotheses are derived from an explanatory model wherein restricted gene fl ow between groups facilitates acquisition and maintenance of character state diff erences that in turn precipitate partial or complete reproductive isolation, and in turn ecological, and phylogenetic divergence alone or in combination. Non-compliance with any of these criteria, where exceptions can be reconciled against likely real processes past or present, is always permitted.

Sampling
We sampled 62 and 25 individuals of R. buccinifera and R. australiana respectively. An additional fi ve individuals belonging to other species of subg. Metaradula were included to increase the stringency of our test of R. buccinifera 's monophyly. We obtained atp Brbc L spacer sequences for 85 accessions, trn G for 86 accessions and trn L-F for 88 accessions, thereby 259 new sequences were generated for this study (Table 1). A further 101 trn G and 89 trn L-F sequences were included for unpublished R. buccinifera sequences and the 93 species in the global phylogeny (Devos et al. 2011a, b). Alignments resulted in datasets of 745 putatively homologous sites for atp Brbc L spacer, 777 for trn G, and 671 for trn L-F, the concatenated dataset had 2193 alignment positions of which 1391 were constant, 216 autapomorphic, and 586 were parsimony informative. Th e combined dataset included 449 of the total 591 sequences, a missing rate of 24%, mostly in the atp Brbc L spacer which was not sequenced by Devos et al. (2011a, b). Th e aligned, concatenated dataset is available at Dryad (http://datadryad. org/) doi: 10.5061/dryad.h43q8

Phylogenetic analyses
All data partitions converged on nearly identical topologies for supported clades, with no signifi cant disagreement. All three partitions recover the subgeneric framework resolved in Devos et al. (2011b). In all analyses of concatenated data the tree topology recovered contained the same seven major lineages as Devos et al. (2011a, b) all of which were again fully supported, with the same relationships between subgenera (Figs 4, 5). Individuals of Radula buccinifera were resolved in two fully supported subgenera, six in subg. Radula and the remainder in subg. Metaradula . Within subg. Metaradula , R. buccinifera individuals are resolved in six strongly or fully supported clades with long stems relative to tips (Fig. 4). Th ese clades subdivide the geographical range, microhabitat diversity, and morphological variation exhibited by R. buccinifera . For clarity and consistency these clades are referred to by name throughout the results.
Th e six individuals of Radula mittenii are resolved in a fully supported clade nested within subg. Radula (Figs 4,5). Radula mittenii exhibits a number of distinctive morphological features that separates it from other members of the species complex. In life, the plants have a distinctive milky-yellow appearance that is preserved to a greater or lesser degree in herbarium material. Th is milky lustre may be due to the fi nely verrucose ornamentation on the surfaces of leaf cells. While fi nely verrucose leaf surface ornamentation is found in some other species of subg. Radula , including R. madagascariensis and R. refl exa , it does not occur in any other member of the R. buccinifera complex. Radula mittenii individuals are relatively large, and regularly pinnately branched compared to other individuals, the lobules have a large ampliate interior free margin that, on primary shoots, covers and obscures the stem in ventral view. Th e stem section of this entity has nodular to confl uent trigones throughout the stem section. Perianths have a low basal stem perigynium and a calyptra perigynium has a multistratose base. Individuals of this lineage occur on tree trunks or the sides of granite boulders in tropical rainforest habitats from sea level to 1600 m in the Wet Tropics Bioregion of north-east Queensland.
Relationships at the base of subg. Metaradula are not well resolved (Figs 4,5). One clade containing a number of epiphyllous taxa from the paleotropics is resolved sister to the remainder of the subgenus without support. Relationships between R. ratkowskiana , a morphologically distinctive species Braggins 2004, 2005), and other species within subg. Metaradula are also unsupported.
Radula strangulata was represented by 11 individuals from New Zealand, one from Tasmania and one from New South Wales, and was resolved with high posterior probability. While this species is phenetically heterogeneous, individuals share several morphological features including rotund leaf-lobes spreading and usually held more or less appressed to the substrate so that the dorsal stem surface is visible between the leaves and tending remote in plants from hyper-humid and wet sites. Th e lobules are typically triangular when small, longitudinally oblong with increasing size, and have an acuminate apex with small but abruptly ampliate free interior margin when maximally developed. A dorsal leaf-free strip two or three cell rows wide is present. Th e perianth has a plane mouth and well-developed stem perigynium. Ecologically, plants occur in a variety of microsites close to ground level within forest interiors, including tree-trunk bases, rotting logs, soil banks, rocks along and within streams, sometimes under fl owing water. Th e morphology of these plants corresponds to a number of types, including those for R. levieri Steph., R. silvosa E.A.Hodgs. et Allison, and R. strangulata Hook.f. & Taylor, with the last name having priority.
Th e seven individuals of Radula notabilis are resolved in a clade with full support, although relationships between this species and others are again unsupported. All individuals of Radula notabilis were collected as trunk, branch or twig epiphytes in tropical lowland forests, growing on naked bark. Th ey all have a distinct brown green hue and hold their leaves closely appressed to the substrate so that the dorsal stem surface is visible between the leaves. In addition the plants are sparingly branched, with shoots tending to run parallel along the substrate. Th e lobule is longitudinally rectangular, the stem in transverse section has heavily brown-pigmented cortical cell walls, and bulging trigones at the angles of the medulla cells. Th e undulate and repand perianth mouth is distinctive.
Other species of the R. buccinifera complex, including R. australiana , are resolved in a fully supported clade.
Radula imposita is represented in the phylogeny by two individuals (one extracted and sequenced twice), one a twig epiphyte on Sloanea australis overhanging a stream in the North Coast of New South Wales, the other an epiphyll on Normandia leaves, also overhanging a stream in the Wet Tropics Bioregion of north-east Queensland. . Majority Rule tree from posterior probability distribution sample taken by MrBayes, shown as a modifi ed proportional tree, with some branch lengths shortened, for presentation purposes, with morphological groups identifi ed. Tree topology, rather than branch length is emphasized in this tree, the branches are not to any scale. For branch lengths proportional to substitution rate refer to Figure 4 which is a phylogram with identical topology. Support values associated with each branch are parsimony bootstrap / likelihood bootstrap / posterior probability. Asterisks indicate full support. Only values for supported branches associated with the Radula buccinifera species complex are shown.
Both plants are small, irregularly pinnately branched, the leaves are obliquely patent, and have lobules with an attenuate apex and pyriform carinal region. Th e leaf cells are bulging and occasionally weakly papillose.
Radula buccinifera and R. demissa are resolved as sister taxa with strong support. Radula buccinifera is resolved with full support. Individuals of R. buccinifera are relatively large, with weakly obliquely spreading leaf lobes that do not cover the entire dorsal stem surface, leaving some of the stem visible in dorsal view. Th e leaf insertion does not obtain the dorsal stem mid-line, leaving one to three cortical cell rows leaf-free; leaf-lobes are rotund, not falcate, the lobules, when fully developed, have an ampliate free interior margin, the apex is obtuse to acute, and the perianths have a plane mouth. Th ese plants occur in a range of microsites in forest interiors close to the forest fl oor, including tree-trunk bases, rotting logs, soil banks, rocks along and within streams, sometimes under fl owing water. Microhabitat diversity decreases northward, with plants from central and northern districts of New South Wales primarily lithophytic on vertical rock surfaces.
Radula demissa is resolved with full support. Individuals of R. demissa have obliquely patent leaf-lobes that completely obscure the dorsal stem surface, the leaflobes are falcate. Th e leaf insertion attains the dorsal stem mid-line, leaving no dorsal leaf-free strip. Th e lobules are rhombic to widely rhombic. Th e perianth mouth is often weakly inrolled in immature perianths whose apex is bicornute as a result, however the mouth is usually plane in mature perianths. Th ese plants are epiphytes on tree trunks, branches, twigs, and occasionally epiphylls in humid, well lit sites.
Radula australiana is resolved in a fully supported sister relationship with an unidentifi ed accession collected as an epiphyll in the wet tropics of north-east Queensland. Phenetically this individual is similar in some respects to Radula gedena Gottsche ex Steph. but lacks gemmae, and has leaf-lobe cells that are uniform in size. Furthermore the leaf-lobules have a curved keel and lack a protruding mamilliform pocket in the carinal region. In the relative size of its lobules, its leaf-lobe shape and orientation, this plant does resemble R. australiana . Although similar to at least one species assigned to sect. Epiphyllae by Castle (1939) and Yamada (1979), broader consideration of Australian species and types reveals that in details of its lobule shape, leaf-lobe cell wall architecture and stem section the plant is a good match with the type of R. robinsonii Steph. (G00042708!) . Radula robinsonii and the R. parvitexta complex will be the subject of a subsequent treatment.
Radula australiana itself is resolved monophyletic with strong support. Despite the sister relationship with tropical species, R. australiana inhabits alpine areas in New South Wales, Victoria and New Zealand. Individuals are characterised by large leaflobules up to one quarter the area of the leaf lobes, with acute to acuminate apex, and broadly ampliate free-interior margin, typically brown-green plants with obliquelypatent leaves that obscure the stem in dorsal view. With one exception (NSW875829) these plants were all collected from rock outcrops and rocky areas in alpine areas, often in association with watercourses, and with the same exception all plants are a good match with the isotype of R. australiana. Radula anisotoma , represented by a single accession from Norfolk Island nested within Radula buccinifera , is morphologically divergent in having small, triangular lobules, caducous leaf lobes, and elliptic-oblong female bract lobes. Th ese features have not been observed, alone or in combination, in R. buccinifera from mainland Australia and Tasmania.

Taxonomic implications
Th e name Radula buccinifera is currently applied to a phylogenetically heterogeneous assemblage of six lineages belonging to two recognised subgenera. Th e reason is simple: characters separating morphologically distinct entities within the R. buccinifera complex have been repeatedly overlooked or misinterpreted. Demolition of substandard taxonomic output by molecular phylogenetic investigations is trivial. However, molecular investigations at species level often fi nd confl ict between Linnean classifi cations and phylogenetic relationships even when species taxonomy based on morphology is well resolved (e.g. Heinrichs et al. 2009aHeinrichs et al. , 2012. Molecular data almost always prompts an increase in recognised diversity (e.g. Ramaiya et al. 2010;Pons et al. 2011;Heinrichs et al. 2009bHeinrichs et al. , 2010 by identifying cryptic and semi-cryptic species that represent knowledge gaps (Johnson and Cairns-Heath 2010;Medina et al. 2012). Th e degree to which our previous understanding was defi cient is refl ected in the severity to which extant diversity was under or over-represented by hypotheses of relationship. By this metric, the current circumscription of Radula buccinifera ranks among the most ineff ective yet identifi ed among bryophytes, rivalled by Conocephalum conicum (Odrzykoski and Szweykowski 1991;Kim et al. 1996Kim et al. , 2001 and Frullania tamarisci ). However, both C. conicum and F. tamarisci have near-global distributions, whereas R. buccinifera is confi ned to Australasia. While cryptic species within both C. conicum and F. tamarisci are mostly allopatric, in R. buccinifera up to three entities may be found growing sympatrically. Furthermore, pairs of entities often grow syntopically, which in part explains the admixtures found in voucher specimens and, more problematically, type material.
Most published examples of cryptic diversity within bryophytes come from the northern hemisphere in particular Europe and North America (e.g. Baczkiewicz et al. 2008;Shaw and Allen 2000, Ramaiya et al. 2010, Medina et al. 2012Feldberg et al. 2007;Oguri et al. 2008), with some examples from the Neotropics (e.g. Heinrichs et al. 2009b;Feldberg et al. 2011;Dong et al. 2012) but only a few examples have been published to date for the southern hemisphere. One example of sympatric cryptic species having a centre of diversity in Australasia has been identifi ed in the lichen Cladia aggregata , where 12 species were resolved (Parnmen et al. 2012).
In liverworts, lineage diversity suggestive of cryptic species in the Australasian Lepidozia ulothrix was identifi ed by Cooper et al. (2011Cooper et al. ( , 2012, and in another study the New Zealand Lejeunea tumida species group was found to be polyphyletic (Renner et al. 2010b. Th e additional example provided by R. buccinifera suggests that crypsis and semi-crypsis are not just widespread, but are likely to be taxonomically and geographically pervasive within bryophytes, and the existence of sympatric cryptic entities may prove a common theme in the cryptogamic fl ora of the Australasian region.
Th ere may be good reason for the persistent failure by traditional approaches to recognise instances of crypsis and semi-crypsis. Investigation of patterns of morphological variation within species belonging to the Lejeunea tumida species group found extensive overlap between species due to substantial intra-individual variation (Renner et al. 2013). Morphological variation that is within the developmental capacity of single genotypes, and even single meristems, can swamp inter-specifi c diff erences and mask diff erences between species, resulting in a variation white-out obscuring them to typological approaches. Furthermore, the degree of overlap between cryptic species can be independent of degree of relatedness. Crypsis can be an intrinsic by-product of the growth and development of liverwort gametophytes and cryptic species may be pervasive throughout the group. Lability in leaf morphology has been observed in the mosses Fontinalaceae (Shaw and Allen 2000), and Tortula (Mishler 1986(Mishler , 1988, suggesting developmental plasticity in gametophyte morphology could contribute to crypsis in mosses.
Within the R. buccinifera complex molecular data serve to emphasise the phylogenetic signifi cance of the subtle morphological diff erences detected between lineages, diff erences that have either been overlooked, or dismissed via ad hoc and untested hypotheses of environmental or other intra-specifi c variation.
Variation and co-occurrence in sympatry complicate determination and may have contributed to the generally poor standard of identifi cation in herbarium material. For this study 533 named specimens of R. buccinifera (not including types) held by Australasian and overseas herbaria were examined, of which 190 (36 %) were actually R. buccinifera . Th is implies a misidentifi cation rate of 64%, even including specimens that have been determined for recent regional revisions and synopses. Th is includes specimens that are now referred to new species. Regardless of how a misidentifi cation rate might be calculated to account for names available at the time of identifi cation, the point is that most specimens identifi ed as R. buccinifera do not belong to that species. Perhaps encouragingly, most of the confusion is between R. buccinifera , R. strangulata , and R. demissa , and the fi rst two are very diffi cult to distinguish using morphology. However, R. demissa diff ers in a number of qualitative and quantitative micro-morphological characters. Th e high rate of mis-identifi cation has two consequences, fi rstly Radula buccinifera is misunderstood as a widespread and variable species, and secondly real phylogenetic diversity is overlooked. Th e set of misidentifi ed specimens comprised 26 diff erent Radula species (Table 3), including four of the new species described below, suggesting the working circumscription of R. buccinifera was broad enough to include almost the entire Australasian Radula fl ora. Biodiversity Th e phylogenetic breadth of molecular phylogenetic investigations that have identifi ed cryptic and semi-cryptic diversity, coupled with a mechanism explaining complicated, and often confusing, patterns of morphological variation make the extrapolation that all bryophyte groups contain overlooked diversity a fairly safe inference. Many studies result in reinstatement of synonyms (Heinrichs et al. 2006 and references therein; Heinrichs et al. 2009b, Medina et al. 2012 thereby reducing global synonymy rates, but many also detect genuinely new entities, eventually increasing the number of published and accepted names (e.g. Kreier et al. 2010 Our study suggests that within R. buccinifera , two synonyms need reinstating, and fi ve new species need naming. Th is represents an increase of about 2% in global diversity in Radula, from study of a single 'species' in a relatively small southern region that does not have a reputation for diversity in this genus in comparison to, for example, the tropics. Th ere may be other Radula species of convenience in paleotropical, neotropical, and oceanic regions. Th e precedent set by R. buccinifera suggests that resolution of 'snowball' taxa could signifi cantly increase estimates of global diversity for this genus. Th e taxonomically pervasive, indeed indiscriminate, distribution of cryptic and semi-cryptic species suggests that estimates of global diversity for liverworts revised upward of 10,000 species might not be unreasonable. New Zealand, one of the regions involved in the R. buccinifera complex, has the best studied liverwort fl ora in the southern hemisphere and is under ongoing investigation by a number of research groups, yet totally novel species are still being discovered (e.g. Glenny 1996;Renner et al. 2006Renner 2010, Engel and Glenny 2008bvon Konrat et al. 2011von Konrat et al. , 2012Glenny et al. 2009), synonyms within accepted species are being reinstated (Engel and Smith-Merrill 2010), taxon circumscriptions are being revised (von Konrat et al. 2010b) and through a combination of these and new geographic records (e.g. Glenny 1995;von Konrat et al. 2006;Renner and Pócs 2011) the liverwort fl ora is increasing at the rate of c. 5 species per annum (von Konrat et al. 2010a). If such increases are possible in a well-studied fl ora, the same is possible in other less well-studied parts of the world.

