The concluding chapter: recircumscription of Goodenia (Goodeniaceae) to include four allied genera with an updated infrageneric classification

Abstract Close scrutiny of Goodenia (Goodeniaceae) and allied genera in the ‘Core Goodeniaceae’ over recent years has clarified our understanding of this captivating group. While expanded sampling, sequencing of multiple regions, and a genome skimming reinforced backbone clearly supported Goodenias.l. as monophyletic and distinct from Scaevola and Coopernookia, there appears to be no synapomorphic characters that uniquely characterise this morphologically diverse clade. Within Goodenias.l., there is strong support from nuclear, chloroplast and mitochondrial data for three major clades (Goodenia Clades A, B and C) and various subclades, which lead to earlier suggestions for the possible recognition of these as distinct genera. Through ongoing work, it has become evident that this is impractical, as conflict remains within the most recently diverged Clade C, likely due to recent radiation and incomplete lineage sorting. In light of this, it is proposed that a combination of morphological characters is used to circumscribe an expanded Goodenia that now includes Velleia, Verreauxia, Selliera and Pentaptilon, and an updated infrageneric classification is proposed to accommodate monophyletic subclades. A total of twenty-five new combinations, three reinstatements, and seven new names are published herein including Goodenia subg. Monochila sect. Monochila subsect. Infracta K.A.Sheph. subsect. nov. Also, a type is designated for Goodenia subg. Porphyranthus sect. Ebracteolatae (K.Krause) K.A.Sheph. comb. et stat. nov., and lectotypes or secondstep lectotypes are designated for a further three names.


Introduction
Representatives of the predominantly Australian family Goodeniaceae R.Br, a close relative to the cosmopolitan Asteraceae Bercht. & J.Presl (Tank and Donoghue 2010), have been the focus of various studies in recent years. The first molecular phylogeny of generic exemplars by Gustafsson et al. (1996) indicated that the monotypic and closely allied Brunoniaceae Dumort. was, in fact, embedded within Goodeniaceae. This was previously hypothesised by Carolin (1992a), due to the shared presence of a unique cup-like pollen presenter positioned at the apex of the style referred to as an indusium . Howarth et al. (2003) (and later expanded in Jabaily et al. 2014), studied the origin of Hawaiian species of Scaevola L., the only genus in the family with significant diversity outside of Australia (see table 1 in Jabaily et al. 2012). These studies confirmed that Scaevola dispersed from Australia into the islands of the Pacific at least four times, starting in the late Miocene and continuing into the Pliocene, and that homoploid hybridisation subsequently contributed to the extant diversity apparent across the islands today (Howarth and Baum 2005).
Clarifying the relationships among Goodenia clades A, B, C and the smaller affiliate genera was necessary in order to identify monophyletic groups for taxonomic recognition. We could not seek to make changes to Goodenia s.l. without, at minimum, full and consistent resolution of the backbone relationships and confidence in the species-level composition of each major clade. To try and address this issue, the power of next-generation sequencing was leveraged for a subset of taxa (Gardner et al. 2016a). Twenty-four taxa representative of almost all major clades within Core Goodeniaceae, including 19 accessions from Goodenia s.l. (except subsect. Scaevolina and a subset of species from subsect. Goodenia placed in Clade C), were sequenced using genomeskimming technology. The majority of coding regions of the chloroplast genome were assembled and analysed, resulting in a nearly fully resolved phylogeny for all but two nodes within Goodenia s.l. This topology was then applied as a constraint and also concatenated on an expanded matrix of 98 Core Goodeniaceae species with trnL-F and matK loci of sequence data, greatly improving phylogenetic support values. This backbone topology has been similarly utilised in the present study. The analyses of Gardner et al. (2016a) confirmed the position of Coopernookia as sister to the remainder of Goodenia s.l., followed by stepwise sisters Clade A, Clade B, and finally Velleia sister to Clade C. However, the composition and relationships of subclades within the most morphologically diverse Clade C were poorly resolved, except for the monophyly of subg. Monochila and subg. Goodenia subsect. Coeruleae. Exploration of the backbone phylogeny derived from additional genomic compartments (nuclear ribosomal complex, several single copy nuclear genes) in the study suggested alternative topologies compared to the plastid, though with low support. To continue our investigation of alternative backbone topologies and delve into the poorly resolved Clade C, we expanded and further explored the next-generation sequencing data across nuclear, chloroplast and mitochondrial genomic compartments (Jabaily et al. 2018). We generated new genome skimming data for four additional taxa from within Clade C, and re-analysed the previously generated raw genomic data for the taxa in clade Goodenia s.l., for a total of 24 taxa. Partial mitochondrial genomes and partial chloroplast genomes expanded beyond the efforts of Gardner et al. (2016a) similarly strongly supported our original backbone relationship within Goodenia s.l. Extensive hypothesis tests were performed to explore congruencies and determine statistical support for all possible relationships within the challenging Clade C. Still, relationships between taxa and subclades within Clade C remained poorly resolved with both mitochondrial and plastid loci, as well as an expanded nuclear data set. In conclusion, while there was strong support for the monophyly of subg. Monochila and all other subclades represented distinct lineages, their position relative to each other remained unresolved, thus precluding their recognition as well-supported genera.

Morphology of Core Goodeniaceae
The Flora of Australia treatment of Goodeniaceae and Brunoniaceae (Carolin 1992a;Carolin et al. 1992) represented more than 30 years of research by Roger Carolin and his students. Over this period, Carolin's team successively revised each genus through a series of detailed anatomical and morphological studies that culminated in the recognition of numerous new species and updated infrageneric classifications. Through his early cladistics work and study of inflorescence types, Carolin determined that there were two distinct assemblages within the Goodeniaceae (Carolin 1967a;Carolin 1977). The first was the Lechenaultia-Anthotium-Dampiera (LAD) group, united by the presence of a cymo-paniculate inflorescence, connate anthers and a base chromosome number × = 9. In contrast, the remaining genera within the 'Goodenia group' had a fundamentally different vascularisation of the ovary, a thyrse or raceme-like inflorescence (Figs 2G,3,4), free anthers, and a base chromosome number of × = 7 or 8. These broad relationships were borne out in subsequent molecular studies (Jabaily et al. 2012;Gardner et al. 2016a) (Fig. 1). Carolin (1992a) rightly pointed out that Brunonia was clearly allied to the LAD group and perhaps should not be supported as a distinct family; however, Brunoniaceae was ultimately retained for his Flora of Australia treatment due to the adoption of the Cronquist (1981) classification system by the Flora at its inception (Kanis 1981). Within his 'Goodenia group', Carolin (1977) also determined that the monotypic south-west Western Australian genus Diaspasis was allied to Scaevola despite the presence of connate anthers and almost radially symmetrical flowers compared to the free anthers and fan-shaped flowers typical of the widely distributed Scaevola (Carolin 1992c;1992d); a relationship subsequently confirmed through phylogenetic analyses (Jabaily et al. 2012). Finally, Carolin (1990;1992e) concluded that Coopernookia, Velleia, Verreauxia, and Pentaptilon were allied to Goodenia, along with the four genera Calogyne R.Br., Symphyobasis K.Krause, Neogoodenia C.A. Gardner & A.S.George, and Catosperma Benth. that were later subsumed into an expanded Goodenia. Coopernookia is the only genus in the family that shows the classic SW-SE Australian disjunction from the Nullarbor uplift around 8.8 to 0.5 million years ago (Jabaily et al. 2014), with three species endemic to the central and south west of the continent and three confined to eastern Australia (Carolin 1992b). Carolin (1967b) recognised this genus as distinct from the rest of his 'Goodenia group' as all species have a base chromosome number of × = 7 (rather than 8), stellate hairs on the stems and leaves, and ovoid, strophiolate seeds. The seed testa is also unique, with thickened, straightsided cell walls that contain no mucilage in contrast to many species of Goodenia that  Carolin (1967a) with his corresponding Bauplan 'Type' concepts stated were applicable and phylogenetic position for exemplar species given in brackets; MI = main inflorescence, EZ = enrichment zone, V = vegetative zone. A Form A (Type 1) is a thyrse with leafy bracts and bracteoles e.g. Goodenia ovata (Goodenia I) or (Type 2) 1(-2)-flowered raceme with leafy bracts and bracteoles e.g. G. laevis (Goodenia I) (inset) B Form B (Type 5) is a basal rosette with leafy bracts and bracteoles e.g. G. hederacea (Goodenia II) C Form C (no Type) with flowers solitary in leaf axils, leafy bracts and bracteoles e.g. G. convexa (Goodenia II) D Form D (no Type) flowers solitary in leaf axils with leafy bracts, bracteoles absent e.g. G. pumilo (Porphyranthus I) E Form E (Type 4) a basal rosette, bracteoles, with leafy bracteose bracts and either a panicle-like form e.g. G. paniculata, raceme e.g. G. gracilis (Porphyranthus II) (inset above) or a thyrse e.g. G. pterigosperma (Coeruleae) (inset below) F Form F (Type 6) with ebracteolate racemes and leafy bracts e.g. G. hispida (Ebracteolatae II), (Type 7) non-leafy bracts e.g. G. cusackiana (Ebracteolatae I) (inset above), or (Type 8) a subumbel e.g. G . pulchella (Ebracteolatae I) (inset below) G Form G (Type 3) represented by a thyrse with reduced bracts and bracteoles e.g. G. scapigera (Monochila) and H Form H (Velleia Type) is a compound dichasium with leafy bracts and bracteoles e.g Velleia lyrata (Velleia).
have somewhat compressed seeds, where the epidermal cells are thickened towards the centre and thinner towards the margins and may contain mucilage that swells when wet (Carolin 1966). The flowers of Coopernookia also have retrorse barbulae inside the corolla that act as pollinator guides, reminiscent to those present in Scaevola. Indeed, Carolin (1967b) insightfully postulated that Coopernookia would have an intermediate position between Scaevola and Goodenia, which was later supported by molecular data as it was shown to be sister to Goodenia s.l. (Gardner et al. 2016a).
Goodenia is the largest and most floristically diverse genus in Goodeniaceae with c. 220 species. Species are largely confined to Australia apart from a few representatives that extend northwards to New Guinea, Indonesia, Malaysia, Philippines, and China, with a single taxon also endemic to the Island of Java (Leenhouts 1957;Carolin 1992e;Hong and Howarth 2011). Goodenia are annual or perennial herbs or low shrubs that occupy a wide variety of habitats in almost every biome across the Australian continent. Many species have yellow, white or blue bilabiate flowers (Figs 2,(5)(6)(7)(8), although there have been multiple independent floral symmetry shifts to a fan-shaped flower form (Gardner et al. 2016b). Fruit structure, seed coat surface patterns and appendages such as wings (Fig. 2F) are also important diagnostic characters for the genus (Carolin 1980). Recently, it was determined that the genus Goodenia had been lectotypified incorrectly as the first named species, G. ramosissima Sm., is in fact a species of Scaevola (≡ S. ramosissima (Sm.) K.Krause). Consequently, a proposal was put forward to conserve the name Goodenia using the conserved type G. ovata Sm. (Shepherd et al. 2017) (Fig. 2A), which was subsequently accepted (Applequist 2019). Carolin's (1992e) infrageneric classification currently recognises two subgenera and various sections, subsections and series (Table 1).
Selliera, a genus of three fan-flowered species from Australia, New Zealand and Chile, was supported as distinct within Carolin's (1977) 'Goodenia group'. However, he later questioned its status (Carolin 1992f ), noting that these species resembled members of Goodenia sect. Goodenia with fruits that show "a striking resemblance to that of G. koningsbergeri (Backer) Backer ex Bold. although somewhat smaller" (Carolin 1966), and he suggested that future work may determine that Selliera should be  Carolin (1959). From left-right Velleia, Goodenia, Verreauxia, and Scaevola, showing fusion of floral parts and structure of the incomplete locules and placentation of ovules.

Figure 5.
Goodenia Clade A phylogeny from combined nrDNA + cpDNA sequence data and exemplar taxa of major subclades. Topology is 50% majority rule cladogram from the partitioned Bayesian inference analysis. Support values above the branches are Bayesian posterior probabilities and below are maximum likelihood bootstrap values. Branch colour corresponds with support values and taxon colour corresponds to the taxonomic classification of Carolin et al. (1992). For updated taxonomy from this paper, see Tables  1, 2 Carolin & al. 1992 Figure 7. Goodenia Clade B_2 phylogeny from combined nrDNA + cpDNA sequence data and exemplar taxa of major subclades. Topology is 50% majority rule cladogram from the partitioned Bayesian inference analysis. Support values above the branches are Bayesian posterior probabilities and below are maximum likelihood bootstrap values. Branch colour corresponds with support values and taxon colour corresponds to the taxonomic classification of Carolin et al. (1992). For updated taxonomy from this paper, see Tables Carolin & al. 1992 subg. Monochila F synonymised under Goodenia. This appears to be supported as a single representative of the genus, Selliera radicans Cav. (Fig. 2I), was shown to group with the unusual Scaevola collaris (Carolin 1992c Verreauxia, a small genus of three species from southwest Western Australia, is distinguished by unusual multi-cellular branched hairs and a unilocular ovary ( Fig. 4 and Fig. 8E) that develops into an indehiscent, nut-like fruit with a single seed that (unusually) does not contain any mucilaginous cells in the seed coat testa (Carolin 1966. Carolin (1977) also included this genus in his 'Goodenia group' along with the closely allied monotypic Pentaptilon, which has similar branched hairs but is distinguished by its uniquely winged ovary and fruit (Carolin 1992h). Pentaptilon together with Verreauxia formed a monophyletic group in molecular analyses within the morphologically variable Goodenia Clade C (Gardner et al. 2016a;Jabaily et al. 2018).  Carolin (1992e) to an expanded Goodenia circumscribed herein that includes the genera Selliera, Verreauxia, Pentaptilon and Velleia, with the updated infrageneric classification and new authorities also provided. Velleia, the final genus in Carolin's (1977) 'Goodenia group', currently includes 21 species endemic to Australia except V. spathulata R.Br., which is also found in Malaysia, western New Guinea and the Louisiade Archipelago (Leenhouts 1957;Carolin 1992g). Many species within this genus have a distinctive inflorescence structure comprised of expanded axillary dichasia (Fig. 3H). Another important diagnostic feature for Velleia is the presence of a predominantly superior ovary with the sepals, corolla and stamens usually adnate to the base (Carolin 1992g) (Fig. 2E). In contrast, the remaining Core Goodeniaceae generally have inferior ovaries, except G. macroplectra (F.Muell.) Carolin; a species in Goodenia subsect. Ebracteolatae (Jabaily et al. 2012) that has free sepals inferior to the ovary while the corolla is fused to the apex (Carolin 2007) (Fig. 7E). However, anatomical examination of Velleia ovary sections by Carolin (1959) revealed the floral parts are in fact fused to the ovary to a degree and the stamens are never fully hypogynous and in many species they appear epigynous (Jeanes 1999) (Fig. 4). Carolin (1977) stated that the flowers, fruits and seeds of Velleia are similar to those of various species of Goodenia and suggested that the morphology of the ovary was not a "reversion to an almost superior ovary but the vestiges of the former inferior condition are retained". The infrageneric classification of Velleia, as currently recognised in the Flora of Australia, includes three sections (Carolin 1992g), based on the presence of three sepals (sect. Velleia) or five, which are either connate into a tube (sect. Euthales (R.Br.) Carolin) or free (sect. Menoceras R.Br.). While Velleia was supported as monophyletic in molecular analyses, it is placed sister to the remaining species in Goodenia Clade C (Gardner et al. 2016a;Jabaily et al. 2018).

Inflorescence structure in Goodenia and allied genera
Genera within Goodenia s.l. display a wide variation in floral form. Carolin (1967a) suggested the inflorescence structure was based on an open, polytelic, thyrsoid form with bracts that may be leaf-like ( Fig. 2H) or reduced. The component axillary cymules of the 'primitive' thyrse may become reduced to form racemes and spikes (Fig. 3A), or the main axis may contract to form subumbels (Carolin et al. 1992) (Fig. 2C). Carolin (1967a) outlined nine reference 'types' that summarise the variation in the "Bauplan" across Goodenia s.l. While some inflorescence forms correspond to various infrageneric groups, Carolin's (1967a) survey of inflorescence structure was not comprehensive enough to extrapolate further. Therefore, a more complete survey was undertaken across the Core Goodeniaceae to determine if patterns in floral form are diagnostic and correspond to monophyletic groups recovered in our molecular analyses.