Biogeography, dispersal and speciation
Th e identifi cation of cryptic species, and reconstruction of spatial structure of genetic diversity informs biogeography and evolutionary ecology. In bryophytes, morphologically circumscribed species generally have larger distribution ranges than angiosperm species (Shaw 2001). Th is study presents another example where the broad morphological circumscription of a bryophyte species is untenable. To the best of our knowledge, Radula buccinifera does not occur in New Zealand. At fi rst glance it may seem that the Australian and New Zealand fl oras have less in common following this revision. On the contrary the number of species shared between Australia and New Zealand doubles, due to the detection of R. strangulata in Australia, and the trans-Tasman distribution of R. demissa , described below.
Morphological similarity and continuity between R. buccinifera and R. strangulata makes inference of R. strangulata 's distribution in Australia diffi cult, but at least two other specimens, both collected from rock under running water on the Australian mainland are morphologically and ecologically compatible with this species, suggesting it may be more widespread there. Th e apparent rarity of R. strangulata in Australia, in contrast to its abundance in New Zealand, and the sequence similarity between the two regions are both consistent with recent east to west dispersal, though there may be other explanations. Other examples of east to west dispersal against the prevailing south-westerlies include two species of Chionohebe that dispersed independently from New Zealand to Australia, (Meudt and Bayly 2008), and two independent dispersals of Asplenium hookerianum that established populations in Victoria and Tasmania (Perrie et al. 2010).
In Radula demissa two reciprocally monophyletic geographic clades are also recovered in phylogenetic analysis (Figs 4,5). Despite genetic divergence, there are no consistent morphological diff erences between individuals from New Zealand and Tasmania. It is possible that dispersal from Australia to New Zealand or vice versa was associated with a single dispersal event at a time when the species had a single chloroplast genotype and the genetic variation in the samples all post-dates that dispersal event. But our sampling, focussed as it is on the chloroplast genome, may have failed to detect descendants of other rare long distance spore-dispersal and events between the two regions by virtue of their low frequency, the fact that chloroplasts are probably maternally inherited, half the colonists will be male, half the off spring of female colonists will be male, and due to purely stochastic processes rare chloroplast haplotypes would be expected to go extinct within the colonised region (Hubbell 2001). Regardless, phylogeny suggests long distance dispersal and establishment between Tasmania and New Zealand is a rare event in R. demissa .
Considerable lineage diversity was recovered within Australian R. australiana . In contrast, diversity within New Zealand was more limited, with fi ve of seven accessions forming a clade nested within Australian accessions. Populations established following dispersal should contain only a proportion of the variation in the parental populations (Templeton 1998), so this pattern is consistent with colonisation of New Zealand from an Australian source. Radula australiana is certainly easier to fi nd on the Australian mainland than in New Zealand, and is associated with most rock outcrops (granite and basalt) in alpine regions. Th e apparent completeness with which R. australiana has fi lled this habitat in Australia may explain its earlier detection there, than in New Zealand where it is widely but sparsely distributed in alpine areas, and is absent from apparently suitable sites. Formation of alpine habitats in New Zealand is believed to have occurred during the Pliocene (5-2.5 mya) (Cox and Findlay 1995;Batt et al. 2000;Winkworth et al. 2005), and dispersal to New Zealand is likely to be younger than this.
All entities within the R. buccinifera complex are dioicous, which is the ancestral and most common condition in Radula (Devos et al. 2011a). Unlike cosexual and autogamous plants, at least two dispersal and establishment events in close proximity would be necessary to establish sexually reproductive populations in a new area. Th e capacity for dioicous bryophytes to form isolated mixed-sex populations was noted by Shaw (2000) from observations of Mielichhoferia . Aside from stochastic processes associated with spore dispersal and establishment, spore characteristics are relevant determinants of dispersability (van Zanten 1976(van Zanten , 1978Gradstein et al. 1983). Tolerance of freezing might be crucial for maintaining viability during transit by highly unstable weather systems such as tropical cyclones, where violent internal convection cells expose spores to rapidly fl uctuating temperatures, and not all species are tolerant of freezing (van Zanten 1978). For example, Hymenodontopsis (Pyrrhobryum) mnioides spores are poor stress tolerators, more so than the average species occurring in New Zealand and Tasmania and this possibly explains divergence between Australasia and Patagonia (van Zanten 1978;McDaniel and Shaw 2003). In contrast Leptotheca gaudichaudii , whose spores germinate following freezing (van Zanten 1978), has haplotypes shared between Australasia and Patagonia (McDaniel and Shaw 2003).
Despite its apparent rarity long-distance dispersal has contributed to diversity within the Radula buccinifera species-complex. One instance where a geographically isolated (Norfolk Island), morphologically distinctive individual is nested within an Australian clade was identifi ed in this study, in R. buccinifera . Rare and stochastic long-distance dispersal has been reported as contributing to diversifi cation in Leptoscyphus (Vanderpoorten and Long 2006), by budding speciation where a population becomes spatially isolated and subsequently diverges, rendering the parental species paraphyletic (Vanderpoorten and Long 2006). Th e small size of the isolated population facilitates rapid drift and selection as parental alleles are lost at a faster rate than in parental populations. Th e molecular data for R. buccinifera are consistent with an instance of budding speciation via dispersal from west to east across the Tasman Sea. Th e acquisition of gemmae by the Norfolk Island species is most likely an adaptive morphological response to challenges posed by its island habitat, i.e. dioicy and small population size resulting in limited opportunity for sexual reproduction. Highly modifi ed gemmae are a common feature of many species within the tropical epiphyllous lineage of subg. Metaradula , e.g. R. protensa , R. assamica , R. epiphylla , but gemmae are virtually unknown within other lineages of the subgenus. Asexual reproduction may be a key innovation underpinning rapid diversifi cation within Mitthyridium a genus widespread across the Pacifi c (Wall 2005), and is a key adaptive trait to existence in a diff erent kind of island habitat (leaves) in Lejeuneaceae (Kraichak 2012). Indeed, the idea that fast and effi cient dispersal mechanisms at the local scale might be selected in volcanic island environments due to the abundance of empty niches has been recently suggested to explain the higher proportion of species producing specialized asexual diaspores on volcanic islands than on continents (Patiño et al. 2013b).
Th e traditional view that morphological evolution in bryophytes takes place over millions, if not tens of millions of years, has been confi rmed in a couple of dated phylogenetic studies, including the moss Hymenodontopsis (as Pyrrhobryum ) (McDaniel and Shaw 2003) and the liverwort Leptoscyphus (Devos and Vanderpoorten 2008). In contrast, the similarity in haplotypes between R. buccinifera and the Norfolk Island species suggests more recent divergence, and more rapid morphological evolution. Age estimates for islands and habitats vary, and known fossils cannot be conclusively tied to nodes within the genus so dating the Radula phylogeny based on the current sample is not possible, Th ough Radula anisotoma on Norfolk Island has a maximum age of 7 Ma so could be used for calibrating the tree, but would need to be supplemented with other dates. Elucidating the time course underpinning processes resulting in contemporary phylogeographic patterns should be the focus of future study.

Taxonomic treatment
Artifi cial key distinguishing species belonging to the Radula buccinifera aggregate. Th e fi rst character presented is usually diagnostic. Other characters are included to 1) facilitate identifi cation as far as possible, 2) identify couplet selection errors at subsequent steps of the key and 3) aid in the identifi cation of species that are not included in this treatment, either because they are novel, or unrelated. 1 Leaf-lobe cell surface roughened, verrucose. Lobules one quarter the lobe area on primary shoots, quadrate, with ampliate interior margin. Shoot systems regularly pinnate and subdimorphic with secondary shoots smaller than primary, and with more rectangular lobules whose antical margin may be refl exed near the stem insertion; plants from exposed situations may comprise mostly secondary shoots and the regularly pinnate branching pattern may not be apparent. Stems relatively massive 190-250 μm diameter, with cortical cells in a single tier of 30-50 rows; cell walls brown pigmented throughout; cortical cell walls heavily and continuously thickened, at times constricting the cell lumen; medulla cells in 80-110 rows, cell walls heavily thickened with coarse nodular trigones that become confl uent, and constrict the cell lumen. Female bracts in one and a half or two pairs. Lobules rhombic to trullate, inner lobule margin free for up to two thirds its length, free portion not ampliate, not extending across stem beyond insertion line, apex narrowly rounded to acute, free exterior margin straight, occasionally with a small knee above the lobe-lobule junction, margins entire; leaf-lobes weakly falcate. Stem anatomy with all cortical cell walls heavily and almost continuously thickened and Lobules quadrate to rhombic when small and large, one eighth to one sixth the lobe area; keel apex and postical lobe margin with shallow notch; interior lobule margin free for one third its length, free portion weakly ampliate in small stature lobules to moderately ampliate on large stature lobules, extending at most half way across the ventral stem surface; acroscopic margin S-shaped (typical in situ) to straight (when fl attened), apical portion inclined toward stem, not exceeding (lying antical to) the lobule apex; apex obtuse to acute; free exterior margin straight curved, occasionally with a small knee above the lobe-lobule junction; margins plane, entire or shallowly repand; lobe-lobule junction slightly antical to, or level with, the acroscopic end of stem insertion . Lobules quadrate when small to oblong, one twelfth to one sixth the lobe area, keel apex and postical lobe margin fl ush; interior lobule margin free for one fi fth to one quarter its length, free portion not ampliate in small stature lobules to moderately ampliate on large lobules, extending at most half way across the ventral stem surface; acroscopic margin S-shaped, apical portion perpendicular to stem, in large lobules exceeding (lying antical to) the lobule apex; apex obtuse to apiculate; free exterior margin straight, margins plane; lobe-lobule junction well postical to the acroscopic end of stem insertion . Diagnosis. Within the Radula buccinifera complex R. anisotoma is most similar to Radula strangulata by virtue of its small, rhomboid lobules whose apex lies close to or over the stem, and its leaf-lobes not interlocking over the dorsal stem surface, such that the stem is visible between leaves in dorsal view, and its habit of growing on rocks in association with waterways, but diff ers by its oblong leaf-lobes that are caducous, fragmenting into several irregular pieces, its narrower and longer female bract lobes, rhombic to trullate lobules and smaller stature.

Description. [From CANB650459
] Forming diff use patches of small shoots, or mixed with other bryophytes, brown in herbarium; shoot systems monomorphic, irregularly branched, Lejeunea-type branching frequent, with additional pseudodichotomous branching due to production of subfl oral innovations below gynoecia; 950-1280 mm wide and up to 20 mm long, branches initially smaller in stature than parent shoot, attaining similar stature to parent shoot after two or three leaf pairs; older shoot sectors denuded of leaf-lobes. Stems 90-150 μm diameter, with cortical cells in a single tier of 15-22 rows; cortical cell walls brown-pigmented; external free cortical cell wall continuously thickened, radial longitudinal cortical walls thin or slightly thickened, inner tangential walls discontinuously thickened; medulla cells in 12-20 rows, cell walls faintly yellow-pigmented or colourless, with small to medium-sized triangular trigones, walls between trigones unthickened. Cortical cells on dorsal stem surface arranged in straight longitudinal rows on young and mature shoot sectors. Leaf insertion not reaching dorsal stem mid-line, leaving one or two dorsal cortical cell rows leaf-free, dorsal leaf-free strip present; leaf insertion not attaining the ventral stem mid-line, leaving two ventral cortical cell rows leaf-free. Leaf lobes oblong-elliptic, 380-710 μm long by 300-490 μm wide, contiguous, not falcate, acroscopic base lying in plane with stem, plane, not interlocking over the dorsal stem surface, stem visible between leaf lobes in dorsal view; margins irregularly repand, marginal cells bulging, the interior lobe margin not or only weakly ampliate, not or hardly riding onto dorsal stem surface, antical margin curved, exterior margin sharply curved through nearly 100°, postical margin straight; angle between postical lobe margin and keel c. 135°. Lobules rhomboid, remote, one tenth to one eighth the lobe area, 140-350 μm long by 105-240 μm wide; keel straight or rarely slightly arched, angle between keel and stem 135°, keel apex and postical lobe margin fl ush; interior lobule margin free for one third to one half its length, free portion not or weakly ampliate, hardly riding onto ventral stem surface, not concealing the stem in ventral view; acroscopic margin straight or curved; inclined inwards toward the stem; apex acute, laying close to or over the stem margin, free exterior margin straight to weakly curved, margins irregular; lobe-lobule junction level with or slightly postical to the acroscopic end of stem insertion; attached to stem along 0.5-0.33 of the interior margin, stem insertion more or less linear, gently curved at acroscopic and basiscopic ends, not revolute; lobule apex bearing a single papilla, with another two papilla situated on the interior lobule margin above the stem insertion. Leaf lobe cells rounded, not arranged in rows, unequally sized, 9-24 μm long by 9-19 μm wide, thin walled with concave trigones, medial wall thickenings absent; cells of lobe margin smaller than those of leaf middle, quadrate to rectangular, 9-14 μm long and wide, interior and exterior cell walls not diff erentially thickened; leaf lobe cell surface smooth. Oil-bodies not known. Asexual reproduction by caducous leaf lobes, fragmenting into several irregular pieces, marginal lobe cells often proliferating to form bud-like shoot primordial. Dioicous. Androecia not known. Gynoecia terminal on branch shoots, subtended by 2 or 3 subfl oral innovations that are the same size as the branch shoot and are again fertile; archegonia 115-130 μm tall, archegonial neck 6 cell columns; 14-15 per gynoecium on a small disc of tissue, encompassed by the protoperianth; female bracts in one or one and a half pairs, symmetrical, imbricate, narrow oblongelliptic, lobe 655-975 μm long by 265-450 μm wide, margins entire or repand; lobules rhomboid to trullate, one fi fth to one quarter the lobe area, apex obtuse to acute, not or shallowly notched, keel arched, margins irregular; bract insertion lines interlocking dorsally and ventrally, insertion equitant. Perianths c. 3100 μm long and 660 μm wide at mouth, mouth repand, more or less parallel sided for upper third, then tapering to tubular stem perigynium comprising the lower third to half, broadest at mouth, walls 2-or 3-stratose at junction with perigynium, unistratose above. Long stem perigynium present, multi-stratose throughout. Calyptral perigynium present.