Aims
Roger Carolin's lifetime of work provides a sound framework to test hypotheses about evolutionary relationships in Goodeniaceae and allow for a re-examination of his ge-neric concepts and infrageneric groups. The aim of this study is to build on our previous research (Jabaily et al. 2012;Gardner et al. 2016a;Jabaily et al. 2018) to produce well-sampled and well-resolved phylogenies combining both nrDNA (ITS) and cp-DNA (trnL-F, matK) molecular markers. These updated phylogenies, in conjunction with a survey of inflorescence structure, will clarify our understanding of the systematic importance of these features to characterise subclades within Goodenia s.l. The time has now come to also update Carolin's Flora of Australia classification to reflect these findings and to formally name and describe monophyletic clades as infrageneric groups in Goodenia s.l. in order to ensure nomenclatural stability going forward.

Taxon sampling
Our study includes over 95% of described species within Goodenia s.l. (Suppl. material 1). This paper includes sequences of trnL-F and matK from Jabaily et al. (2012), trnL-F, matK, and nrITS samples from Gardner et al. (2016b) with the majority of accessions new to this study. In some instances, multiple accessions of a taxon were included, including some subspecific taxa, to test for monophyly. A number of informal taxa or phrase-named taxa have also been included such as Goodenia sp. Dampier Peninsula (B.J. Carter 675) (Western Australian Herbarium 1998-; Council of Heads of Australasian Herbaria 2006-) to test the genetic uniqueness of these taxa and confirm allied species. Initially, a dataset with all taxa was aligned and analysed, but the backbone relationships between Goodenia Clade A, Clade B, and Clade C were unresolved, as expected from our prior work with these loci. Datasets were then created and analysed separately for all taxon sets (A, B, C), following following Jabaily et al. (2012) and Gardner et al. (2016a). Separating the total dataset by clade allowed for more precise alignment across taxa, particularly within trnL-F. As the genus Coopernookia was confirmed as sister to Goodenia s.l. (Gardner et al. 2016a), Coopernookia polygalacea (de Vriese) Carolin was used as the outgroup for all three taxon sets; within Clade B and Clade C, two accessions from Clade A (Goodenia benthamiana Carolin and G. tripartita Carolin) were included as additional outgroups.

Sequencing and Phylogenetic inference
Molecular sequencing primers and protocols follows Gardner et al. (2016a). Sequencing was completed by Macrogen (Seoul, South Korea). Individual loci were aligned in Geneious v. 11.0.2 (Kearse et al. 2012) using the Geneious tree building algorithm, with subsequent manual correction. For all taxon sets (A, B, C), three separate alignments were made for the chloroplast loci (matK and trnL-F), nuclear ribosomal locus (nrITS), and all data combined (matK, trnL-F, and nrITS). For each taxon set, individual loci were analysed for models of molecular evolution with Akaike information criterion (AIC) implemented in jModelTest2 (Guindon and Gascuel 2003;Darriba et al. 2012), implemented in CIPRES Science Gateway (Phylo.org; Miller et al. 2010). For the nrITS dataset of Clade B, the model selected was SYM + G + I. For the nrITS and cpDNA datasets of Clade A, the model selected was GTR + G. For the nrITS and cpDNA datasets of Clade C and the cpDNA dataset of Clade B, the model selected was GTR + G +I.
Bayesian phylogenetic analyses using MrBayes 3.2.2 (Ronquist et al. 2012) were conducted in CIPRES Science Gateway. For individual datasets each locus varied independently under the parameters specified by the individual model of molecular evolution. For each Bayesian analysis, two runs were conducted, each with three heated and one cold chain and uniform priors. The heated chain temperature was adjusted to ensure adequate mixing. Each analysis was set to run for up to 100 million generations, autoclosing when the standard deviation of split frequencies reached 0.01. Trees were sampled every 10000 generations, and 25% was discarded as burn-in. The adequacy of each analysis was completed by ensuring effective sample size >100, potential scale reduction factor of ~1.0 for all parameters, and acceptance rates of swaps between adjacent changes was between 0.1-0.7 in Tracer 1.6 (Rambaut et al. 2014). Majority rule consensus trees with posterior probabilities were generated in Geneious.
Maximum likelihood analyses using RAxML v. 8.0 (Stamatakis 2014) were conducted at the high-performance computing cluster (HiPerGator) at the University of Florida using the optimal models of molecular evolution for each dataset as discussed with 1000 bootstrap replications, summarized onto the best ML tree.

Taxonomy and morphology
Typification, synonymy and taxonomy largely follow the Flora of Australia treatment (Carolin et al. 1992) and/or the Australian Plant Name Index (https://biodiversity. org.au/nsl/services/APNI). Field work was conducted over several years in southern Western Australia facilitating the collection of fresh samples for DNA sequencing and examination of plants in situ. Types and specimens at various herbaria or on loan (AD, BRI, CANB, CGG, DNA, K, LD, MEL, PERTH, W) were also critically examined for the morphological survey of inflorescence structure and for lectotypifications. Further material was viewed using Global Plants (http://plants.jstor.org/) and the Museum National d'Histoire Naturelle online database (https://science.mnhn. fr/institution/mnhn/search) (indicated by "image!" in the citation). Images of seeds of various species were viewed on the Seeds of South Australia website (https:// spapps.environment.sa.gov.au/seedsofsa/). Non-Australian species of Selliera were assessed using online images available through the Flora of New Zealand (http:// www.nzflora.info/search.html).

Phylogenetic inference
The cpDNA and nrITS topologies were highly congruent for each taxon set representing Clades A, B and C and no taxon moved between named clades in the nrDNA, cpDNA and combined analyses. For each taxon set, both the chloroplast and nrITS trees (Suppl. materials 2-7) and original alignments (Suppl. materials 8-12) are available. Further, there were no substantial conflicting positions of strongly supported taxa, except where noted below.
Sixty-five accessions were included in the Clade A dataset, representing 50 named taxa (species, subspecies) and four unnamed taxa. Twenty-five of these were not included in previous studies, a 50% increase in taxon coverage. Clade A, representing the majority of species in subsect. Goodenia, resolves into two well-supported subclades (Goodenia I and II) with roughly similar numbers of taxa (Fig. 5). The backbone of subclade Goodenia I was poorly supported, with the position of G. phillipsiae Carolin (a species previously included in subsect. Ebracteolatae) differing between datasets. Similarly, accessions of Selliera placed in slightly different subclades. In the nrITS analysis, species of Selliera resolve as sister to G. viscida (previously included in subg. Monochila), several clades removed from Scaevola collaris and G. laevis Benth.; however, they are placed sister to these species in the cpDNA and combined analyses (Fig. 5). Goodenia II is congruent between datasets and resolves into two subclades. These were congruent between datasets except for a weakly supported subgroup comprising G. atriplexifolia A.E.Holland & T.P.Boyle, G. disperma F.Muell. and G. viridula Carolin that was recovered in the nrITS dataset but not retained in the cpDNA or combined data analyses.
The Clade B dataset comprised 175 accessions of 132 taxa (species, subspecies, and unnamed spp.) including 26 unnamed species. Seventy-seven taxa are newly included in this study, representing 58% of our sampling. Clade B comprises well-supported subclades (Porphyranthus I and II) of sect. Porphyranthus that are successively sister to subsect. Ebracteolatae, which resolves into strongly supported subclades Ebracteolatae I and II (Figs 6, 7). Taxon composition of these subclades and relative support values are congruent between datasets. In addition to including representatives of Carolin's sect. Porphyranthus, the Porphyranthus I clade also comprises two representatives G. kakadu Carolin and G. pumilio R.Br. from sect. Amphichila; a small section of diminutive species found in damp habitats in Northern Australia with the latter species also extending to New Guinea. Our analyses show that G. chthonocephala Carolin, a poorly known and unusual cushion-like plant previously included in ser. Borealis Carolin, is also allied to these two species (Fig. 6). Further, G. neogoodenia Carolin, an atypical species currently included in subsect. Ebracteolatae, is allied to another group of northern Australian species in the Porphyranthus I clade. The remaining representatives of Carolin's (1992e) ser. Borealis are included in the Ebracteolatae II clade, intermingled with species previously included in ser. Calogyne (Fig. 7), while G. wilunensis Carolin (subsect. Goodenia), is sister to the Ebracteolatae I clade (Fig. 6).
Clade C represents the most morphologically diverse group. Analyses included 92 accessions representing 67 taxa (with 4 being unnamed), a 31% increase in the number of species previously sampled across this clade. Seven individual subclades were well supported: a small group in sect. Goodenia, subsections Scaevolina and Coerulea, subg. Monochila and the genera Velleia, Verreauxia and Pentaptilon (Fig. 8). However, the relationships between clades remains unclear. In the combined analysis, subg. Monochila and subsect. Scaevolina were supported as sister, but this relationship was not found in the individual cpDNA and nrDNA trees. Similarly, G. xanthotricha de Vriese and G. arthrotricha Benth. were weakly supported as sister to subsect. Coeruleae in the combined analysis only. Surprisingly, G. quadrilocularis R.Br. is supported as sister to Velleia on the nrITS tree while it was sister to the subset of species from subsect. Goodenia in the cpDNA and combined analyses.