Distribution and ecology. Radula anisotoma is currently known only from Norfolk Island where, at the only known location, it occurred in dense forest at 130 m within Mt Pitt Reserve. Th e plants grew beside a creek, presumably within a gully, on a rock admixed with Lejeunea anisophylla Mont., Metzgeria sp. and Radula cf. novaehollandiae (see comments below under R. farmeri Pearson for explanation).
Recognition. Identifi cation of Radula anisotoma should present no diffi culty. Th e most accessible morphological character of R. anisotoma that diff ers from all other members of the R. buccinifera complex is the production of caducous leaves and associated proliferation of marginal lobe cells to form bud-like shoot primordial. Caducous leaves are not produced by any other member of the R. buccinifera complex, with the exception of R. mittenii which diff ers in a number of macro-and micro-morphological characters, and bud-like shoot primordia have only been observed twice, on two different specimens of R. strangulata (M.A.M. Renner pers. obs.).
Radula anisotoma is most similar to R. strangulata and, notwithstanding the diff erences described above, could be confused with that species. However, there are several subtle diff erences between R. anisotoma and R. strangulata in the shape of leaf lobes, lobules, and female bracts, which will aid identifi cation. Th e leaf lobes of R. anisotoma are oblong, whereas they are round in R. strangulata . Th e lobules of R. anisotoma are rhombic to trullate, whereas they are rhombic to longitudinally rectangular in R. strangulata . Th e female bract lobes are narrowly oblong in R. anisotoma , whereas they are elliptic-ovate in R. strangulata . Finally, R. anisotoma is generally a smaller plant than R. strangulata . Comparison with known material is recommended in order to appreciate the degree of diff erence in shape between, as well as variation within, each species when making determinations.
Specimens examined. Australia Description. Forming pure turfs or mats of shoots, dark brown in herbarium; shoot systems regularly pinnately branched, with additional pseudodichotomous branching in female plants due to production of pairs of subfl oral innovations below gynoecia; dimorphic, primary shoots 1.5-1.8 mm wide and up to 40 mm long, secondary shoots smaller in stature than parent shoot, 0.8-1.0 mm wide, and either apparently terminating growth after 4 to 7 leaf pairs, or producing reproductive structures and, in female plants, continuing vegetative growth; older shoot sectors retaining leaf-lobes.
Stems 120-160 μm diameter, with cortical cells in a single tier of 23-29 rows, cortical cell walls yellow-brown pigmented, external free cortical cell wall continuously thickened, radial longitudinal cortical walls thin or slightly thickened, inner tangential walls thickened; medulla cells in 23-45 rows, medulla cell walls faintly yellowpigmented, thin walled, small triangular trigones, medial walls unthickened. Cortical cells on dorsal stem surface arranged in straight longitudinal rows on young and mature shoot sectors. Leaf insertion reaching dorsal stem mid-line, leaving no dorsal cortical cell rows leaf-free; leaf insertion not attaining the ventral stem mid-line, leaving two ventral cortical cell rows leaf-free. Leaf lobes rotund, 475-920 μm long by 400-780 μm wide, contiguous, not falcate, acroscopic base not sharply defl exed away from stem, weakly concave, not or weakly interlocking over the dorsal stem surface, stem visible between leaf lobes in dorsal view; margins entire or crenulated, not repand, the interior lobe margin shallowly ampliate, reaching the opposite stem margin, antical and exterior margins more or less continuously curved, postical margin shallowly curved or straight; angle between postical lobe margin and keel 140-180°. Lobules quadrate on leading shoots, one sixth to one quarter the lobe area, 330-605 μm long by 370-595 μm wide; keel straight to shallowly curved, angle between keel and stem 100-135° , keel turning through up to 30° , keel apex and postical lobe margin fl ush; interior lobule margin free for one quarter to one third its length, free portion ampliate, extending half way across the ventral stem surface or more; acroscopic margin S-shaped to straight, apical portion slightly inclined toward stem or perpendicular to it; apex obtuse but usually weakly apiculate; free exterior margin straight, margins plane, entire; lobelobule junction level with or slightly postical to the acroscopic end of stem insertion; attached to stem along 0.66-0.75 of the interior margin, stem insertion gently curved, not revolute; lobule apex bearing a single papilla, another two papilla situated on the interior lobule margin above the stem insertion. Leaf lobe cells rounded-oblong, not arranged in rows, unequally sized, 13-35 μm long by 11-21 μm wide, thin-walled with small triangular trigones, medial wall thickenings absent; cells of lobe margin smaller than those of leaf middle, quadrate to rectangular, 11-18 μm long by 9-13 μm wide, interior walls moderately and continuously thickened, exterior wall moderately and diff erentially thickened at mid-wall, forming a conspicuous bulge and imparting a crenulate appearance to lobe margin; leaf lobe cell surface unornamented, smooth. Oilbodies 2 or 3, light brown, granular, internally homogeneous, fi lling the cell lumen. Asexual reproduction absent. Dioicous. Androecia on branches that usually terminate after production of 4 or 5 pairs of antheridial bracts, but rarely branches indeterminate, bearing ∞ pairs of antheridial bracts; lobules epistatic, keel deeply curved, bucket-like, free apical portion triangular, apex acute, interior margin ampliate, covering ventral stem surface, and imbricate with adjacent antheridial lobules; lobes rounded, not caducous, antheridia not seen. Gynoecia terminal on branch shoots, subtended by two or three subfl oral innovations that are full-sized and again fertile; archegonia 125-150 μm tall, archegonia neck fi ve cell columns, 10 per gynoecium on a small disc of tissue, encompassed by a low protoperianth; female bracts in one pair, symmetrical, tightly imbricate, elliptic-obovate, weakly falcate, lobe 690-770 μm long by 430-535 μm wide, margins crenulate; lobules rectangular, one half to two thirds the lobe area, apex obtuse, keel straight to arched, margins crenulate; bract insertion lines interlocking dorsally and ventrally, insertion equitant. Perianths 4200-4700 μm long and 1050-1200 μm wide at mouth, mouth entire to irregular, parallel sided for upper two thirds, widening to fl ask shaped, faint bulb in basal third, broadest in middle of this bulb, 1200-1350 μm wide, then tapering to base. Perianth walls unistratose above, with bistratose bands extending up to half way up perianth, increasing in width toward base, becoming confl uent, basal perianth walls progressively increasing in thickness, 2-3-stratose. Long stem perigynium present, 5-6 stratose, cell walls heavily thickened and brown-pigmented. Calyptral perigynium present, base of calyptra 2-4 stratose at base, strata progressively lost, unistratose above, unfertilised archegonia elevated on surface of calyptra.
Etymology. Australian. Distribution and ecology. Radula australiana occurs on mainland Australia (NSW, ACT, VIC), and in Tasmania and New Zealand, usually well above treeline in alpine or subalpine shrublands, grasslands and tussocklands where it grows in association with seepages and running water over and around exposed bedrock and boulders on alpine bluff s, rock outcrops and rock piles. However, R. australiana also inhabits rocky open sites within forest habitats, particularly in Australia where high altitude Eucalyptus forest occurs in the alpine zone, and on bluff s associated with watercourses in New Zealand montane beech forest. Radula australiana is primarily a lithophyte on a wide range of sedimentary, metamorphic, and igneous rocks including greywacke and schist in New Zealand, and granite and basalt in Australia. Microsites occupied by R. australiana are typically sheltered and shaded, such as the back walls of recesses in rock bluff s, crevices, between boulders within ephemeral streambeds. Radula australiana may also occur in bryophyte turfs on soil, usually in the shade of surrounding woody vegetation or rock. It shares all of these habitats with Radula helix , Herzogobryum teres, Nothogymnomitrion erosum, Cheilolejeunea mimosa , and Andreaea spp. Variation. Individuals vary in branching density, New Zealand plants are typically densely branched, and this also occurs in Australia. Openly branched individuals typically have larger shoots and correspondingly larger lobules that produce more pronounced acuminate lobule apices. Th ese diff erences may be associated with both patch age and microsite. Shoots colonizing naked rock are always openly branched. Th ose growing in bryophyte turfs on soil are often closely branched.
Recognition. One of the fi rst clues to the identity comes from the habitat and microhabitat occupied by Radula australiana , as it is one of the few Australasian Radula species that inhabits subalpine and alpine areas, frequently in association with exposed rock. Among alpine species it is one of two having brown pigments, the other being R. demissa M.A.M.Renner. Despite the considerable morphological disparity between them, three unrelated species have been confused with Radula australiana . Radula aneurismalis is yellow-green or orange-green, has a single botryoidal, light brown oil-body per cell, possesses microphyllous branches, has a distinct narrowly infl ated lobule carinal region that extends the length of the keel, produces androecia on short spike-like lateral branches that may be determinate, and the perianths lack a stem perigynium. In contrast R. australiana is brown-green, has two or three granular, brown oil-bodies per cell, does not produce microphyllous branches, though lateral branches may be smaller in stature than primary shoots, has a broadly infl ated carinal region, produces androecia on long branches and on primary axes, and has perianths with a long stem perigynium.
Radula helix is almost as diff erent from R. australiana as is R. aneurismalis . Radula helix is yellow-green, has 3-5 smooth, hyaline, oil-bodies per cell, is paroicous, with androecia immediately below gynoecia, and the perianth lacks a stem perigynium . In contrast, R. australiana is brown-green, has 2-3 granular, brown oil-bodies per cell, is dioicous, and the perianth has a stem-perigynium.
Radula acutiloba is similar in its lobules having an acuminate tip on main shoots, and on this basis Radula australiana was confused with R. acutiloba in Devos et al. (2011) by the fi rst author of this contribution who followed herbarium determinations. Th ough sharing similar lobule shapes, and some phenetic similarity, Radula australiana diff ers from R. acutiloba in lacking subdiscoidal gemmae on the leaf lobe margin, by its brown-green not yellow-green colour, by the presence of two or three light to tan-brown oil bodies in each leaf lobe cell, lack of secondary thickening on cell walls in the stem medulla, and by its occurrence in subalpine and alpine habitats. In all populations of R. acutiloba examined, subdiscoidal gemmae have been present on the leaf lobe margin. Th is feature alone is suffi cient to discriminate between these two species.
Two related species, also members of the R. buccinifera complex, occasionally cooccur with R. australiana and may be confused with it. In Australia, R. buccinifera is typically a forest inhabitant, but at some sites in Victoria and Tasmania where forest occurs over and among exposed granite boulders at high altitude, the ecological envelopes of R. australiana and R. buccinifera overlap, and they may co-occur. Morphological characters by which R. australiana may be distinguished from R. buccinifera are presented in the recognition section for R. buccinifera .   [from NY00831294 and MEL38047] Forming interwoven mats of shoots, brown in herbarium, shoot systems regularly pinnately branched, with additional pseudodichotomous branching due to production of pairs of subfl oral innovations below gynoecia; dimorphic, primary shoots 1.2-1.7 mm wide and up to 40 mm long, secondary shoots smaller in stature and either apparently terminating growth after fi ve to seven leaf pairs, or continuing vegetative growth and attaining similar stature to primary shoots by fourth to sixth pair of leaves; older shoot sectors retaining leaf-lobes.
Stems 130-155 μm diameter, with cortical cells in a single tier of 25-31 rows, medulla cells in 20-35 rows, cortical cell walls yellow-brown pigmented, ventral cortical walls occasionally yellow pigmented, external free cortical cell wall continuously thickened, radial longitudinal cortical walls thin or slightly thickened, inner tangential walls continuously thickened; medulla cell walls faintly yellow-pigmented, with small triangular trigones, walls between trigones lacking thickenings; cortical cells on dorsal stem surface arranged in straight longitudinal rows on young and mature shoot sectors. Leaf insertion variable within single individuals, reaching the dorsal stem mid-line or not, leaving zero to three dorsal cortical cell rows leaf-free, dorsal leaf-free strip usually present; leaf insertion not attaining the ventral stem mid-line, leaving two to fi ve ventral cortical cell rows leaf-free. Leaf lobes rotund-ovate, 600-845 μm long by 400-655 μm wide, contiguous, not to weakly falcate, acroscopic base not sharply defl exed away from stem, plane, not interlocking over the dorsal stem surface, stem visible between leaf lobes in dorsal view; margins irregularly but minutely repand, otherwise entire, the interior lobe margin shallowly ampliate, not or only just reaching the opposite stem margin, antical margin curved, exterior margin sharply curved through nearly 100°, postical margin shallowly curved or straight; angle between postical lobe margin and keel c. 135°. Lobules quadrate to rhombic when small and large, one eighth to one sixth the lobe area, 310-475 μm long by 215-420 μm wide; keel curved in small stature lobules, to straight, to arched in large stature lobules, angle between keel and stem 135°, keel turning through 90° mostly at keel-lobe junction, keel apex and postical lobe margin with shallow notch; interior lobule margin free for one third its length, free portion weakly ampliate small stature lobules to moderately ampliate on large stature lobules, extending at most half way across the ventral stem surface; acroscopic margin S-shaped (typical in situ) to straight (when fl attened), apical portion inclined toward stem; apex obtuse to acute; free exterior margin straight to curved, occasionally with a small knee above the lobe-lobule junction; margins plane, entire or shallowly repand; lobe-lobule junction slightly antical to, or level with, the acroscopic end of stem insertion; attached to stem along 0.66 of the interior margin, stem insertion more or less linear, gently curved at acroscopic and basiscopic ends, not revolute; lobule apex bearing a single papilla, with another two papilla situated on the interior lobule margin above the stem insertion. Leaf lobe cells rounded-oblong, not arranged in rows, unequally sized, 10-23 μm long by 11-19 μm wide; thin walled with small triangular trigones, medial wall thickenings absent; cells of lobe margin smaller than those of middle, quadrate to rectangular, 9-15 μm long and wide, interior and exterior cell walls not diff erential thickened, cell lumen not bulging medially, leaf lobe cell surface unornamented, smooth. Oil-bodies not known. Asexual reproduction absent. Dioicous. Androecia on lateral branches that usually terminate following production of 2-4 pairs of antheridial bracts, occasionally these branches continue vegetative growth; bract lobules epistatic, keel deeply curved, bucket-like, free apical portion triangular, apex obtuse, inner margin ampliate, plane; lobes rounded, not caducous; antheridia 1-2 per bract. Gynoecia terminal on leading shoots, subtended by two subfl oral innovations, usually full-sized and again fertile; archegonia 130-155 μm tall, archegonia neck six cell columns, 6-8 per gynoecium, on a small disc of tissue, interspersed with paraphyses of 1-3 moniliform cells capped by a hyaline papilla, not encompassed by a protoperianth. Female bracts in one pair, symmetrical, tightly imbricate, elliptic-obovate, weakly falcate, lobe 725-845 μm long by 390-610 μm wide, margins entire; lobules rectangular, one half the lobe area, apex obtuse to broadly acute, keel arched, margins entire; bract insertion lines interlocking dorsally and ventrally, insertion equitant. Perianths 2670-3650 μm long and 630-930 μm wide at mouth, mouth repand, more or less parallel sided for upper third, then tapering to tubular stem perigynium comprising the lower third to half, faint bulb in basal third, broadest c. one third from mouth where 660-950 μm wide, walls bi-or tri-stratose at junction with perigynium, unistratose above, cell walls with triangular trigones. Long stem perigynium present, 5-6 stratose throughout, cell walls not thickened or brownpigmented, hyaline, perianth-calyptra junction elevated above female bracts on 9-15 tiers of cells. Calyptral perigynium present, 2-4 stratose at base, unistratose above, unfertilised archegonia elevated on surface of calyptra.
Etymology. Horn-bearing. Distribution and ecology. Radula buccinifera is endemic to Australia, where it is widespread throughout Tasmania, and Victoria but more restricted in New South Wales, being confi ned to the eastern side of the Great Dividing Range and associated escarpments. In Western Australia it is confi ned to the far south-west. Th e northern limit of this species has not been identifi ed, currently the northernmost locality is Point Lookout in New England National Park, in association with Nothofagus moorei . Cool temperate rainforests dominated by Nothofagus also occur in south-east Queensland and R. buccinifera may be be found in these areas.
Casual fi eld observation suggests the elevational range occupied by R. buccinifera is correlated with latitude. In Tasmania and Victoria R. buccinifera occurs across a broad elevational range from sea level to 1500 m, encompassing a range of habitat types from lowland to montane forests, including wet sclerophyll forest and cool temperate rain- forest, wetlands, and in alpine scrub and within habitats may occupy a range of microsites from twigs, branches, tree trunks and tree bases, exposed tree roots on the forest fl oor, to rotting logs, exposed soil on forest banks, dripping rocks adjacent waterfalls, and on rocks within stream beds, sometimes under running water.
Th e ecological range decreases with latitude; the further north, the more restricted to foothills, escarpments, and mountains in association with cool temperate rainforests. Th is altitudinal contraction is associated with restriction in the diversity of microhabitats occupied, such that at the northern end of its range, where R. buccinifera occurs at 1000 m or higher, and is always encountered as a lithophyte most frequently on vertical rock faces associated with bluff s and outcrops.
Radula buccinifera is one of three [ R. novae-hollandiae , R. strangulata , R. buccinifera ] Radula species in south-eastern Australia that may be found growing under running water. On rocks it frequently co-occurs with Radula novae-hollandiae , forming mixed mats. Th e type specimen of R. novae-hollandiae comprises such a mixed patch. In north-eastern New South Wales it may co-occur and form mixtures with a number of other Radula species, initially presenting as confusing and exceptionally variable individuals.
Variation. Individual populations exhibit generous amplitude in shoot stature and lobule shape. Lobules vary from rhombic-quadrate with little ampliation of the interior margin and a straight antical margin, to quadrate with a pronounced ampliate interior margin and an S-shaped antical margin having a distinct, obtuse, lobule apex. Colour of individuals varies, from glaucous yellow-green to dark green, in part this appears correlated to microhabitat, with epiphytic plants tending yellow green, and lithophytic plants being dark green.
Th e amplitude of morphological variation expressed between individuals appears negatively correlated with latitude. Th e greatest morphological variation between individuals in plant colour, shoot size and lobule shape occurs in the southern end of the distribution. Individuals vary from brown-green, mid-green and glaucous green and from small to large, with associated diff erences in lobule shape. In Central and Northern Coast, and Northern Tableland regions in the northern part of R. buccinifera range, individuals are fairly consistent in morphology being large, mid-green, and having lobules whose antical margin is more or less straight, with an pronounced ampliate free interior margin.
Recognition. Despite the variability exhibited by R. buccinifera this species is relatively easy to recognize. Th e fi rst clue to identity comes from the habitat and microhabitat the plants occupy; Radula buccinifera inhabits forest interiors, frequently in microsites on or close to the forest fl oor that are often well shaded. Other species of the R. buccinifera complex with which R. buccinifera could be confused on morphological grounds almost all occupy quite diff erent habitats and microsites.
Radula australiana inhabits subalpine and alpine habitats typically dominated by shrubland, tussockland or grassland, and is almost always found above tree-line. However, R. australiana and R. buccinifera may co-occur in forest interiors on forested mountain tops in Victoria and Tasmania, particularly in association with exposed large granite boulders, for example at Mt. Ellery in Errinundra National Park, Victoria. Th e most accessible morphological character by which R. buccinifera diff ers from R. australiana is the shape of the leaf lobes, where the junction between the lobe and keel forms a simple angle in R. buccinifera and the leaves are not or weakly falcate as a result. In R. australiana the keel runs more or less seamlessly into the lobe outline, the two following the same curve. If this character proves ambiguous, lobule shape is a good source of discriminating characters. Lobule shape should always be assessed on the basis of hydrated, slide-mounted material, and at least 50× magnifi cation. When hydrated the lobules of Radula buccinifera are quadrate to rhombic, and one eighth to one sixth the lobe area. Th e antical margin is usually (northern plants excepted) S-shaped in situ (straight when fl attened), but this is more pronounced with a deep medial curve, and the apex is obtuse. Th e lobules of R. australiana are quadrate to rectangular, and one quarter the lobe area. Th e antical margin is S-shaped in situ and when fl attened, but the medial curve is relatively shallow, and the apex varies from obtuse to acute. Further diff erences exist in the lobe marginal cells that bulge in R. buccinifera , but are crenulate, due to medial thickening on the external cell wall in R. australiana . Finally, if all of these characters prove ambiguous, diagnostic diff erences can be derived by counting the number of rows of dorsal cortical cells that are not crossed by the leaf insertion lines. In R. buccinifera 0-3 cortical stem cell rows are leaf free, whereas no cell rows are leaf-free in R. australiana. Obviously multiple counts from a few shoots are required. If none of these characters leads to a satisfactory conclusion, consider the possibility that the material at hand comprises a mixed collection.
Radula buccinifera may be confused with R. demissa , with which it co-occurs in Victoria and Tasmania where the two species may occupy similar microhabitats. For characters distinguishing these two species, see the recognition section of R. demissa .
Radula buccinifera could also be confused with R. strangulata , as the two species occupy similar microhabitats, both have a dorsal leaf-free strip, and are identical in overall appearance. Lobule shape diff erences are possibly the best, and only, source of discriminating morphological characters. In R. buccinifera the antical lobule margin is typically straight and perpendicular to the stem and the lobule apex is obtuse at least in large lobules. In contrast the antical margin of R. strangulata is either inclined toward the stem or arched and the lobule apex is broadly acute in large lobules. Aquatic morphs of R. strangulata have longitudinally rectangular lobules, whose straight exterior margin runs parallel with the stem and whose interior margin is ampliate over the ventral stem surface. In R. buccinifera the lobule is trapeziform or quadrate, and not ampliate to the same degree. More substantial characters have not been identifi ed. Some interpretations of morphology would fi nd (and have found) strong evidence for the synonymisation of R. buccinifera with R. strangulata from this circumstance. However, here as in other liverwort complexes morphological unity and the existence of morphological continua are immaterial to the phylogenetic structure underpinning the observed morphological diversity, and the existence of populations of both species intermediate in size and shape does not indicate phylogenetic unity (Renner et al. 2013).
Th e name Radula buccinifera has been applied to specimens belonging to a wide variety of unrelated species from tropical Queensland, even as far north as Moa Is- land in the Torres Strait (i.e. CANB9500180). However, with the exception of two specimens collected by Pentzke in 1882, one of which is a mixture of elements from at least two locations (see discussion of the type of R. mittenii ), all records of known provenance from tropical Queensland are based on misidentifi cations. Radula buccinifera can be distinguished from virtually every species in Queensland by 1) the presence of a stem perigynium in the perianth, 2) the production of two subfl oral innovations, 3) the absence of secondary cell wall thickening in the stem medulla, 4) the sub-dimorphic shoot systems, 5) the rhombic lobules, 6) the presence of two papillae on the interior lobule margin, 7) the smooth leaf-lobe cell surfaces. Radula buccinifera may be confused with R. notabilis , and R. imposita ; for distinguishing characters see the recognition section of those species.
Remarks. Radula buccinifera is among the most widely misunderstood species of Radula, having been confused in herbarium collections with almost every other Radula species in Australia, and several that aren't. Misidentifi cations of Radula buccinifera have been the basis for several records of species now excluded from the Australian fl ora, including Radula physoloba and R. plicata . Confusion with R. physoloba seems attributable to R.A. Bastow, all of whose specimens identifi ed as R. physoloba in MEL ex herb. R.A. Bastow are R. buccinifera , and confusion with R. plicata is attributable to Stephani, who misidentifi ed several gatherings by Weymouth and Bastow.
Radula buccinifera was understandably confused with R. strangulata by Mitten (1855) who cited collections by Colenso, Stephenson, and Lyall from the southern North Island in his treatment for Hooker's Flora of New Zealand, and this confusion has been perpetuated in various manifestations. Allison and Child (1975) illustrated R. strangulata as R. buccinifera , and again in the next fi gure as R. levieri Steph. Renner (2005) made a diff erent, and less forgivable, error when he presented a key to New Zealand species that included R. buccinifera . Th e plants he referred to were actually Radula demissa M.A.M.Renner, and his error was based on mis-interpretation of type material.
Castle, however, appears to have been hopelessly confused regarding the identity of R. buccinifera , arbitrarily placing specimens of R. buccinifera , R. demissa , and R. strangulata under the names R. buccinifera , R. mittenii , and R. wattsiana . As an example, the fi gure illustrating R. buccinifera in Castle (1967) includes a male shoot of Radula strangulata . Castle maintained that Radula mittenii diff ered from R. buccinifera in its more falcate leaf lobes, and the shorter perianths and he may have had R. demissa in mind in this assessment. Yamada (1984) synonymised R. mittenii with R. buccinifera on the basis of his examination of 'holotype' material held in g, and his assertion that the diff erences observed were environmental. Yamada's identifi cation of holotype material is incorrect, his interpretation was derived from examination of duplicate material in Geneva and the implications of this are discussed below under R. mittenii . Th e synonymisation of R. mittenii , having a type from tropical Queensland, with R. buccinifera by Yamada (1984) seems to have encouraged the application of the name R. buccinifera to a range of specimens from tropical Queensland, to the point where nearly every species occurring in Queensland has, at some point, been identifi ed as R. buccinifera . Beyond the specimen collected by Pentzke, no populations attributable to R. buccinifera from tropical Queensland exists within herbaria, and the species has not been observed during the course of recent fi eldwork in the Wet Tropics of Queensland.
Nomenclature. So (2005) identifi ed a Hooker collection from Van Diemens Land as the holotype of R. buccinifera . Th is is inappropriate as Hooker and Taylor (1844) did not identify a single type collection, the only specimen details given in the protologue were "Van Diemen's Land". Collections from "Van Diemen's Land" by both R. C. Gunn and J. D. Hooker were made prior to 1844 and were studied by Hooker and Taylor for their contribution to Hepaticae Antarcticae, as made clear in the subtitle to Hooker and Taylor's (1844) publication, and by the presence of duplicates from both collectors in Taylors own herbarium. At least three gatherings may comprise the syntype series, being the more or less pure gatherings made by Gunn and Hooker, and a gathering on Sticta , whose collector has not been recorded. Furthermore, Castle (1967) had attempted to lectotypify R. buccinifera , though we are unclear about exactly what Castle intended in his lectotypifi cation. Castle's (1967) lectotypifi cation of R. buccinifera is somewhat obscured by his own commentary. In the list of specimens examined he states 'Van Dieman's Land, the Type (BM K and Y) and Van Dieman's Land, R.C. Gunn (BM and K)' [italics ours]. Here Castle appears to identify two groups of specimens, one of which was collected by Gunn and was not identifi ed as the lectotype. But Castle then notes 'we may assume that those collections which bear the label Van Diemen's Land are portions of the collection upon which Th omas Taylor based his Jungermannia buccinifera . Several samples of this type, preserved in the herbaria of the British Museum and of Kew Gardens, also include in the label the name Gunn. Gunn presumably made the type collection as his name, as collector, appears in the title of the article in which the original description of J. buccinifera was published, without data'. Here Castle implies that the Gunn collection should be regarded as the lectotype. All gatherings comprising the syntype series we have seen, being duplicates of both Hooker and Gunn's collections, and the specimen sine. coll. are unambiguously assignable to R. buccinifera , so application of the name Radula buccinifera will remain the same regardless of which gathering and which particular specimen the species is lectotypifi ed on. Castle's lectotypifi cation at best requires narrowing because he did not identify a single specimen as type, and at worst lectotypifi cation requires repeating because it is not quite clear whether Castle identifi ed a single gathering.
Specimens examined. Australia Diagnosis. Similar to Radula buccinifera and R. strangulata but diff ers from both species in the leaf insertion attaining the dorsal stem mid-line, leaving no dorsal cortical cell rows leaf-free, the leaf lobes overlapping across the dorsal stem, such that the stem is not visible from above, the falcate leaf lobes, the rhombic lobules with antical margin sloping inwards toward the stem at an angle of c. 45°, and in its typically epiphytic and epiphyllous habit.