Inflorescence morphology
Carolin (1967a) originally classified the various inflorescence structures evident in Goodenia s.l. into nine different types, based on a relatively limited number of species. A survey of inflorescence morphology across Goodenia s.l. was undertaken here, utilising published information, images and herbarium specimens, to confirm key diagnostic characters such as the position and insertion of leaves and bracts, the presence or absence of bracteoles, and overall inflorescence form (Table 2). It should be noted that it is not always easy to distinguish leafy bracts from cauline leaves or between bracts and bracteoles in this group. For example, Albrecht (2002) observed that while Goodenia halophila Albr. and G. cylindrocarpa Albr. have structures subtending the flowers that look like bracteoles, axillary buds are sometimes present. For that reason, he decided to follow the classification of Briggs and Johnson (1979) and used the term "opposite or sub-opposite bracts" rather than bracteoles. While more accurate in some respects, this terminology is not entirely satisfactory and subsequent authors have continued to use the term bracts for reduced cauline leaves that subtend flowers, and appendages on the flower stalk, when present, are termed bracteoles (Holland and Boyle 2002;Cowie 2005;Pellow and Porter 2005;Sage and Shepherd 2007;Lang 2014). This survey also follows Carolin's concepts for floral structure; however, a more comprehensive evodevo study of floral development that considers the genetic mechanisms that control branching patterns of the floral-axis (i.e. inflorescences), would greatly improve our understanding of these complex structures. Carolin's (1967a) original classification of inflorescence structure is now revised to eight different morphologies, characterised as Forms A-H (Fig. 3). Carolin's Type 1 form, seen in the type species G. ovata (Fig. 3A), was characterised by the terminal shoot ending with a main inflorescence (MI), subtended by a zone of enrichment (EZ) and then the zone of inhibition or vegetative zone (V) (these labelled HF, BZ and V respectively in Carolin 1967a). The overall structure is a thyrse with leafy bracts, which Carolin observed in other species in subsect. Goodenia such as G. mueckeana F.Muell. Table 2. An updated linear sequence and classification for Goodenia s.l. including phylogenetic position (NS = species not sequenced) and a summary of morphological characters such as leaf position, inflorescence form and type (as characterised in Figure 2), the presence of leafy (L), bractose (N) or disc-like (D) bracts, and presence (1) or absence (0) of bracteoles. Authorities for most taxa are available in APNI (https://biodiversity.org.au/nsl/services/apni). *Species with an uncertain placement within Clade C.  1 in Carolin 1967a, note the infrageneric classification in that work follows Krause 1912). The subsequent year's growth in this inflorescence type is continued by lateral buds in the inhibition zone of the previous year's growth. In Carolin's Type 2 form, every partial inflorescence is reduced to 1(-2) flower(s) per raceme, as observed in species of Coopernookia and other members of subsect. Goodenia such as G. laevis (Fig. 3A inset) and G. calcarata (F.Muell.) F.Muell. Type 1 and 2 inflorescence forms intergrade somewhat, as some species such as G. varia R.Br. and G. grandiflora Sims were recorded by Carolin (1967a) as having both a Type 1 and Type 2 Bauplan. As such, these two inflorescence types have been combined into a new category designated as Form A. This form occurs in all species in the Goodenia I subclade of Clade A, and several species allied to G. atriplexfolia in the Goodenia II subclade (Table 2). Other species in this latter clade also exhibit inflorescence Form B previously known as Carolin's Type 5, which have a basal rosette but with leafy bracteoles, as observed in G. hederacea Sm. (Fig. 3B) and species allied to G. affinis de Vriese in subsect. Goodenia. A small group in this category from subg. Porphyranthus may have inflorescences that are panicle-like; however, some of the lateral inflorescences from the main stem appear to be monochasial cymes, where the youngest flowers emerge from the axil of a bract below older flowers or may be reduced to a single flower (D. Albrecht pers. comm.). Carolin (1967a) noted that G. rotundifolia R.Br. exhibited a Type 5 inflorescence (see table 1, Carolin 1967a), but this species tends to have cauline leaves and leafy racemes rather than a basal rosette, so its inflorescence is correctly categorised as Form A. Carolin also observed that among G. hederacea and allied species the main stem may not produce an inflorescence in a given year, presumably due to poor growing conditions, continuing with vegetative growth potentially for several years, as the inflorescences are entirely a product of the enrichment zone. Two inflorescence groups that were previously not characterised by Carolin are recognised here. A number of small, tufted herbaceous species in the Goodenia II subclade allied to G. convexa Carolin show a reduction of the inhibition zone and the vegetative branching zone to form a basal rosette of leaves, but with solitary, bracteolate flowers produced in the axils of the basal leaves. These species are categorised as having a Form C inflorescence (Fig. 3C). While another group of herbs from the Porphyranthus I clade including the diminutive G. chthonocephala Carolin, as well as G. kakadu and G. pumilio previously included in sect. Amphichila along with the recently described G. cravenii (Fig. 6C) and G. oenpelliensis R.L.Barrett, all have ebracteolate, solitary flowers in leaf axils, which is categorised here as Form D (Fig. 3D). It should be noted that G. oenpelliensis, which is currently only known from a single locality in the Northern Territory, has dimorphic inflorescences with both solitary flowers and short, ebracteolate cymes (Barrett and Barrett 2018), the latter type being categorised as Form F (see below and Fig. 3F).
Species characterised as having a Form E inflorescence herein include a diverse group from the Porphyranthus subclades of Clade B and representatives of the Coreulea, Tetrathylax, Verreauxia and Scaevolina subclades of Clade C that have a basal rosette of leaves and inflorescences with non-leafy bracts and bracteoles that form panicles (e.g. G. paniculata Sm., Fig. 3E), racemes (e.g. G. gracilis R.Br., Fig. 3E inset above), or a thyrse-like inflorescence (e.g. G. pterygosperma Krause, Fig. 3E inset below).
Species of Goodenia lacking bracteoles from subsects. Ebracteolatae and Borealis currently placed in the Ebraceolatae subclades of Clade B were variously categorised under Carolin's inflorescence Types 6-8. Carolin's Type 6 inflorescence morphology was characterised by leafy, ebracteolate racemes, as observed in various species in subsect. Ebracteolatae such as G. hispida R.Br. (Fig. 3F). Previously, G. pumilio, from sect. Amphichila, was also recorded as having a Type 6 inflorescence but as stated above, this morphology is now treated as Form D. Carolin's Type 7 group included species with an inflorescence similar in form to Type 6 but with reduced, non-leafy bracts (e.g. G. pinnatifida Schltdl., G. fascicularis F.Muell. & Tate (Fig. 3F inset above), and allied species). Carolin (1967a) noted that G. cycloptera R.Br. had both Type 6 and Type 7 inflorescences, while G. filiformis R.Br., also from subsect. Ebracteolatae, was documented as having both Type 7 and Type 8 inflorescences, the latter form characterised by an inflorescence where the internodes are shortened to form a subumbel and the bracts are leaf-like, as observed in G. concinna Benth. (Fig. 2C) and G. pulchella Benth. (Fig. 3F inset below). It is now evident that many species in the Ebraceolatae subclades of Clade B may exhibit variations in inflorescence morphology, particularly when growing under varying seasonal conditions, and so Carolin's inflorescence Types 6-8 are grouped together here under Form F (Table 2).
Carolin's Type 3 inflorescence (here treated as Form G) was defined as being the same as Type 1 but the bracteoles are reduced rather than leafy. This was observed in members of subg. Monochila, for example G. scapigera R.Br. (Figs 2G, 3G) and G. racemosa F.Muell., as well as Verreauxia reinwardtii (de Vriese) Benth.
Finally, Carolin (1967a) treated species in the genus Velleia as having a Bauplan that was a modification of the Type 1 form (recognised here as Form H), where the whole of the terminal "paracladium" (the enrichment zone) is contracted and the inflorescences are elongated with each partial inflorescence expanding into complex branching dichasia or "dichotomous axillary cymes" forming a significant component of the overall plant habit (Carolin 1967c). Carolin (1967a) also noted that the terminal bud apparently continues to grow from year to year.