Description. [from CHR587329 and AK280339] Forming hanging wefts or patches of interlocking shoots, brown-green when fresh, fading to a glossy tan or brown in herbarium; shoot systems regularly pinnately branched in male plants and sterile female plants, but pseudodichotomous in fertile female plants due to production of pairs of subfl oral innovations below gynoecia; dimorphic, 1.0-2.0 mm wide and up to 40 mm long, branches smaller in stature than parent shoot; older shoot sectors retaining leaf-lobes. Stems 115-175 μm diameter, with cortical cells in a single tier of 20-35 rows. Cortical cell walls yellow-brown to brown pigmented; external free cortical cell wall heavily and continuously thickened, radial longitudinal cortical walls thin, inner tangential walls thin or continuously thickened; medulla cells in 22-43 rows; cell walls yellow-brown pigmented, with small to medium triangular trigones, walls between trigones lacking thickenings. Cortical cells on dorsal stem surface arranged in straight longitudinal rows on young and mature shoot sectors. Leaf insertion reaching dorsal stem mid-line, leaving no dorsal cortical cell rows leaf-free, except in instances where stem growth appears to have introduced an additional cortical cell row to the dorsal stem surface, in which case a single row of dorsal cortical cells is leaf-free, but this row is discontinuous between adjacent leaf pairs; leaf insertion not attaining the ventral stem mid-line, leaving two or three ventral cortical cell rows leaf-free. Leaf lobes ovate-falcate, 650-1100 μm long by 430-900 μm wide, larger on leading shoots, contiguous to imbricate, falcate, acroscopic base not sharply defl exed away from stem, fl at, obliquely patent, interlocking over the dorsal stem surface, stem not visible between leaf lobes in dorsal view; margins irregularly but minutely repand, entire to weakly minutely crenulate, the interior lobe margin ampliate, covering the dorsal stem surface and reaching or exceeding the opposite stem margin, antical margin curved, exterior margin curved, postical margin curved; angle between postical lobe margin and keel 60-90°. Lobules rhombic when small to widely rhombic, one eighth to one sixth the lobe area, 255-535 μm long by 180-385 μm wide, larger on leading shoots; keel straight to curved, angle between keel and stem 135°, keel apex and postical lobe margin notched; interior lobule margin free for one fi fth to one third its length, free portion ampliate, extending at most half way across the ventral stem surface; acroscopic margin S-shaped, apical portion inclined towards stem; apex rounded to apiculate-rounded; free exterior margin curved, margins plane, entire; lobe-lobule junction antical to the acroscopic end of stem insertion in small lobules to level with or below it in larger lobules; attached to stem along 0.66 to 0.8 of the interior margin, stem insertion curved, particularly at basiscopic end, not revolute at acroscopic end; lobule apex bearing a single papilla, with another two papilla situated on the interior lobule margin above the stem insertion. Leaf lobe cells rounded-oblong, not arranged in rows, unequally sized, 14-25 μm long by 9-18μm wide, thin walled with small triangular trigones, medial wall thickenings absent; cells of lobe margin smaller than those of leaf middle, quadrate to rectangular, 9-15 μm long and wide, interior and exterior cell walls not diff erential thickened, cell lumen not or only slightly bulging medially; leaf lobe cell surface unornamented, smooth or weakly bulging; upper lobe wall diff erentially thickened over cell lumen, forming a weak papilla. Oil-bodies two or three per cell, ellipsoidal, fi lling cell lumen, light-brown, surface granular, internally homogeneous or with a hyaline droplet. Asexual reproduction absent. Dioicous. Androecia on indeterminate branches that continue vegetative or reproductive growth, androecial bracts in 4-20 pairs, lobules epistatic, keel deeply curved, bucket-like, free apical portion triangular, apex obtuse, moderately defl exed, lobes rounded, not caducous, antheridia not seen. Gynoecia terminal on leading shoots, subtended by two subfl oral innovations that are full-sized and again fertile; archegonia 145-190 μm tall, archegonia neck six cell columns, 6-8 per gynoecium on a small disc of tissue, encompassed by the protoperianth; female bracts in one pair, symmetrical, tightly imbricate, ovate-falcate, lobe 870-940 μm long by 450-560 μm wide, margins entire to repand; lobules ovate to trullate, one third to one half the lobe area, apex rounded to obtuse, keel arched, margins entire; bract insertion lines interlocking dorsally but not ventrally, insertion equitant. Perianths 2300-3900 μm long and 710-760 μm at mouth, mouth entire to irregularly lobed, perianth narrowing gently from slightly fl ared mouth to approximately one third to one half length above base, where 670-790 μm wide, then tapering to narrow cylindrical base, often infl ated independent of sporophyte growth, in which case perianth labia roll inward to overlap one another, rather than laying in plane, the inrolling may cause perianths to appear weakly bicornute, perianth walls unistratose above, with bistratose collar 3 or 4 cell tiers high above the perianth-calyptra junction; basal stem perigynium present, 5-6 stratose, cell walls not thickened or pigmented; perianth-calyptra fusion elevated above female bracts on 9-15 tiers of cells; calyptral perigynium present, base of calyptra bistratose, unistratose above, unfertilised archegonia elevated on surface of calyptra.
Etymology. From Latin demissa, hanging, in reference to the hanging fan shaped wefts formed by this species when growing on twigs and branches, also in reference to the frequent colonisation of leaves of the fern Hymenophyllum demissum .
Distribution and habitats. South-eastern Australia in Victoria and Tasmania, and New Zealand. On mainland Australia R. demissa occurs only on tablelands and highlands in southern and eastern Victoria, but in Tasmania the species is widespread. Radula demissa grows in cool temperate and warm-temperate rainforests and occasionally in wet sclerophyll forests, particularly in the southern part of its range.
In New Zealand R. demissa occurs throughout all three main islands, and extends east to the Chatham Islands and south to the Auckland and Campbell Islands. It inhabits all forest environments, but is rare in coastal forests with high exposure to salt-laden winds. Radula demissa occupies a broad elevation range, from the coast to around 1600 m asl, where it grows in subalpine scrub above the treeline.
In forests Radula demissa is typically an epiphyte on tree trunks, branches and twigs where it forms characteristic fan-shaped hanging wefts, and is frequently encountered in riparian areas and edge habitats where local light environments are brighter than adjacent forest interiors. Common host species in New Zealand include Melicytus ramifl orus and Beilschmiedia tawa . Radula demissa is also commonly encountered as an epiphyll on the leaves of ferns and broadleaved shrub and tree species. In New Zealand Hymenophyllum species, including H. demissum , H. ferrugineum and H. scabrum are favorite fern hosts, and Beilschmiedia tawa is a frequent broadleaf host. In lowland to montane podocarp-broadleaf forest of tall stature with sparse sub-canopy and shrub layers R. demissa can be found as a trunk epiphyte and epiphyll on B. tawa , and as an epiphyll on leaves of H. demissum where this species grows in dense carpets on the forest fl oor. On bark Radula demissa is often associated with R. allisonii, R. plicata, R. strangulata , Metzgeria spp. and various species of Lejeuneaceae including Drepanolejeunea aucklandica and Metalejeunea cucullata , Sematophyllaceae, various Plagiochilaceae species including Dinckleria pleurata , Frullania spp. Co-occurrence with phenetically similar species can pose a challenge to identifi cation. On twigs R. demissa grows with Plagiochila colensoi , Lepidolaena taylorii, L. palpebrifolia , R. plicata , and on leaves R. demissa grows with Echinolejeunea papillata , Cololejeunea laevigata , and members of the Lejeunea epiphylla Colenso aggregate.
In subalpine scrub R. demissa occurs on damp rock and adjacent soil, sometimes in highly insolated microsites. Alpine plants exhibit some subtle morphological diff erences from forest plants, as discussed in the next section.
Variation. Included within Radula demissa are forms that grow on wet rocks in alpine habitats, typically between 800 and 1600 m asl. Alpine forms exhibit subtle morphological diff erences from forest plants. Th e leaves are remote to contiguous rather than imbricate, and the lobes posses a small auricle at the dorsal base of the stem insertion, which is not found in forest plants. Th e carinal region is typically narrowly infl ated along the length of the keel, rather than more broadly infl ated across the width of the lobule. Although alpine plants expressing these features are known from several sites in New Zealand, alpine plants are not currently known from Tasmania or mainland Australia.
Th e existence of subtly diff erent, ecologically distinct, alpine plants of R. demissa could be explained in several ways. Alpine plants could be an ecotype within a reproductively cohesive species that also occupies forests. Th ey could be a partially reproductively isolated population that retains genetic contact with forest populations. Th ey could be fully reproductively isolated, morphologically cryptic species. Th ese alterna- tives remain untested. Alpine accessions nest within R. demissa in our phylogeny, but may warrant investigation through additional genetic, ecological, and physiological data. Given the species straddles what are eff ectively diff erent biomes (forest and alpine environments: Crisp et al. 2009) the transition from forest to alpine environments concomitant with a change in microsite occupancy could represent a striking example of phenotypic plasticity.
Recognition. Despite being confused with a range of related and unrelated species, Radula demissa is easy to recognise. Th e fi rst clue to identity comes from the habitat and microhabitat the plants occupy. Radula demissa is typically an inhabitant of well-lit microsites, and is an epiphyte or epiphyll in forest, and a lithophyte in subalpine shrublands and grasslands. Th ree other species of the R. buccinifera complex occupy similar microsites across the range of R. demissa, and may co-occur with it.
In Australia, Radula buccinifera may co-occur with R. demissa on tree trunks. Diff erentiating these two species is only possible on the basis of hydrated, slidemounted material because deformation associated with drying renders void examination and interpretation of leaf-lobe and lobule shape and orientation in dry material. Th e most accessible character by which Radula demissa and Radula buccinifera diff er is in the orientation and spacing of the leaf lobes, however diff erences are not as clear cut as between other pairs of species (i.e. between R. demissa and R. strangulata ) and this character may not prove reliable in all instances. In R. demissa the leaves are contiguous to imbricate, and obliquely patent, spreading up and away from the substrate so that they overlap across the dorsal stem surface, such that the dorsal stem surface is not usually visible from above. In R. buccinifera the leaves are patent to obliquely patent, and while they may spread upward away from the substrate, do not usually overlap completely across the dorsal stem surface so the stem can usually be seen from above, at least in part. However, as there is variation in both leaf lobe orientation and how falcate the lobes are in both species, this character should be used as a guide only. Leaf shape provides another useful clue to identity. In R. demissa the leaves are falcate, with a distinct notch at the junction of the lobe and keel. In R. buccinifera the leaves are not or only weakly falcate, and the junction between the lobe and keel forms a simple angle. Lobule shape is also a good source of diagnostic diff erences, though the diff erences between these two species are subtle. For convenience, some conceptual re-orientation of the shoot is required to describe these diff erences clearly. Imagine the shoot is being held vertically with the apex top-most. In this orientation the lobule apex of R. demissa always lies well above the uppermost point of the ampliate portion of the lobule margin. Between these two points the lobule margin is shallowly S-shaped, and this slopes downward at c. 45° toward the stem. In R. buccinifera the lobule apex lies variably between the same level as, or slightly above, the uppermost point of the ampliate portion of the lobule margin. Between these two points the lobule margin varies from straight to S-shaped in situ (straight when fl attened), but when S-shaped there is a pronounced medial curve. Th e slope of the antical margin varies between sloping downward toward the stem at up to 45° and remaining level. Sev-eral shoots should be examined to gauge the patterns of variation in lobule shape, and comparison with known material is recommended.
Th e lobule keel diff ers between R. demissa and R. buccinifera . In R. demissa the keel is curved, on lobules from both main shoots and branches. In R. buccinifera the keel is typically straight, but may be curved on some lobules, particularly those on branches.
If these more accessible characters prove ambiguous, diagnostic diff erences can be derived by counting the number of rows of dorsal cortical cells that are not crossed by the leaf insertion lines. In R. demissa no rows are leaf-free, whereas zero to three rows are leaf-free in R. buccinifera. To achieve this count, leaves must be removed from the stem in such a way as to preserve the leaf insertion yet make the dorsal stem surface visible, which is a challenging proposition at the best of times.
Radula demissa may co-occur with Radula strangulata on tree trunks and on living leaves close to the forest fl oor. For guidance on separating these two species on morphological grounds see the recognition section of R. strangulata .
Radula demissa has been confused with the New Zealand endemic Radula plicata . Radula demissa may co-occur with Radula plicata on tree trunks, branches, and leaves. Although R. plicata is not closely related to R. demissa , confusion may arise due to the similarity in size and shape of lobes and lobules, in the curved keel with infl ated carinal region, the S-shaped antical lobule margin, and the falcate leaf lobes. When fertile female plants are encountered Radula plicata is readily separated by the plicate perianth surfaces. However, sterile and male herbarium material may cause problems if gross morphological characters are relied on for identifi cation. Radula demissa is readily separable from sterile and male material of R. plicata by its unornamented leaf-lobe cell surfaces, and bulging, rather than crenulate leaf-lobe margin. In R. plicata the leaf-lobe cells bear conspicuous coarse granular ornamentation on their surface. Th is is easily observed in hydrated slide-mounted material on the lobe marginal cells. Th e cells of the lobe margin are also distinctly crenulate, although comparison with known material is recommended as a guide to the degree of diff erence between these and the bulging cells of R. demissa . Diff erences in stem anatomy also diff erentiate these two species. In R. demissa the cortical stem cell walls are brown-pigmented, and the medulla walls are yellow-pigmented. In R. plicata the cortical and medulla stem cells walls are unpigmented. If living material is being examined, diagnostic diff erences may be found in the oil-bodies. In R. demissa there are 2-3 large, light brown, internally homogeneous oil-bodies that have a granular surface and completely fi ll the cell lumen of each cell. In R. plicata there are 3-5 small, clear or grayish oil bodies with a smooth surface each containing a hyaline droplet, that are arranged in a loose submarginal ring and do not fi ll the lumen in each cell.
In alpine environments, R. demissa may be confused with R. australiana . However, these two species are readily separated by diff erences in lobe shape that, with care, can be accessed in the fi eld. In R. demissa the leaf lobes are falcate, with a notch at the lobe-lobule junction. In R. australiana the leaf lobes are not falcate, rather the lobule keel runs more or less seamlessly into the lobe outline. Lobule size and shape also diff ers between these two species. In R. demissa the lobules are rhombic to widely rhombic and one eighth to one sixth the lobe area. In terms of the same conceptual re-orientation described above, the lobule apex of R. demissa lies above the ampliate inner lobule margin. In R. australiana the lobules are quadrate and one sixth to one quarter the lobe area, with the lobule apex lying level with the ampliate inner lobule margin. Furthermore, the ampliate portion of the lobule is typically much larger, and the lobule apex more acute in R. australiana than in R. demissa .
Perianth morphology may lead to confusion between R. demissa and R. ratkowskiana , particularly given comments by  that inrolling perianth labia and the associated bicornute perianth is unique to R. ratkowskiana . Th is is not the case, both features may be developed in young perianths of R. demissa , which can be distinguished from R. ratkowskiana by the small rhombic lobules, and the absence of domeshaped papillae on the leaf lobe cells.
Remarks. Mitten (1855) was the fi rst to record R. buccinifera for New Zealand, citing collections by Colenso, Stephenson and Lyall. Hodgson (1944) evidently followed Mitten in accepting R. buccinifera for New Zealand, but in the absence of an examined type, equated R. demissa with R. buccinifera . Hodgson's (1944) description of R. buccinifera having glossy dried plants with falcate leaf lobes whose margins extend over the stem and rhomboid lobules agrees with R. demissa , and on the basis of these characters Hodgson accurately distinguished R. demissa (as R. buccinifera ) from R. strangulata (as R. levieri for which she had examined type material).  followed Hodgson's (1944) conception of R. buccinifera , and the plants presented in his key as R. buccinifera also correspond to R. demissa . Th is error was made despite the fact that he had examined syntype material of R. buccinifera held in bm in early 2004, apparently the diagnostic diff erences in leaf lobe and lobule shape were overlooked. As discussed under R. buccinifera , Mitten's New Zealand record of that species must be rejected.
In Australia R. demissa has remained unrecognized, being subsumed within an increasingly broad working circumscription of R. buccinifera , as discussed under that species.
Colenso's description of a bipinnately branched plant with dimorphic shoots, 'peduncled' perianths with a truncate mouth with slightly uneven labia, and subtended by two divergent subfl oral innovations, growing epiphytic on Hymenophyllum is consistent with R. demissa . However, no defi nite type material for Colenso's R. epiphylla has been located, and our attribution of this invalid name to synonymy under R. demissa is tentative. Diagnosis. Similar to both R. buccinifera and R. demissa and could easily be mistaken for a poorly developed morph of either, but diff ers in the crenulate leaf-lobe margins and low dome-shaped papillae over each leaf-lobe cell. Diff ers also in the relatively small size, the densely but irregularly pinnate branching pattern, and the irregular margins to the female-bract lobes, which are also crenulate.