Discussion
Taxonomic stability is important, particularly in species-rich groups that are horticulturally popular such as the family Goodeniaceae. Under-sampling in phylogenetic studies can result in premature taxonomic decisions as the addition of further taxa or more informative data may highlight significant incongruencies. This is often most problematic in groups with poor backbone resolution. In light of this, we have been reluctant to make taxonomic changes based on our previous molecular phylogenetic studies, particularly within the morphologically diverse Goodenia s.l. (Jabaily et al. 2012;Gardner et al. 2016a;Jabaily et al. 2018). Through ongoing studies, we now be-lieve we have addressed key sampling and data issues by including multiple accessions and combining genome skimming and Sanger sequencing from across multiple gene regions and genomic compartments. A number of potentially new phrase-named species (designated by 'sp.' and a relevant phrase name e.g. Goodenia sp. Mount Bomford (M.D. Barrett 423)) that are currently recognised on Australian plant name databases (Council of Heads of Australasian Herbaria, 2006-; Western Australian Herbarium, 1998-), and variants that show an affinity to but may be distinct from current species ('aff'), were also included in this study. Australia has a high level of species discovery and description (Wege et al. 2015), and yet many taxa that have been provisionally recognised as new are yet to be formally named and described. Western Australia is a centre for diversity for many groups including Goodenia with more than 70% of known species of this genus found there. Given that approximately 44% of the state's undescribed taxa are listed as being poorly known and of conservation concern (Smith and Jones 2018), with many facing continued significant threat due to land clearing and habitat fragmentation, fire, weed invasion, disease and climate change, it is essential that the description of new taxa is expedited. By including variants and phrase-named taxa in molecular phylogenies, their closest allied taxa can be confirmed, thus focusing taxonomic study to the most relevant species group to facilitate their taxonomic resolution.
The results of this molecular study reconfirm our earlier findings where Goodenia s.l. is paraphyletic with respect to Pentaptilon, Selliera, Velleia, and Verreauxia (Jabaily et al. 2012;Gardner et al. 2016a;Jabaily et al. 2018). Unfortunately, there are no obvious synapomorphies that define this broad group and yet the various included genera are relatively morphologically well circumscribed. This led to earlier suggestions that Goodenia could potentially be more narrowly defined to represent only the species within Clade A (including Selliera and Scaevola collaris), as the newly conserved type G. ovata (Shepherd et al. 2017) falls within this clade. In this case, Pentaptilon, Velleia and Verreauxia would be retained along with several newly reinstated or circumscribed segregate genera. However, this outcome would be significantly more taxonomically disruptive as around 160 name changes would be required, mostly in the species-rich Clade B where the earliest available name is Calogyne. Moreover, phylogenetic under-resolution and conflict remains within floristically diverse Clade C, likely due to recent radiation, possible hybridisation and incomplete lineage sorting. This, in conjunction with our re-assessment of key morphological characters including inflorescence form and ovary structure, and expanded molecular phylogenetic data, has led us to the pragmatic taxonomic decision to synonymise Pentaptilon, Selliera, Velleia, and Verreauxia into an expanded Goodenia.

Synonymisation of Diaspasis
Scaevola is not discussed in detail in this study other than to provide the new combination for the monotypic Western Australian genus Diaspasis (Jabaily et al. 2012). Diaspasis was first recognised as distinct by Robert Brown (1810) due to its nearly actinomorphic flowers (see Fig. 1E in Jabaily et al. 2012) and connate anthers. Scaevola by contrast, typically has fan-like flowers with free anthers (Carolin et al. 1992); however, shared characters between these genera include a dry indehiscent fruit with a hard endocarp and a spathulate embryo (cf. terete in the LAD clade sensu Jabaily et al. 2012) and similar trichomes (Carolin 1971). Moreover, Carolin (1959) noted that D. filifolia had a similar anatomical floral and stylar structure to Scaevola albida (Sm.) Druce and S. hookeri (de Vriese) F.Muell. ex Hook.f. Recent floral morphometric analyses of the Core Goodeniaceae have shown that floral symmetry is quite labile across the family with various species of Scaevola tending towards a pseudo-radial symmetry (e.g. S. phlebopetala F.Muell.: see Gardner et al. 2016b) somewhat like the form of D. filifolia. Molecular data also supports Diaspasis as congeneric with Scaevola (Howarth et al. 2003;Jabaily et al. 2012;Gardner et al. 2016a) and so a new combination for this species is provided.