Type Description. [From NSW896812] Forming tufts of overlapping and interlocking shoots on leaves and twigs, bright-to mid-green when fresh, fading to a glossy tan or brown in herbarium; shoot systems densely irregularly pinnately branched in female plants, with additional pseudodichotomous branching in fertile female plants due to production of pairs of subfl oral innovations below gynoecia; monomorphic, 0.8-1.2 mm wide and up to 30 mm long, branches smaller in stature than parent shoot; older shoot sectors retaining leaf-lobes. Stems 80-130 μm diameter, with cortical cells in a single tier of 15-30 rows; cortical cell walls yellow-brown to brown pigmented; external free cortical cell wall heavily and continuously thickened, radial longitudinal cortical walls thickened or not, inner tangential wall more or less continuously thickened; medulla cells in 10-25 rows; cell walls yellow-brown pigmented, with medium-sized triangular to weakly bulging trigones, walls between trigones lacking thickenings or continuously thickened; cortical cells on dorsal stem surface arranged in straight longitudinal rows on young and mature shoot sectors. Leaf insertion reaching dorsal stem mid-line or not, leaving zero or one dorsal cortical cell rows leaf-free, in some shoots stem growth appears to have introduced an additional cortical cell row to the dorsal stem surface, in which case a single row of dorsal cortical cells is leaf-free, but this row is discontinuous between adjacent leaf pairs; leaf insertion not attaining the ventral stem mid-line, leaving two to four ventral cortical cell rows leaf-free. Leaf lobes elliptic-ovate, 430-700 μm long by 320-450 μm wide, larger on primary shoots, contiguous to imbricate, not to weakly falcate, acroscopic base not sharply defl exed away from stem, fl at, patent, not interlocking over the dorsal stem surface, stem visible between leaf lobes in dorsal view; margins irregularly but minutely repand, crenulated due to medial wall thickenings on marginal cell walls, the interior lobe margin weakly ampliate, covering part of the dorsal stem surface but not often reaching the stem midline, and never exceeding the opposite stem margin, antical margin curved, exterior margin curved, postical margin curved to straight; angle between postical lobe margin and keel 45-60°. Lobules rhombic when small to almost quadrate, one sixth the lobe area, 175-345 μm long by 150-265 μm wide, larger on leading shoots; keel straight to curved, angle between keel and stem 135°, keel apex and postical lobe margin running into lobe margin smoothly or shallowly notched; interior lobule margin free for one eighth to one sixth its length, free portion not or weakly ampliate, not or hardly extending onto the ventral stem surface; acroscopic margin shallowly S-shaped, apical portion inclined towards stem; apex apiculate-rounded; free exterior margin straight to curved, margins plane, entire; lobe-lobule junction level with the acroscopic end of stem insertion in small and large lobules; attached to stem along 0.83 to 0.88 of the interior margin, stem insertion straight to curved, not revolute at acroscopic end; lobule apex bearing a single papilla, with another two papilla situated on the interior lobule margin above the stem insertion. Leaf lobe cells rounded-oblong, not arranged in rows, unequally sized, 11-22 μm long by 8-16 μm wide, walls slightly thickened their entire length, sometimes with small triangular trigones, medial wall thickenings absent; cells of lobe margin smaller than those of leaf middle, quadrate to rectangular, 9-14 μm long and wide, interior walls not diff erentially thickened, exterior walls with pronounced medial thickening, and medially bulging cell lumen; leaf lobe cell surface papillose, upper lobe wall diff erentially thickened over cell lumen forming a single low dome-shaped papilla over each cell. Oil-bodies not known. Asexual reproduction not known. Dioicous. Androecia not known. Gynoecia terminal on leading shoots, subtended by two subfl oral innovations that are full-sized and again fertile; archegonia 145-160 μm tall, archegonia neck fi ve or six cell columns, 6-8 per gynoecium on a small disc of tissue, encompassed by the protoperianth; female bracts in one pair, slightly asymmetrical with lobule on one larger than the other, imbricate, ellipticovate, lobe 520-570 μm long by 240-320 μm wide, margins irregular and crenulate; lobules rectangular-ovate, one half to two thirds the lobe area, apex rounded to obtuse, keel arched, irregular.
Etymology. Imposita : laid-upon, in reference to both the epiphyllous habit and the way the small, densely packed branches grow over one another in the type.
Distribution and ecology. Known to date from four specimens, all growing over running streams on either leaves or bark including tree-trunks and branches. Th e specimens were collected in the North Coast of New South Wales, and in the Wet Tropics Bioregion of north-east Queensland.
Variation. Within its already reduced size, in comparison to its relatives, R. imposita exhibits limited variation in shoot size and lobule morphology within individuals, and much more variation between individuals. However, knowledge of this species is based on only four specimens, from extremes of a geographical distribution spanning 15 degrees of latitude, so more variation than described here should be anticipated.
Recognition. Radula imposita could easily be mistaken for a poorly developed morph of either R. buccinifera or R. demissa . Th e ecological envelope and geographical distribution of both species exhibits some overlap with R. imposita. However, the crenulate leaf-lobe margins and low dome-shaped papillae over each leaf lobe cell are distinctive and will immediately distinguish R. imposita from both R. demissa and R. buccinifera . Several other features are distinctive of R. imposita including the relatively small size, the dense irregularly pinnate branching pattern with almost monomorphic shoots, and the irregular and crenulate margins to the female-bract lobes.
[From NSW897201] Forming extensive pure sheets of interwoven pendulous shoots on tree trunks and rocks; living plants opaque yellow-green to glaucous brown-green, fading to milky pale-brown in herbarium; shoot systems regularly pinnate, subdimorphic, with shoots 1.6-2.5 mm wide and up to 80 mm long, branches typically slightly smaller in stature than parent shoot; older shoot sectors becoming ragged in appearance due to irregular leaf fragmentation. Stems 190-250 μm diameter, with cortical cells in a single tier of 30-50 rows; cell walls brown pigmented throughout; cortical cell walls heavily and continuously thickened, at times constricting the cell lumen, dorsally arranged in an oblique zig-zag on young shoot sectors, cell elongation somewhat obscuring this pattern in mature shoot sectors; medulla cells in 80-110 rows, cell walls heavily thickened with coarse nodular trigones that become confl uent, and constrict the cell lumen, thin walls occasional. Leaf insertion exceeding dorsal stem mid-line, insertion lines interlocking over two dorsal cortical cell rows, dorsal leaf-free strip absent. Leaf lobes oblong-falcate, 770-1360 μm long by 560-870 μm wide, contiguous to weakly imbricate, acroscopic base sharply defl exed away from stem, otherwise leaves weakly convex, not interlocking over the dorsal stem surface, stem visible in dorsal view, margins may be irregular in outline but always entire, the interior lobe margin curved, not auriculate, antical margin shallowly curved, exterior margin broadly rounded, postical margin straight or substraight. Lobules on leading shoots typically one quarter the lobe area, more or less quadrate, 430-720 μm long by 510-750 μm wide, keel straight to shallowly arched, angle between keel and stem 135°, keel turning through 45-55° at the apex; interior free margin weakly ampliate, acroscopic margin straight or shallowly arched, usually more or less perpendicular to shoot axis, and apices obtuse or broadly acute; attached to stem along 0.4-0.5 of the interior margin, stem insertion gently S-shaped but abruptly revolute at acroscopic end; lobule apex bearing a single papilla, with another papilla situated on the interior lobule margin above the stem insertion; lobules on leading shoots typically larger than those on branches, lobules on branches more rhomboid than quadrate, one fi fth to one quarter the lobe area, keel shallowly arched to straight to slightly curved, angle between keel and stem 135°, keel turning through 45-55° at the apex, interior free margin weakly ampliate, acroscopic margin typically S-shaped to shallowly arched, inclined to shoot axis, apex typically broadly acute. Leaf lobe cells rotund to roundedoblong, 19-28 μm long by 15-23 μm wide, thin walled with concave to triangular trigones, medial wall thickenings absent; cells of lobe margin smaller than those of leaf middle, quadrate to rectangular, 10-15 μm long by 7-11 μm wide, long axis orientated parallel to lobe margin, exterior cell walls each with a medial wall thickening that bulges into the cell lumen margin; cell surface weakly bulging, bearing heavy verrucose ornamentation. Oil-bodies not known. Asexual reproduction by caducous leaf lobes, sporadic, typically but not always on old shoot sectors, fragmentation scars irregular, shoot primordia forming as irregular buds on leaf lobe margins in older shoot sectors prior to leaf fragmentation. Dioicous. Androecia on lateral branches that continue vegetative growth, androecial bracts in 3-5 pairs, smaller than vegetative leaves, lobes 510-660 μm long and 360-430 μm wide, ovate, imbricate, lobules hypostatic, keel deeply curved, bearing 1-2 antheridia each. Gynoecia terminal on axes, with one pair of female bracts subtended by one (on branches) or two (on leading shoots) full sized subfl oral innovations that may again be fertile; archegonia 160-200 μm tall, archegonia neck fi ve cell tiers, cells regularly arranged, 18-20 per gynoecium on a small raised disc of tissue encompassed by the base of the protoperianth; female bracts equal, ovate-falcate, lobes 1200-1310 μm long by 685-870 μm wide, lobules ovate, one half the lobe area, apex rounded to obtuse, keel strongly arched, insertion interlocking both dorsally and ventrally, insertion equitant. Perianths 3000-3500 μm long and 800-900 μm wide at mouth, cyathiform, fl aring widely from a narrow base, which is a low stem perigynium 4-5 stratose with brown-pigmented walls, broadest immediately above base, straight-sided, gradually tapering to the mouth, which has irregularly sinuous labia; perianth walls 2-3 stratose at base, unistratose above; calyptral perigynium present, unfertilised archegonia 'riding' onto base of calyptra, calyptra 2-3 stratose at base, tapering to unistratose above.
Etymology. Named for William Mitten (1819-1906) pharmaceutical chemist and bryologist, contemporary of and collaborator with W.J. and J.D. Hooker, fatherin-law to Alfred Russell Wallace; and author of the fi rst fl oristic treatment of liverworts of New Zealand in Hooker's Handbook of the New Zealand fl ora.
Distribution and ecology. In Australia R. mittenii is known from a range of localities in the Wet Tropics Bioregion of north-eastern Queensland. Radula mittenii occurs over a broad elevational range, from near sea level to 1500 m, encompassing an array of tropical forest types from lowland mesophyll forests on river fl oodplains, to cyclone disturbed forests on hillslopes, to notophyll-vine forests on summit peaks. Within these habitats Radula mittenii occupies a range of microsites, from boulders, to tree trunks, and branches, twigs, and liane stems. As a lithophyte on the sides of large boulders R. mittenii may form extensive pendulous mats of milky yellow-green shoots.
Variation. Within individuals the dimorphic nature of shoot systems means lobules on primary shoots diff er in size and shape from those on secondary shoots, a fea- ture fi rst documented by Renner et al. (in press) that is widespread among species of subg. Radula, Volutoradula and Cladoradula . Lobules on secondary shoots tend toward transverse rectangular shape, and the development of a refl exed antical lobule margin, which causes the lobule apex to appear acute. Plants from exposed situations, including those growing on branches and twigs, may appear to consist only of tangled shoots lacking architectural regularity, bearing lobules of this type, and so diff er considerably from the lush, regularly pinnate plants found on rocks and tree-trunks in forest interiors. Close inspection of plants from exposed situations will reveal a primary shoot closely appressed to the substrate from which secondary branches arise. Comparison with plants from sheltered situations will reveal commonalities in the shape and size of lobules from secondary shoots.
Recognition. Among Australasian Radula , R. mittenii is easily recognized by its large size, regularly pinnate branching, more or less quadrate lobules, verrucose ornamentation on leaf-lobe cell surface, and the leaf-insertion interlocking over the dorsal stem mid-line. highly distinctive, and easily recognized species. Th e outstanding feature of R. mittenii is the verrucose ornamentation that it imparts a milky yellow appearance to plants in life and in herbaria. However, colour is not always a reliable guide to identity.
Radula mittenii shares with R. sumatrana Steph. A number of characters, including shoot size and branching architecture, the verrucose cuticle, and the large more or less quadrate lobules, and the two are probably closely related. However, the type of R. sumatrana (G00112217!) diff ers in that the lobules tend to become much larger, imbricate, and obscure the stem surface in ventral view. Th is is consistent with other specimens seen from Borneo. Yamada describes stem cell walls of the type of R. sumatrana as being thin-walled with large trigones. Th in walls in the stems of Australian material examined are rare, and in general trigones are more confl uent and less discrete than illustrated by Yamada (1974). Radula sumatrana occurs throughout South East Asia from Th ailand to Indonesia including Sumatra, Java and Borneo (Yamada 1979), and plants corresponding to it do not occur in Australia. However, in the unlikely circumstance that these two names do refer to the same entity, R. mittenii has priority. Some forms of R. mittenii bear resemblance to the type specimen of R. similis (G00042714!) in their lobule shape and smaller stature. However, R. similis , a taxon from New Caledonia, has apparently smooth leaf lobe cell surfaces, so is readily distinguished on this basis. However, granular ornamentation in some species of subg. Radula can be very diffi cult to detect with some light microscopes, and we may have missed this feature. At any rate if ornamentation is present, it is not as pronounced as in R. mittenii .
Radula mittenii can be diff erentiated from members of the R. buccinifera complex on the basis of many characters, including 1) the verrucose lobe cell surfaces, in R. buccinifera the lobe cell surfaces are smooth or bear low dome-shaped papillae; 2) the presence of heavy, continuous, irregular, brown-pigmented secondary thickenings on the medullar cell walls of the stem formed by fusion of nodular trigones, in R. buccinifera the medullar walls are pigmented or not, and if secondarily thickened bear trigones that rarely fuse to form irregular continuous thickenings, 3) the large quadrate lobules on leading shoots one quarter the lobe area, in R. buccinifera complex the lobules are one eighth to one fi fth the lobe area and are rhomboid, not quadrate; 4) the production of caducous leaves, all members of the R. buccinifera complex bar R. anisotoma lack specialised means for asexual reproduction; 5) the regular pinnate, subdimorphic branching of female plants, in the R. buccinifera complex female plants are irregularly or pseudodichotomously branched, as a function of patterns of subfl oral innovation production (male plants may be pinnately branched).
Radula mittenii could be confused with a number of related species that occur in the wet tropics of north Queensland, however as the identity of several of these has not yet been fully resolved, the diff erentiation of R. mittenii from these species will be dealt with in a subsequent treatment of R. refl exa.
Stephani (1884) identifi ed a single specimen gathered by Pentzke in 1882 from the 'Zamtree River' held in the herbarium of Mitten as the type of R. mittenii , in so doing eff ectively identifi ed a holotype. Zamtree River is a lexigraphical error for the hand-written locality on the holotype, which is "Zaintree". Th is is a corruption of 'Daintree', the river immediately south of Cape Tribulation in north Queensland.
In this holotype (NY00831342) are three elements: two separate pieces of material and three shoots loose within the packet. Th ese three elements comprise three diff erent Radula species.
Th e fi rst element has shoot systems that are sparingly dichotomously branched, and lobules whose interior margin is not ampliate over the stem, and in these characters it does not agree with the protologue.
Th e second and third elements are pinnately branched, and lobules with infl ated base and truncate exterior margins, so agree in some pertinent details with the protologue. However, besides the description of shoot length, the protologue is insuffi ciently detailed to discriminate between elements two and three. Element two has shoots approaching 50 mm long which element 3 does not.
Of the three elements only two, the fi rst and second, are known to occur in the vicinity of the Daintree River. Th e third is a southern temperate species whose northern limit is somewhere in the vicinity of the Clarence River, northern New South Wales.
As there is not much basis, beyond shoot length, for selecting between elements 2 and 3 given the protologue, Article 9, Recommendation 9A.4 of the ICN could be invoked, to preserve current usage of the name R. mittenii , as a synonym of R. buccinifera following Yamada (1984), as the third element is this species. However, R. buc- cinifera does not occur at the Daintree River, from where the holotype was collected, and the holotype contains other elements that do occur there, one of which agrees with the protologue, making this undesirable. Lectotypifi cation of the name R. mittenii on the second element capitalizes on an available name for a species that has not otherwise been described. Historical misinterpretation of R. mittenii, particularly by Castle (1967), is immaterial as Yamada's proposed synonymy with R. buccinifera resolved this at the specimen level. Th is means the name R. mittenii has no current usage to preserve. We therefore lectotypify R. mittenii on that element agreeing with the diagnosis and originating from the type locality, being element two within NY00831342, which is separated as a subpacket.
Of the other elements, the fi rst corresponds to R. notabilis M.A.M.Renner, described below and the third, comprising three loose shoots within the packet corresponds to R. buccinifera .
Th e origin of these three loose R. buccinifera shoots is unclear -based on current knowledge of R. buccinifera 's distribution it cannot have come from the Daintree River region. Th e most likely explanation is that the type in Mitten's herbarium comprises an aggregate of material from two geographical locations. It is not known how this mixing may have occurred, however there is in MEL a collection, also attributed to Pentzke in 1882 from the Daintree River, and this is a large mat of pure R. buccinifera . We can rule no explanation for the attribution of R. buccinifera to the tropics out at this stage. Mitten's specimen was communicated from the Phytologic Museum of Melbourne by F. von Mueller, and it is possible the contamination, and confusion, occurred at the time duplicates were made. A duplicate of Pentzke's collection ex herb. Mitten in herb. Stephani (G00067469!) also contains a shoot of R. buccinifera , including branches, as does another duplicate ex herb Steph in BM (BM000969294!), these specimens can therefore be regarded as duplicates of the type specimen but are not isolectotypes because they do not contain the element upon which the name is lectotypifi ed.
[From NSW909500, NSW909501 and BRI-AQ722865] Forming loosely interwoven mats on tree trunks, branches and twigs, either tightly adherent on or hanging from substrate. Live plants nitid brown-green, fading to brown in herbarium. Shoot systems irregularly and often infrequently branched, female plants predominantly pseudodichotomous due to production of pairs of subfl oral innovations below gynoecia. Shoot systems monomorphic, 1.4-2.0 mm wide and up to 40 mm long, branches initially smaller in stature than parent shoot but attaining similar stature by fourth of fi fth pair of leaves. Older shoot sectors retaining leaflobes, though older leaf lobes may partially fragment on some shoots. Stems 135-160 μm diameter, with cortical cells in a single tier of 14-26 rows; outer half brownpigmented, inner half tan-pigmented; external free cortical cell wall heavily and continuously thickened, radial longitudinal cortical walls thin or slightly thickened, inner tangential walls heavily and more or less continuously thickened by fusion coarse nodular trigones; medullar cells in 15-32 rows, with coarse nodular trigones, lacking thickenings between trigones, occasionally with heavily and continuous thickenings, all unpigmented or faintly yellow pigmented. Cortical cells on dorsal stem surface arranged in straight longitudinal rows on young and mature shoot sectors. Leaf insertion not reaching dorsal stem mid-line, leaving one or two dorsal cortical cell rows leaf-free; leaf insertion not attaining the ventral stem mid-line, leaving two ventral cortical cell rows leaf-free. Leaf lobes rotund-elliptic to elliptic-oblong, 950-1145 μm long by 610-820 μm wide, contiguous to imbricate, not falcate, acroscopic base not sharply defl exed away from stem, plane, not interlocking over the dorsal stem surface, stem visible between leaf lobes in dorsal view; lobe margins irregular and crenulate, the interior lobe margin sometimes minutely auriculate, not or only just reaching the opposite stem margin, antical margin shallowly curved, becoming substraight in larger leaf lobes, sharply curved through nearly 90° in exterior quarter, exterior margin shallowly curved or straight, sharply curved through nearly 90° in postical quarter, postical margin straight; angle between postical lobe margin and keel c. 135°. Lobules rhombiform when small, transitional as stature increases to trapeziform with exterior and interior margins nearly parallel, one sixth to one fi fth the lobe area, 370-735 μm long by 260-480 μm wide, keel straight in rhombiform lobules, arched in trapeziform lobules, angle between keel and stem 135°, keel gradually turning through 90°, keel apex and postical lobe margin fl ush, interior lobule margin free for one quarter to one third its length, free portion weakly ampliate on rhomboid lobules to moderately ampliate on trapeziform lobules, extending at most half way across the ventral stem surface, acroscopic margin S-shaped to curved, apical portion inclined toward stem in smaller lobules, transitional to perpendicular to stem axis in larger lobules, apex acute in rhomboid lobules transitional to obtuse in trapeziform lobules, free exterior margin straight to shallowly curved, occasionally with a small knee above the lobe-lobule junction, margins plane, crenulate; lobe-lobule junction antical to the acroscopic end of stem insertion, lobule attached to stem along 0.66-0.75 of the interior margin, stem insertion more or less linear, gently curved at acroscopic and basiscopic ends, not revolute; a single papilla present at the lobule apex and another two papilla situated on the interior lobule margin above the stem insertion. Leaf lobe cells rounded-oblong, not arranged in rows, unequally sized, 16-26 μm long by 11-19 μm wide, thin walled with triangular trigones, medial wall thickenings absent; cells of lobe margin smaller than those of leaf middle, quadrate to rectangular, 9-15 μm long and wide, interior cell walls evenly and continuously thickened, exterior cell wall thickened diff erentially at midwall, causing exterior margin to be crenulated, cell lumen not bulging medially; leaf lobe cell surface unornamented, smooth. Oil-bodies not known. Asexual reproduction possibly by caducous leaf lobes but sporadic, older shoot sectors usually retaining most or all of their leaf-lobes, with fragmenting leaf-lobes tearing into several pieces, fragmentation scars jagged, irregular, typically leaving part of basiscopic leaf margin attached beyond keel, shoot primordia forming as irregular buds on leaf lobe after leaf fragmentation. Dioicous. Androecia on indeterminate branches that continue vegetative, androecial bracts in 4-8 pairs, lobules epistatic, keel deeply curved, bucket-like, free apical portion triangular, apex obtuse, moderately defl exed, lobes rounded, not caducous, antheridia not seen. Gynoecia terminal on leading shoots and branches, subtended by one or two full sized subfl oral innovations that are again fertile, where a single subfl oral innovation is present, a 'resting' shoot primordium occurs in place of the second subfl oral innovation; archegonia 140-170 μm tall, up to 16-18 per gynoecium on a small raised disc of tissue, encompassed by the protoperianth, archegonia neck eight cell columns; female bracts in one pair, asymmetrical, tightly imbricate, oblongobovate, larger lobe 665-720 μm long by 440-475 μm wide, smaller lobe 620-650 μm long by 350-370 μm wide, lobules rectangular, one half the lobe area, apex obtuse to broadly acute, keel arched, margins crenulate, insertion interlocking dorsally but not ventrally, insertion equitant. Perianths 2800-3800 μm long, conical and fl ared at mouth, mouth irregularly repand 880-950 μm wide. Perianth walls unistratose above, with bistratose bands extending up to half way up perianth, increasing in width toward base, becoming confl uent, basal perianth walls progressively increasing in thickness, 2-3 stratose. Long stem perigynium present, 5-6 stratose, external cell wall thickened and brown-pigmented, internal walls unthickened and unpigmented. Calyptral perigynium present, base of calyptra bistratose at base, unistratose above, unfertilised archegonia elevated on surface of calyptra.
Etymology. From Latin notabilis : notable, for trumpet shaped perianths with a fl ared repand mouth that are distinctive among both the Radula buccinifera species complex, and Australasian Radula .
Distribution and ecology. Radula notabilis is endemic to the Wet Tropics of north-eastern Queensland, where it is a common epiphyte in riparian rainforest in the tropical lowlands, from the edge of the coastal plain to approximately 300 m asl. Radula notabilis is rarely found far from running freshwater, and is usually encountered on tree trunks and branches over or adjacent to watercourses where it often forms closely adherent mats or pendant-procumbent wefts on bark, and does not often inhabit dense multi-species epiphytic turfs. Variation. Beyond variation in shoot stature and associated size related changes in lobule shape, Radula notabilis is morphologically consistent across individuals.
Recognition. Radula notabilis can be recognized by the combination of its infrequently branched monomorphic shoot systems, its slightly nitid appearance when fresh, the stem being visible between the leaves in dorsal view, the leaves being held in plane with the stem rather than being obliquely patent or dorsally assurgent, the presence of a dorsal leaf-free strip one or two cells wide; the trapeziform lobules with an obtuse apex; the perianths having a long tubular stem perigynium and walls that fl are abruptly to the mouth, whose labia are undulate-repand and often partially inrolled.
Radula buccinifera is the only named species to have been confused with Radula notabilis . Th e two species are similar in size and colour, the presence of a dorsal leaffree strip, gross lobule and perianth morphology. However, several character diff erences are accessible via critical examination. Lobules are quadrate to rhombic when small and large with typically S-shaped antical margin and obtuse to acute apex with and keel curved to straight or arched in R. notabilis (Fig. 22H) vs rhombiform when small, transitional to trapeziform when large, with a straight antical margin and obtuse apex and straight to arched keel in R. buccinifera (Fig. 11B). Th e stem anatomy of R. notabilis has coarse nodular trigones on cortical and medulla cell walls vs small triangular trigones in cortical and medulla walls in R. buccinifera . Th e perianth mouth is repand, inrolled and lobed in R. notabilis. vs plane and entire in R. buccinifera . In the fi eld colour diff erences are sometimes apparent, with R. notabilis being nitid browngreen whereas R. buccinifera may be milky yellow-green through dull brown-green to mid-green, but colour is not always a reliable indicator of identity. Geography also provides a good clue to identity in that Radula buccinifera does not occur in the Wet Tropics Bioregion of north-east Queensland.
Remarks. Radula notabilis is one of three elements present in the type specimen of Radula mittenii in herb. Mitten.
Specimens examined. Australia  Diagnosis. Radula pugioniformis is outwardly similar to R. buccinifera , but can distinguished by the presence of three female bracts in association with the gynoecium, the trullate to pugioniform lobules, the stem anatomy, where the cortical cell walls are heavily and continuously thickened which partially constricts the cell lumen, and the medulla cell walls are also continuously and heavily thickened, somewhat more so at cell junctions.
Type. Australia Description. [From NSW770504] Forming loosely interwoven mats of adherent shoots on soil and rock; live plants unknown, brown in herbarium; shoot systems monomorphic, 1.0-1.5 mm wide and up to 40 mm long, irregularly branched, though female plants predominantly pseudodichotomous due to production of pairs of subfl oral innovations below gynoecia, branches initially smaller in stature than parent shoot but attaining similar stature to parent shoot by second or third pair of leaves; older shoot sectors retaining leaf-lobes. Stems 110-140 μm diameter, with cortical cells in a single tier of 19-25 rows, cortical cell walls yellow-brown pigmented, all walls heavily and continuously thickened, partially constricting individual cell lumen; medullar cells in 12-17 rows, medulla cell walls yellow-pigmented, continuously and heavily thickened, somewhat more so at cell junctions; cortical cells on dorsal stem surface arranged in straight longitudinal row on young and mature shoot sectors. Leaf insertion not reaching dorsal stem mid-line, leaving one to three dorsal cortical cell rows leaf-free; leaf insertion not attaining the ventral stem mid-line, leaving four or fi ve ventral cortical cell row leaf-free. Leaf lobes ovate, 620-860 μm long by 500-700 μm wide, imbricate, weakly falcate or not, acroscopic base plane, not sharply defl exed away from stem, not interlocking over the dorsal stem surface, stem visible between leaf lobes in dorsal view, margins entire, the interior lobe margin weakly ampliate, not auriculate, not or only just reaching the opposite stem margin, more or less straight in larger lobes, sometimes with a single triangular tooth near the stem insertion, more often in smaller lobes, antical margin continuously curved, exterior margin curved, postical margin straight or slightly curved; angle between postical lobe margin and keel 100-135°. Lobules rhombic to trullate, one eighth to one seventh the lobe area, 240-485 μm long by 185-310 μm wide, keel slightly curved or straight, angle between keel and stem 100-135°, keel apex and postical lobe margin weakly notched, inner lobule margin free for one half to two thirds its length, free portion not ampliate, not extending across stem beyond insertion line, acroscopic margin straight to weakly curved, apex narrowly rounded to acute, free exterior margin straight, occasionally with a small knee above the lobe-lobule junction, margins plane, entire; lobe-lobule junction postical to the acroscopic end of stem insertion, attached to stem along 0.33-0.5 of the interior margin, stem insertion gently curved its entire length, not revolute; lobule apex bearing a single papilla, no other papilla, or papilla scars, observed on the interior lobule margin. Leaf lobe cells hexagonal-oblong, not arranged in rows, unequally sized, 11-21 μm long by 10-12 μm wide, walls moderately and continuously thickened. Cells of lobe margin smaller than those of leaf middle, quadrate to rectangular, 6-12 μm long by 6-10 μm wide, interior cell walls evenly and continuously thickened, exterior cell wall unthickened. Leaf lobe cell surface unornamented, smooth. Oil-bodies not known. Asexual reproduction absent. Dioicous. Androecia on short lateral branches or terminal on leading shoots, either terminating following androecia production, or continuing vegetative growth; antheridial bracts in 3-4 pairs; lobules epistatic, keel deeply curved, bucket-like, free apical portion triangular, apex acute, plane, lobes rounded, not caducous; antheridia not seen. Gynoecia terminal on leading shoots and branches, subtended by one or two full sized subfl oral innovations that are again fertile, more often subtended by a single subfl oral innovations on branches with a 'resting' shoot primordium in place of the second subfl oral innovation. Archegonia 175-190 μm tall, archegonia neck fi ve or six cell columns, 12-14 per gynoecium on a small raised disc of tissue, encompassed by the protoperianth, with several large single celled or stalked (on 2 or 3 cells) papillae scattered among gynoecia. Female bracts three, symmetrical, imbricate, obovate-falcate, lobe 1140-1255 μm long by 670-795 μm wide, lobules triangular, one half the lobe area, 585-930 μm long by 340-645 μm wide apex acute, keel arched, margins entire, insertion interlocking dorsally and ventrally, insertion equitant. Perianths and sporophytes not known.
Etymology. From Latin Pugioniformis, with the form of a dagger, in reference to the resemblance of the lobules to broadly triangular iron daggers adopted by several ancient cultures, including in the Roman Empire as the 'Pugio'.
Distribution and ecology. Radula pugioniformis is known from a range of localities in the Central and Southern Tablelands of New South Wales. Th e two specimens with details on substrate indicate R. pugioniformis grows on wet mud and rocks in and around streams or other watersources, in gullies or on steep slopes. Radula pugioniformis has been collected growing with Radula buccinifera , Acrophyllum dentatum , Th uidium furfurosum , and Bryum sp. on soil, and with Lejeunea sp. on rock.
Variation. Within individuals, shoot stature varies, and this is correlated with changes in lobule shape, which tend to be shorter on smaller shoots.