No synapomorphic characters define Goodenia s.l.
Key characters previously used to distinguish Goodenia (Carolin 1992e) include bilabiate flowers, an inferior ovary that has 2 incomplete locules with > 2 ovules either present in two rows or scattered over the surface of the placenta, and fruits being dry, bi-valved, dehiscent capsules with flat seeds that have a rim or wing (Fig. 2F). Carolin (1992e) did note there were exceptions, such as G. neogoodenia and some representatives of the fan-flowered subg. Monochila, which have 1-seeded, indehiscent nuts, as does the newly included Scaevola collaris (Fig. 2B). Moreover, molecular sequence data show that species of Selliera that have indehiscent dry or fleshy fruits are also embedded within Goodenia. Clearly these various indehiscent fruits are superficially similar, but evidently nonhomologous, to the indehiscent fruits of Scaevola; a diagnostic character for that genus. Further, the fan-flower form typical for Scaevola has also evolved independently across every major clade in Goodenia (Gardner et al. 2016b). It was evident to Carolin (1966) that while Scaevola and Goodenia were allied, these genera had distinct evolutionary histories and that the "similarities in the ovary structure are the result of convergence rather than common origin" as the locules in the ovary of Goodenia are derived from two carpels rather than one as evident in Scaevola (Carolin 1959). While there are no easily discernible synapomorphies available for Goodenia s.l., this is a well-supported clade, so a combination of characters is required to recircumscribe this genus. Thus, a revised classification based on our understanding of phylogenetic relationships within the newly expanded Goodenia is outlined, recognising monophyletic groups at infrageneric levels (Table 1) including three newly circumscribed subgenera Goodenia, Porphyranthus and Monochila representing Clades A, B and C respectively.
In the combined molecular analysis, Goodenia I includes G. phillipsiae; a species previously placed in sect. Ebracteolatae despite the presence of bracteoles, although Carolin (1992e) did acknowledge that this was a species of "uncertain affinity" as apparent related species were ebracteolate. The typical subclade with G. ovata and allied species includes prostrate or decumbent subshrubs, many of which have long stoloniferous branches that may root at the nodes (Fig. 2I). These plants are usually glabrous or viscid, with bright yellow bilabiate flowers where the dorsal lobes are erect and sometimes overlapping (Fig.  5A, E-G). One exception is G. viscida (Fig. 5C), an erect fan-flowered subshrub from south-west Western Australia. Carolin et al. (1992) was also uncertain of the systematic placement of this species but included it in subg. Monochila due its yellow fan-shaped flowers, despite the fact G. viscida "has a seed-surface pattern unlike any other species [within the subgenus] and is therefore likely to be misplaced" (Carolin 1980).
The second subclade in Goodenia I includes Selliera, a small genus of prostrate, woody perennials with white or pale pink fan-flowers that have fleshy fruits that become woody and corky with age, which are found near coastal, winter-wet or saline flats (Fig. 5B) in Australia, New Zealand and Chile. Also in this clade is Scaevola collaris (Fig. 2C), a widespread fan-flowered species often found around the margins of salt lakes across arid Australia with a uniquely beaked fruit with a sponge-like woody endocarp. This unusual species is placed sister to G. laevis, which is confined to the southern regions of Western Australia and has capsular fruits as seen in the rest of sect. Goodenia. However, in a surprising twist, recent analysis of new ITS sequence data has shown that the enigmatic species Velleia exigua (F.Muell.) Carolin is in fact more closely allied to G. laevis (R. Jabaily, unpublished data). Initial attempts to molecular sequence V. exigua had failed, and its systematic position was equivocal. This species was previously included in Goodenia (as G. exigua F.Muell.) and while Carolin (1992g) transferred it to Velleia sect. Euthales, he noted that it was unlike other species of the genus due to the presence of solitary and almost sessile flowers, and sepals that were adnate to the ovary in the lower half (in contrast to all other species where the sepals are adnate to the base of the ovary). The indusium of V. exigua was also considered to be unique within Velleia, but on close inspection it is remarkably similar to that seen in Goodenia viscida and indeed Scaevola collaris, which was described by Carolin (1992g) as being obloid and longer than it is wide. Moreover, the indusium in these species is notched at the apex and has no obvious fringing hairs on the lips (Fig. 5C), although tiny hairs are present on the indusium of S. collaris. Based on morphological and molecular evidence, it is now clear that V. exigua should be included in sect. Goodenia.
The final subclade recovered in the Goodenia I clade includes several erect, glabrous subshrubs allied to G. kingiana Carolin, that have articulated pedicels and large bilabiate flowers where the dorsal petals are spreading to expose a long style supporting a broad indusium. The South Australian species G. saccata Carolin was not sequenced, but it is morphologically allied to G. grandiflora Sims and so is included in the typical section. The Goodenia II subclade, treated here as sect. Rosulatae (K.Krause) K.A.Sheph., includes erect or decumbent shrubs and herbs with rosulate and/or cauline leaves that are often covered in a dense tomentum of soft, multicellular hairs. Some species such as G. rotundifolia are variable and may be glabrous or have a mix of simple, glandular and multicellular hairs. There are two well supported subclades in this section that are not taxonomically recognised. In the first, G. fordiana Carolin is placed sister to a subclade of small tufted herbs from southern Australia allied to G. convexa. These species have cottony, multicellular hairs and solitary flowers that are sometimes supported by a distinctly geniculate pedicel that is sharply bent at the point of bracteole attachment. Sister to this, is a less well-resolved group allied to the widespread G. glabra R.Br. that includes decumbent herbs or subshrubs from northern and central Australia, which may also have a geniculate pedicel. The second monophyletic group in the Goodenia II subclade includes decumbent plants that occur in more arid inland regions of Australia with a centre of diversity in Queensland. Two potentially new species informally known as G. sp. Carnarvon Range (D.J. Edinger Nats 30) and G. sp. Mt Castletower (M.D. Crisp 2753) (Council of Heads of Australasian Herbaria 2006-), also fall within this clade. Finally, G. arenicola Carolin, a species currently only known from the type locality on Stradbroke Island in Queensland, as well as G. robusta (Benth.) K.Krause and G. rupestris Carolin, were not successfully sequenced, but are included in sect. Rosulatae as they share key diagnostic characters such as low prostrate habit and a tomentum of multicellular hairs. According to Carolin (1992e), G. stephensonii F.Muell. is allied to G. heterophylla Sm. and this species is placed in sect. Rosulatae for now; however, G. stephensonii is glandular hairy and somewhat viscid, features that are common to species included in sect. Goodenia.
Goodenia wilunensis, G. xanthotricha, and a group of species allied to G. quadrilocularis were all previously included in Carolin's (1992e) subsect. Goodenia, but based on molecular and morphological data these species are now excluded from our recircumscribed subg. Goodenia and will be discussed in later sections.

Recognition of Porphyranthus as a new subgenus (Goodenia Clade B)
Section Porphyranthus is elevated to subgeneric rank herein as subgen. Porphyranthus (G.Don) K.A.Sheph., which encompasses the variation evident across the monophyletic Clade B. Two sections corresponding to the two major subclades within this new subgenus are also recognised. G. Don's (1834) original sect. Porphyranthus, represented by Porphyranthus I and II in the molecular analyses, is expanded to include sect. Amphichila. While Krause's (1912) subsect. Ebracteolatae, encompassing the Ebracteolatae I and II clades, is recircumscribed to include both ser. Borealis and Calogyne of sect. Borealis and elevated to a section, recognised here as sect. Ebracteolatae (K.Krause) K.A.Sheph. (Table 1).
The reinstatement of G. rosulata Domin and recognition of a further three new species (Albrecht 2002;Holland and Boyle 2002;Pellow and Porter 2005;Sage and Shepherd 2007) in sect. Porphyranthus in recent decades has resulted in the expansion of this section from Carolin's (1992e) concept. The majority of species in this group are herbaceous annuals or perennials that grow in sandy soils and winter wet situations and creek beds in eastern and northern Australia, with G. purpurascens R.Br. also found in New Guinea (Carolin 1992e). These species generally have basal leaves and spreading, bracteolate inflorescences that may comprise a large part of the plant, and glossy, round seeds that are small (less 1 mm wide) with a very narrow mucilaginous wing c. 0.1 mm wide. Goodenia kakadu, G. pumilio and the recently described G. cravenii (Fig. 6C) and G. oenpelliensis Barrett 2014, 2018), are currently included in sect. Amphichila. These small, sometimes mat-like herbs have tiny reddish-purple fan-like flowers and small convex seeds with a minute wing. All four species are here transferred to sect. Porphyranthus along with G. chthonocephala, a species previously included in ser. Borealis that has an unusual cushion-like habit and tiny flowers held in a dense head at ground level. The Western Australian G. neogoodenia, originally recognised as the monotypic Neogoodenia minutiflora due to its tiny wingless corolla, and enlarged, indehiscent, 1-celled fruit (Gardner and George 1963), was transferred by Carolin (1990) to Goodenia (in subsect. Ebracteolatae). On molecular evidence it is clear G. neogoodenia is allied to a group of species in sect. Porphyranthus that are often associated with the margins of inland salt lakes including G. halophila and G. gypsicola Symon (Fig. 2H).
A were not sequenced; however, they are here included in sect. Porphyranthus due to the shared presence of a floriferous inflorescence, leafy bracts, bracteoles, and small seeds, with G. nocoleche also recorded as producing aquatic leaves under flood conditions (Pellow and Porter 2005).
Over the last two decades, 10 new species have been recognised in Carolin's (1992e) subsect. Ebracteolatae (Sage 2000;Sage 2001;Sage and Dixon 2005;Sage and Shepherd 2007;Barrett and Barrett 2014;Barrett and Barrett 2018). This is currently the largest infrageneric group in Goodenia and is characterised by a lack of bracteoles, generally yellow flowers, and distinctively winged seeds. Many species are annuals or herbaceous perennials found in the more arid regions of the Australian continent. Indeed, it is evident that the Eremaean interior has been an important source and sink for diversification within this group (Jabaily et al. 2014). In these arid regions, many species are confined to damp areas around the margins of creeks and lakes that germinate or regenerate from rootstock after significant cyclonic rainfall (Sage and Pigott 2003), thus 'avoiding' harsher seasonal conditions during the long dry season. Of interest is the Western Australian bracteolate G. wilunensis that placed sister to the Ebracteolatae clade. While this section is characterised as generally being ebracteolate, there are a few other species within this clade that do retain this character, such as G. nigrescens Carolin and G. cirrifica F.Muell. As stated, sect. Ebracteolatae as recognised here, is expanded to include the former ser. Borealis and Calogyne of subsect. Borealis, a group characterised by leafy inflorescences, a lack of bracteoles and seeds with a prominent rim rather than an obvious wing. Since the Flora of Australia treatment (Carolin 1992e), three new species, G. inundata L.W.Sage & J.P.Piggot, G. debilis A.E.Holland & T.P.Boyle and G. elaiosoma Cowie, have been included in ser. Borealis (Sage 2001;Holland and Boyle 2002;Cowie 2005), a group segregated on the presence of a simple style and broad sepals to 2.5 mm wide, in contrast to the divided style and narrow sepals to 0.4 mm wide that distinguished ser. Calogyne. Species in this former subsection are found in northern Australia, with the exception of G. armstrongiana de Vriese (ser. Borealis), which also occurs in New Guinea, while the widespread and variable G. pilosa (R.Br.) Carolin (ser. Calogyne) is found in damp areas in Northern Australia, New Guinea, Indonesia, Malaysia, Philippines and China.
Further species currently included in Carolin's subsect. Ebracteolatae, such as G. concinna (Fig. 2C), G. symonii (Carolin) Carolin, G. fascicularis, the recently recognised G. effusa A.E.Holland (Holland 2015), and reinstated G. pritzelii (Barrett and Barrett 2018), were not successfully sequenced. Material was also unavailable for the very poorly understood G. salmoniana (F.Muell.) Carolin and G. pallida Carolin, which are only known from type collections from the Gascoyne and Pilbara regions in Western Australia. All of these species are retained within this newly circumscribed section due to the presence of diagnostic characters and the confirmed phylogenetic position of morphologically allied species; however, the position of G. salmoniana is equivocal as this species was originally placed in Velleia by Mueller (as V. salmoniana F.Muell.), most likely because its sepals are fused to the lower half of the ovary and the indusium lips are glabrous, unlike other species in this group.

Expansion of subg. Monochila (Goodenia Clade C)
Clade C is the most morphologically diverse clade in Goodenia s.l. and, while relationships between some of the subclades are unclear, each is generally supported as monophyletic. As such, subg. Monochila is expanded herein to include all members of Clade C, with most subclades formally named at the sectional level.
Typical sect. Monochila is easily recognised as all members (except the newly included G. sericostachya C.A.Gardner) have white, fan-shaped flowers (Fig. 2D) and a narrow indusium that is supported by a style covered in stiff, short, spreading hairs (Carolin 1992e). G. sericostachya, a narrow range endemic from Western Australia, was previously included in subsect. Scaevolina due to its dense indumentum of silver white hairs and pink fan-like flowers with a yellow throat (Fig. 8G). However, on close inspection it is evident that this species has the distinctive short hairs on the style and the narrow indusium that are diagnostic for sect. Monochila, confirming its affinity to other species in this clade. Two subsections are further recognised in sect. Monochila. The typical subsection includes all species with short white hairs on the style and capsular fruits, while the remaining species with short purple hairs on the style and a nut-like fruit are now included in a new subsection named herein as subsect. Infracta K.A.Sheph.
The majority of Carolin's (1992e) species included in his subsect. Goodenia fall within Goodenia Clade A; however, a small clade of yellow-flowered species were found to be nested within Goodenia Clade C (Jabaily et al. 2012;Gardner et al. 2016a). Don's (1834) sect. Tetrathylax G.Don, which previously only comprised the Western Australian narrow range endemic G. quadrilocularis (Figs 2A, 8D), is here resurrected and expanded to include this group, represented by several diploid and polyploid taxa (Peacock 1963) from eastern Australia. Sect. Tetrathylax is superficially similar to species in Clade A, due to the presence of yellow, bilabiate flowers, but this section is characterised by distinctive inflorescences comprising long, leafless, interrupted spikes, racemes or panicles where the upper pedicels have short, linear bracteoles close to the flowers with the lower ones being more distant. Carolin (1980) also noted that this group is characterised by unique seeds, that have a "sinuous-areolate" seed coat with the shape of the radial wall thickening as "type 4". Goodenia rostrivalvis Domin was not sequenced, but is included in sect. Tetrathylax due to its morphological similarity to allied species such as G. decurrens R.Br.
In Carolin's (1992e) Flora of Australia treatment, sect. Coeruleae comprises subsections Scaevolina and Coeruleae, and is represented by blue-flowered species of Goodenia where the septum of the ovary is at least 2/3 as long as the locule. In our molecular analyses, these subsections are each supported as monophyletic, but they never group together (Jabaily et al. 2018). Accordingly, they are treated as separate sections in our new classification. Sect. Coeruleae is re-circumscribed here to only include the members of the former subsect. Coeruleae, representing the blue-flowered perennial herbs and low subshrubs from southwest Western Australia with seeds that have a dry, membranous wing greater than 0.1 mm wide. This section now also includes G. katabudjar Cranfield & L.W.Sage and G. lancifolia L.W. Sage & Cranfield (Cranfield and Sage 1997;Sage 2000). The latter species was not placed in the phylogeny, along with G. leptoclada Benth., due to poor sequence results but both species share the diagnostic characters of this section.
Sect. Scaevolina represents the predominantly northern Australian blue-flowered perennials that have seeds with a narrow, mucilaginous wing c. 1 mm wide and has been expanded to include G. azurea subsp. hesperia L.W.Sage & Albr., G. hartiana L.W.Sage and G. splendida A.E.Holland & T.P.Boyle since the publication of Carolin's Flora treatment (Holland and Boyle 2002;Sage and Albrecht 2006).
Two species previously included in subsect. Scaevolina have a more southern distribution than typical. The first is the unusual fan-flowered G. sericostachya that is now included in sect. Monochila. The second is the rare species G. arthrotricha (Smith and Jones 2018), whose broader phylogenetic relationships remain unclear. In all molecular analyses, this species forms a well-supported clade with the short-range endemic G. xanthotricha (Fig. 8A), but their relationship to other subclades remains equivocal.
While G. xanthotricha was previously included in sect. Goodenia, Carolin (1992e) acknowledged that it was "a species difficult to place" noting that even though it has blue flowers, the seed coat ornamentation is "aculeate" rather than "colliculate-punctate" as seen in other members of subsect. Scaevolina. Furthermore, though members of the current sect. Scaevolina generally have an indusium that is longer than wide, the indusium in G. arthrotricha is wider than long, similar to that observed in many of the species of sect. Coeruleae. Goodenia arthrotricha and G. xanthotricha form a very weakly supported association with the southern Coeruleae clade in the nrITS analysis, but neither species has a seed with a dry wing > 0.1 mm wide (see the seed rim in Fig.  2F) that characterise this group. Both G. arthrotricha and G. xanthotricha are naturally rare, but somewhat surprisingly, their distribution overlaps as both species are found in a nature reserve situated on the Dandaragan plateau in Western Australia, although never observed as co-occurring (Western Australian Herbarium 1998-). One could hypothesise that their relatively close situation and morphological features that show some congruence to both sect. Coeruleae and Scaevolina, could suggest that one or both may be of possible hybrid origin. As these species are difficult to place systematically and no obvious synapomorphy supports them as distinct from other blue-flowered species, they currently remain unplaced within subg. Monochila.
Verreauxia is a small genus of three species from southwest Western Australia characterised by simple, unbranched and branched multicellular hairs, glandular hairs with multicellular heads (Carolin 1971), and a unilocular ovary that becomes indehiscent and nut-like in fruit. The closely allied monotypic Pentaptilon (Carolin 1992h), which occurs around the northernmost border between the South-West and Eremaean Botanical Provinces in Western Australia, has similarly unusual branched hairs; however, it was recognised as distinct due to its uniquely winged ovary and fruit, and morphologically distinct seeds. These genera group together in a monophyletic subclade within Clade C and so are recognised here as sect. Verreauxia in subg. Monochila.
The final monophyletic group that consistently placed sister to the rest of the morphologically diverse Clade C (Jabaily et al. 2018) is the genus Velleia characterised by inflorescences that are axillary dichasia, which form most of the plant habit (although sometimes flowers may be solitary), and a predominantly superior ovary (Fig. 2G). Carolin (1980) also noted that while the seeds in some Velleia showed similarity to various species of Goodenia, a number displayed a 'characteristic wrinkling'. Carolin's (1992g) infrageneric classification of Velleia recognised three sections, based on the presence of three sepals (sect. Velleia) or five, which may be either connate into a tube (sect. Euthales) or free (sect. Menoceras); however, in our analyses sect. Menoceras was shown to be paraphyletic and there was only moderate support for some of these former sections in the chloroplast analyses. As such, we propose here to reduce Velleia to a section of subg. Monochila and to not formally recognise any further groups within it. Velleia parvisepta Carolin and V. perfoliata R.Br., a narrow range endemic from New South Wales, while not sequenced are retained within this section as both species have the typical Form H inflorescence and sepals fused to near the base of the ovary, which are characters typical for sect. Velleia. It should be noted that while V. perfoliata is placed after V. macrocalyx de Vriese in the proposed updated linear sequence for the genus (Table 2), Carolin (1992g) noted that this poorly known species had connate bracteoles that form a disk-like funnel unlike other species in his sect. Velleia. Three species in the former sect. Menoceras (V. discophora F.Muell., V. panduriformis A.Cunn. ex Benth. and V. connata F.Muell.) have similarly modified bracteoles to V. perfoliata (Table 2) and future sequencing of this species may confirm it is more closely allied to this group of taxa.