Recognition. Radula pugioniformis is outwardly similar to R. buccinifera , and inhabits a subset of microhabitats occupied by that species. Diff erentiating these two species is best achieved on the basis of hydrated, and preferably slide mounted material. Th e most accessible character by which Radula pugioniformis and Radula buccinifera diff er is in shape of the leaf lobules. In R. pugioniformis the lobules are conspicuously trullate, the antical margin is straight to curved, and slopes steeply toward the stem at 45-70°, such that the lobule apex lies well above the antical end of the stem insertion. In R. buccinifera the lobules are quadrate to rhombic, the lobule apex lies at variably between the same level as, or slightly above, the uppermost  point of the ampliate portion of the lobule margin. Between these two points the lobule margin varies from straight to S-shaped in situ (straight when fl attened), but when S-Shaped there is a pronounced medial curve. Th e slope of the antical margin varies between sloping downward toward the stem at up to 45° and remaining level.
Other diagnostic diff erences between R. pugioniformis and R. buccinifera can be found in the stem anatomy. In R. pugioniformis the cortical cell walls are heavily and continuously thickened, which partially constricts the cell lumen, and the medulla cell walls are continuously and heavily thickened, somewhat more so at cell junctions. In R. buccinifera the free external cortical cell wall is diff erentially thickened, but all other cortical and medulla cell walls bear triangular trigones at most, and are otherwise unthickened.
Radula pugioniformis is similar to R. iwatsukiana K.Yamada from New Caledonia , and may be related to this species. However, the type of R. iwatsukiana was not available for study due to CITES restrictions, and interpretation of morphology via descriptions and illustrations is fallible. Th e size of female bracts in both absolute terms and relative to vegetative leaves diff ers in R. pugioniformis from that described and illustrated by Yamada (1985), the illustrations also suggest a sharp distinction between marginal and medial cells in R. iwatsukiana that does not occur in R. pugioniformis. If Yamada's illustration is representative of R. iwatsukiana , there are also diff erences in the relative sizes of cortical and medulla cells in the stem; in R. iwatsukiana these are approximately the same area in transverse section, while in R. pugioniformis the cortical cells are half to one quarter the area of the medulla cells. Th ey are also far more numerous.
Remarks. Female bract number suggests this species belongs to subg. Odontoradula , not subg. Metaradula. Although apparently not closely related to species of the R. buccinifera complex, R. pugioniformis is included here because it has been misidentifi ed as R. buccinifera .
Specimens examined. Leaf lobes elliptic-ovate, 550-1050 μm long by 400-830 μm wide, remote to contiguous, not falcate, acroscopic base not sharply defl exed away from stem, fl at, lying in plane with stem, not interlocking over the dorsal stem surface, stem visible between leaf lobes in dorsal view; margins irregularly but minutely repand, minutely crenulate, the interior lobe margin not or at most shallowly ampliate, hardly covering the dorsal stem surface and not reaching the opposite stem margin, antical margin curved, exterior margin curved through nearly 100°, postical margin straight; angle between postical lobe margin and keel c. 135°. Lobules quadrate when small to oblong, one twelfth to one sixth the lobe area, 270-490 μm long by 140-270 μm wide; keel straight to shallowly arched, angle between keel and stem 135-150°, keel turning through 90° at keel-lobe junction, keel apex and postical lobe margin fl ush; interior lobule margin free for one fi fth to one quarter its length, free portion not ampliate in small stature lobules to moderately ampliate on large lobules, extending at most half way across the ventral stem surface; acroscopic margin S-shaped, apical portion perpendicular to stem; apex obtuse to apiculate; free exterior margin straight, margins plane, crenulated; lobe-lobule junction well postical to the acroscopic end of stem insertion; attached to stem along 0.75 to 0.8 of the interior margin, stem insertion more or less straight, not curved at acroscopic or basiscopic ends, not revolute; lobule apex bearing a single papilla, with another two papilla situated on the interior lobule margin above the stem insertion. Leaf lobe cells rounded-oblong, not arranged in rows, unequally sized, 10-25 μm long by 9-19 μm wide, thin walled with small triangular trigones, medial wall thickenings absent; cells of lobe margin smaller than those of leaf middle, quadrate to rectangular, 9-15 μm long and wide, interior and exterior cell walls not differential thickened, cell lumen bulging medially; leaf lobe cell surface unornamented, smooth. Oil-bodies two or three per cell, ellipsoidal, fi lling cell lumen, light-brown, surface granular, internally homogeneous or with a hyaline droplet. Asexual reproduction usually absent, however two specimens have been observed producing budlike shoot primordia from leaf lobe margins. Dioicous. Androecia on indeterminate branches that continue vegetative or reproductive growth, androecial bracts in 4-∞ pairs, lobules epistatic, keel deeply curved, bucket-like, free apical portion triangular, apex obtuse, moderately defl exed, lobes rounded, not caducous, antheridia not seen. Gynoecia terminal on leading shoots, subtended by two full subfl oral innovations that are usually full-sized and again fertile; archegonia 130-155 μm tall, archegonia neck fi ve or six cell columns, 6-8 per gynoecium on a small disc of tissue, not encompassed by the protoperianth; female bracts in one pair, slightly asymmetrical, tightly imbricate, elliptic-ovate, weakly falcate, lobe 840-1015 μm long by 460-545 μm wide, margins entire; lobules rhomboid to trullate, one quarter to one half the lobe area, apex acute to acuminate, keel straight to arched, margins entire; bract insertion lines interlocking dorsally and ventrally, insertion equitant. Perianths 3200-3800 μm long and 840-980 μm at mouth, mouth entire to irregularly lobed, perianth shape variable, either broadening from mouth to widest point at approximately one third to one half length above base, where 850-950 μm wide, then tapering to base, or tapering from mouth to base; perianth walls unistratose above, with bistratose collar 3 or 4 cell tiers high above the perianth-calyptra junction; long stem perigynium present, 5-6 stratose, cell walls not thickened or pigmented, perianth-calyptra fusion elevated above female bracts on 9-15 tiers of cells; calyptral perigynium present, base of calyptra bistratose, unistratose above, unfertilised archegonia elevated on surface of calyptra. Etymology. Strangulating, in reference to the entwining growth about a shoot of Weymouthia cochlearifolia exhibited by the type specimen.
Distribution and ecology. Radula strangulata is widely distributed throughout the New Zealand Botanical Region, from the Kermadec Islands in the north, throughout the North, South, and Stewart Islands, south to the Auckland and Campbell Islands, and east to the Chatham Islands. Radula strangulata occupies an elevational range from sea level to c. 800 m encompassing lowland to montane habitats, including coastal scrub, mature and regenerating lowland podocarp-broadleaf forest and beech forest on both sides of the main axial ranges in North and South Islands. Radula strangulata also has a broad ecological amplitude, occupying a range of microsites from tree trunks and bases, exposed tree roots on the forest fl oor, to rotting logs, exposed clay on forest banks, dripping rocks adjacent to waterfalls, and on rocks within stream beds, sometimes under running water. In hyperhumid locations R. strangulata may grow epiphyllously on fern fronds. Th is is the only Radula species in New Zealand commonly encountered growing aquatically, typically submerged on basaltic boulders within cool, clear, fast fl owing streams. It is also the only Radula species encountered in suburban areas where it is opportunistic in a range of man-made habitats in suitably moist sites, for instance the surfaces of rotting wooden roof tiles.
Radula strangulata is the most commonly collected species of Radula in New Zealand, and despite this accessibility bias, is probably the most common species of Radula in New Zealand.
Variation. Radula strangulata exhibits a broad amplitude of morphological variation, as might be expected from the diverse array of habitats occupied by this species. Th is, in conjunction with the relatively deep phylogenetic divergences between groups of individuals suggests R. strangulata , as circumscribed here, could well be a complex of weakly morphologically diff erentiated sister species. Variation in gametophyte morphology appears at least partly correlated with moisture regime of the occupied microsite. Plants growing in dry habitats and microsites have contiguous to imbricate leaf lobes, small rhomboid lobules, and are glaucous green. Th e origin of this glaucous colouration is not known, however it is not due to cell surface ornamentation. Th e type material of Radula silvosa is typical of plants growing in dry sites. Plants growing at the wet end of the moisture spectrum have remote leaf lobes, large rectangular lobules with a well developed ampliate interior margin when large, and are dark green. Th e type material of R. levieri is typical of plants from wet sites.
Part of the justifi cation presented by Hodgson (1944) for the recognition of R. silvosa was that the perianths abruptly infl ated above a tubular base, as opposed to gradually narrowed from mouth to base in R. levieri (a species she accepted). Th e former feature is apparent in juvenile perianths, but is lost as the stem perigynium subtending the perianthium grows and matures prior to eruption of the sporophyte from the calyptra. All perianths we have seen wherein the sporophyte has ruptured are gradually tapered from apex to base. In addition to developmental changes in perianth shape, perianth length at maturity varies, and this variation is positively correlated with capsule valve length ( R 2 =0.368, F (1, 22) =12.81, P =0.0017) (Fig. 33).
Recognition. Despite the variability exhibited by R. strangulata this species is relatively easy to recognize. Radula strangulata inhabits forest interiors, frequently in microsites on or close to the forest fl oor, and often in relative shade. Th e other species of the R. buccinifera complex with which R. strangulata could be confused on morphological grounds occupy quite diff erent microsites. Radula australiana inhabits subal- pine and alpine habitats typically dominated by shrubland, tussockland or grassland, and almost always above tree-line. Like R. strangulata , R. demissa is a forest inhabitant, but is typically an epiphyte on tree trunks, branches, twigs and occasionally lea ves, in reasonable light. When fresh, plants of R. strangulata growing in relatively dry microsites have a distinct glaucous bloom. Th is is a distinctive character that, in combination with microsite can facilitate identifi cation in the fi eld. However, not all plants are glaucous, and microsite diff erences from other members of the complex are not absolute.
Th e most accessible morphological character by which R. strangulata diff ers from R. demissa is the orientation of the leaf lobes. In R. strangulata the leaves lay in plane and alongside the stem, such that the dorsal stem surface is at least partially visible from above. In Radula demissa the leaves are obliquely-patent, and spread upward away from the stem and overlap one another across the dorsal stem surface, such that the dorsal stem surface is not usually visible from above.
If this character proves ambiguous, lobule shape is a source of diagnostic diff erences. Radula strangulata has lobules with a straight keel, and the carinal region is a weakly infl ated mound set back from the keel. In contrast, R. demissa typically has a curved keel, and the carinal region is strongly infl ated along the length of the keel. Radula strangulata expresses a range of lobule morphologies within a single population, from small rhomboidal lobules to large rectangular lobules. In rhomboidal lobules the apex lies close to the stem, and the exterior margin is noticeably inclined toward the stem and the exterior margin is straight. In R. demissa , lobule morphology is more conservative, with the smallest lobules being approximately quadrate such that the apex is closer to the keel apex, and the exterior margin is not noticeably inclined toward the stem. Th e exterior margin of R. demissa is curved. Another useful diff erence is found in the junction of the lobule and the lobe. In Radula strangulata this junction forms a simple angle of c. 135°, and the postical lobe margin then continues perpendicular to the stem in a straight line. In R. demissa this junction forms a notch, and the postical margin is weakly to strongly falcate, and curved. Diagnostic diff erences between R. strangulata and R. demissa can also be derived by counting the number of rows of dorsal cortical cells that are not crossed by the leaf insertion lines. In R. strangulata 2-5 rows are leaffree, whereas no rows are leaf free in R. demissa .
Radula strangulata has been confused with two other Radula , R. australiana and R. buccinifera . As noted above R. australiana is primarily an alpine species. Lobule shape provides the best morphological diff erences between R. strangulata and R. australiana. Whereas the lobules in R. strangulata are rhomboid to longitudinally rectangular, in R. australiana they are more or less quadrate. Lobules in R. australiana are typically larger in comparison to the lobe size than in R. strangulata , and may be up to one quarter the lobe area. Th e lobule keel and its junction with the leaf lobe also diff erentiate these two species. In Radula strangulata the keel is straight, and forms an angle of c. 135° at the junction with the lobe, the postical lobe margin then continues perpendicular to the stem in a straight line. In R. australiana the keel is curved, and continues evenly into the rounded outline of the leaf without forming a notch or angle at the junction. Diagnostic diff erences between R. strangulata and R. australiana can again be derived by counting the number of rows of dorsal cortical cells that are not crossed by the leaf insertion lines. In R. strangulata 2-5 rows are leaf-free, whereas no rows are leaf free in R. australiana.
As far as is known, R. buccinifera does not occur in New Zealand. However, it is always worth checking unusual plants against either of these species, particularly as the vagrant occurrence of R. buccinifera in New Zealand cannot be ruled out. Th e reverse is also true, and some aquatic forms of R. buccinifera are diffi cult to diff erentiate from R. strangulata , and both species occur in Tasmania and south-east Australia. For guidance on separating R. strangulata and R. buccinifera , see the recognition section of R. buccinifera .
Remarks. Th e variation in morphology exhibited by Radula strangulata seems to have encouraged the description of several species that we retain in synonymy. Th e type of R. strangulata consists of small male plants on Weymouthia cochlearifolia , the lobules on this individual represent only a part of the range of variation expressed by the species. Th e other end of lobule shape variation, having heavily ampliate lobule base extending over the ventral stem surface, is found in the type specimen of Radula levieri. Th is specimen also has remote leaves and in these two characteristics corresponds to morphotypes associated with microsites that are permanently saturated. Th e type of R. silvosa possesses lobules of the same shape range as exhibited by the type of R. strangulata, however the plants are larger and female. Hodgson's (1944) application of names is explained by her stated belief that the type of R. strangulata was the male plant of R. levieri . Th e type of Radula parviretis exhibits variable leaf size on single shoots, reddish pigments, variable leaf cell size, and grows on peat. Variation in cell size within leaves is not atypical of R. strangulata . Th e collections made on peat may refl ect occupancy of a rarely explored habitat. Radula parviretis is known from only two specimens, both from Port Pegasus on Rakiura (Stewart) Island. Reddish pigments are not known in other specimens of this species, and have not been observed in the fi eld.
Th e plants illustrated for Radula buccinifera and R. silvosa in Allison and Child (1975) are actually both Radula strangulata . Th e material illustrated for R. buccinifera corresponds to a hygromorphic phenotype whereas the material illustrated for R. silvosa corresponds with plants occupying drier microsites. Both drawings illustrate clearly how the dorsal stem surface is visible between the leaves, however as described above the perianth diff erences illustrated for these two species are misleading. Hodgson (1944) identifi ed a monoicous specimen of R. silvosa collected by N. Kemble Walsch. Th is specimen (CHR587344) contains a packet labeled 'perianth showing monoecious [sic]' containing three shoots two of which are male and one of which is female. If the female shoot was ever attached to either of the male shoots, it was not so attached at the time of this investigation. Although the apex of one of the male shoots is missing, the sequence of leaves suggests that the female shoot does not comprise that apex, as the male shoot ends with an entire right leaf, and the female shoot begins with a fragment of the same. Furthermore, the remainder of the specimen consists of shoots that are either male, or female, or sterile. Th ere is no defi nite evidence that this specimen is not a mixture of a male and a female individuals, a common occurrence for this, and other dioicous Radula species.  Remarks. Radula farmeri was recorded for Australia by So (2006). Th e specimens CANB650456 and CANB650457 are sterile, and while specimen CANB650459 is fertile the perianth mouth is entire. As identifi cation of R. farmeri is dependent on observation of perianths with fi mbriate mouths; her determinations are not substantiated by the specimens, and this species should be excluded from the Australian fl ora.