Conclusion
While this could be considered the final chapter of our detailed study of Goodenia s.l., resulting in a new understanding and an updated classification of this captivating group, it is not likely to be the final word. Goodenia s.l. represents an outstanding clade for further studies, particularly of inflorescence and floral form, seed traits, and the potential impacts of adaptations on rates of diversification. These well sampled and resolved phylogenies also allow for the inclusion of Goodeniaceae in meta-studies of diversification patterns across Australia and in other biodiversity hotspots like the Southwest Australian Floristic Region (SWAFR) (Jabaily et al. 2014). Furthermore, this framework will support more in-depth studies at the species level to hopefully expedite the recognition of many new but as yet unnamed taxa.

Taxonomic treatment
In this present treatment, revised descriptions of infrageneric groups are provided with a synopsis of the species currently recognised therein, including updated nomenclatural changes. An updated key to genera in the family, including Brunonia (previously placed in the monotypic Brunoniaceae), and incorporating Selliera, Pentaptilon, Velleia and Verreauxia into Goodenia is also provided. A key is also provided to infrageneric groups as recognised in this paper.
Key to Genera (modified from Carolin et al. 1992, previously  Perennial shrubs or subshrubs, or annual or perennial herbs, sometimes stoloniferous and rooting at nodes; glabrous, or with simple (sometimes multicellular) hairs, or viscid with glandular hairs. Leaves basal and/or cauline, petiolate or sessile, entire to pinnatifid, usually with axillary hairs. Inflorescence a raceme, thyrse, spike, panicle, subumbel, axillary dichasia, or flowers solitary in axils of basal leaves; pedicels sometimes articulate, rarely geniculate, with or without * For synonymy, see below under infrageneric taxa. bracteoles. Sepals 5 or 3, fused or free, variously adnate to ovary. Corolla bilabiate or fan-like (lobes almost equal), white, cream, yellow, orange, pink, mauve, blue or purple; corolla-lobes usually winged, sometimes unequally; with hairs in the throat (rarely glabrous), sometimes with enations; often auriculate; sometimes with pouch or spur; stamens free, epigynous or hypogynous; style simple or 2-4-fid, glabrous or with simple hairs; indusia 1-4, 2-lipped, usually with bristles on lips; ovary inferior or superior, rarely winged, usually incompletely 2-locular with few to many ovules either in two rows or scattered over surface of the placentas, or solitary. Fruit a 2-or 4-valved capsule (rarely fleshy), 1-seeded nut, 4-seeded hard drupe or rarely a soft, indehiscent fruit with wings (G. careyi). Seeds flat or biconvex, usually with a rim or wing that is sometimes reduced.
Number of taxa and distribution. The genus has c. 251 taxa and is predominantly Australian. Goodenia pilosa extends to New Guinea, Indonesia, Malaysia, southern China and Philippines, while G. armstrongiana, G. purpurascens and G. pumilio extend to New Guinea and G. koningsbergeri occurs in India, Thailand, Cambodia, Malaysia and Indonesia according to Karthigeyan et al. (2009). Species previously included in Selliera also occur in coastal habitats in New Zealand and South America.

Goodenia Sm. subg. Goodenia
Description. Shrubs, subshrubs or herbs, erect, decumbent or prostrate, sometimes stoloniferous and rooting at nodes. Leaves basal, cauline or both with the upper leaves sometimes smaller and narrower. Flowers in thyrses, racemes, spikes, or flowers solitary in leaf axils; bracts usually leafy; bracteoles usually present, pedicel infrequently geniculate, articulate or not. Corolla bilabiate or fan-like, usually yellow, sometimes white, cream or blueish purple, rarely pink; throat usually with scattered hairs, often with enations, not auriculate and often with a pouch. Style simple. Ovary with a variable septum from very short to 2/3 as long as locule; ovules in 2 rows in each locule or scattered, rarely solitary. Fruit a capsule with 2 valves, persistent or deciduous, rarely a fleshy fruit or 1-seeded nut. Seeds with a wing 0.1-0.2 mm wide and mucilaginous or obsolete.
Number of taxa and distribution. This subgenus currently includes 51 species, with 47 confined to Australia and three also occurring in New Zealand, Chile and southern Asia.  Description. Herbs or occasionally subshrubs, usually with multicellular hairs sometimes becoming glabrescent, or rarely with simple and glandular hairs. Leaves basal and/or cauline. Flowers usually in racemes or solitary in leaf axils; pedicels usually not articulate. Corolla bilabiate.

Number of taxa and distribution.
This section includes 28 species found in every state of Australia across a range of biomes with some species extending into arid central inland regions.
Number of taxa and distribution. The subgenus Monochila includes 58 species across six sections. Western Australia is a centre of diversity for this group with species from sect. Velleia also found in Eastern Australia and one species present in New Guinea. Description. Shrubs, subshrubs or herbs. Leaves basal, cauline or both. Flowers in thyrses or spikes; usually with bracteose bracts; bracteoles present. Sepals equal. Corolla fan-like, white with brown or purple spots at the base of each lobe or pink with a bright yellow throat, with stiff hairs in throat, enations absent, without auricles; pouch inconspicuous, to 1/2 ovary length. Ovary 2-locular with ovules solitary or to 40, usually in 2-rows in each locule. Fruit either a capsule with valves bifid or indehiscent and nutlike. Seeds to c.1 mm, wing < 0.5 mm and mucilaginous or obsolete.
Number of taxa and distribution. A Western Australian section of 10 species. Note. Don (1834) recognised the section Tetrathylax (meaning four -pouch) to include G. quadrilocularis. The name was formed providing the Greek and Latin translations for tetras (four-fold) and thylax (a cell) in recognition of the 4-celled condition of the capsule. de Vriese treated the section at generic rank with the incorrect spelling 'Tetraphylax', which was followed by subsequent authors until corrected by Carolin (1992e Description. Herbs with short stems. Leaves basal or cauline. Flowers in axillary dichasia with scapes erect to prostrate or flowers solitary in axils usually with bracteose bracts; bracteoles present and sometimes disc-like. Sepals equal or adaxial one larger. Corolla bilabiate yellow, orange, mauve, pink or white, with or without hairs in the throat, enations absent or present, auriculate, pouch absent or present, sometimes forming a spur. Ovary 2-locular with 4-25 ovules. Fruit a capsule with with 2 or 4 valves. Seeds 1.5-6 mm long, wing 0.5-2 mm wide or with a thickened rim.   Morris 1976 andSymes 1969, respectively). A collection at AD (AD 97604803) may represent original material of this name. The collection comprises two flowering branchlets of Goodenia collaris, is labelled "Scaevola collaris F.v.Muell. / Lake Eyre" in an unknown hand, and matches the description given in the protologue, with the exception of fruits,  de Vriese (1854) in the protologue for Dampiera verreauxii, although de Vriese indicates that he examined material of this taxon made available by Verreaux at the Muséum National d'Histoire Naturelle in Paris (P) in 1850. Verreaux visited Australia during 1842-1846 but his collecting efforts were confined to Tasmania and the east coast of Australia (George 2009) and so it is unlikely he would have obtained material from Western Australia directly. Carolin (1992i) postulated that "The type was probably collected by J. Drummond (Drummond 4: 186 cited by Krause, loc. cit. [= Pflanzenr. 54: 170 (1912)])". This is possible, as Drummond did collect extensively through south-west Western Australia (including the region where this species occurs) and his specimens were sent to various institutions throughout Australia and Europe. Three Drummond collections of this taxon have been located (MEL 42188 [image!], P 03035588 [image!] and P 04057856 [image!]). However, it is not clear whether these specimens represent original material and neither specimen at P is an exact match for the plant illustrated in the protologue. Accordingly, the illustration included in the protologue is here designated as the lectotype for Dampiera verreauxii de Vriese. Russell Barrett, and associate editor Clifford Morden. We would also like to finally thank our families, friends and past students for their support. This study is the culmination of many years of research by an international collaboration on a predominantly Australian plant group, as such we acknowledge the Traditional Custodians of the lands on which these plants grow and pay our sincere respects to their Elders past, present and emerging.