An illustrated checklist of the genus Elymnias Hübner, 1818 (Nymphalidae, Satyrinae)

Abstract We review the genus Elymnias Hübner, 1818, a morphologically diverse satyrine butterfly clade involved in multifarious Batesian mimicry relationships throughout Asia and Africa. A variety of different model species are mimicked, and many Elymnias species are sexually dimorphic mimics, with males and females resembling different model species. We revise species and subspecies delimitations in light of an integrative taxonomic investigation using external morphology, male and female genital morphology, and a multi-locus molecular phylogeny. There is little interspecific genitalic variation among species in this group, and previous taxonomists therefore relied almost entirely on wing patterns. Our molecular phylogenetic analysis reveals several examples of polymorphism or wing pattern divergence within a single species currently classified as two or more different species. We also found examples of wing pattern convergence among disparate lineages that mimic the same widespread model species. Frequently, two or more phenotypically similar species were classified as a single species. This comprehensive checklist reviews all names associated with Elymnias to align its taxonomy with the evolutionary history of the group. All available information on nomenclature, type localities, repositories of type specimens, and geographical distributions is summarized, and images of adult specimens and genitalia are provided along with distribution maps of all species and selected subspecies. We identify 2 species incertae sedis, establish 15 monophyletic species groups (including 1 species unplaced in any species group), and make 49 taxonomic changes, including 35 new synonyms, 7 new combinations (2 of which have new status), 1 resurrected combination, 1 resurrected subspecies, and 7 status changes.


Introduction
Elymnias Hübner, 1818 (Nymphalidae: Satyrinae) is a species-rich and widespread butterfly genus distributed throughout the Old World tropics (Aoki et al. 1982). It derives its name from Elymnias jynx Hübner, 1818 (now Elymnias hypermnestra hypermnestra (Linnaeus, 1763)) (Fruhstorfer 1907). The genus' widespread range and remarkable wing pattern diversity, together with the economic importance of several species, has attracted the attention of taxonomists and agricultural entomologists for centuries (Merrett 1993;Wallace 1869). However, the entire group has not been examined systematically in over 100 years (Fruhstorfer 1907;1911), and no studies to date have used genetic data to substantiate taxonomic hypotheses. Most species are found in the Indo-Australian Archipelago, a geographically complex and geologically dynamic area with over 20,000 islands that are likely to have contributed to diversification in this taxon (Lohman et al. 2011).
We recognize a single Afrotropical species with two subspecies and 52 Australasian species with 181 subspecies distributed from Nepal to Sri Lanka in the west, throughout tropical and subtropical Asia, and extending east to Taiwan and south to Australia's Cape York and the Bismark Archipelago of Papua New Guinea. A few species are widespread across several countries or landmasses, but many are restricted to single islands. Several new species have been discovered recently (Monastyrskii 2004;Okubo 2010;Saito and Koshi 2012), highlighting the rarity of many species and their predilection for relatively inaccessible locales, such as high mountains and remote islands.
Species in the genus differ markedly in wing color, pattern, shape, and size, making Elymnias one of the most morphologically heterogeneous butterfly genera (Feltwell 1993). This morphological diversity is apparently because most species are Batesian mimics of strikingly different, unpalatable model butterfly species (Corbet 1933;. Many Elymnias species are monomorphic; conspecific males and females mimic the same model species (e.g., E. paradoxa and E. vasudeva). However, some species are sexually dimorphic mimics (Moore 1894;Punnett 1915;Vane-Wright 1976), with males and females mimicking different models and differing markedly in wing color and pattern (e.g., E. kuenstleri and E. harterti; Corbet et al. 1992;Parsons 2000). A few sexually dimorphic Elymnias species mimic a single sexually dimorphic model species, such as E. casiphone and E. saueri, which both mimic Euploea mulciber. A few species exhibit variability in sexual dimorphism: males and females in some populations mimic the same model species and are monomorphic, while the same species is strongly dimorphic in other locales (e.g., E. hypermnestra and E. agondas).
Larvae of all species with known life histories feed exclusively on palms (Arecaceae) (Bascombe et al. 1999;Ek-Amnuay 2012;Parsons 2000;Robinson et al. 2017), and several species are also agricultural pests on economically important oil palm, Elaeis guineensis (Koh and Gan 2007;Merrett 1993). Adults are known to feed on exudates from rotting fruit (Treadaway and Schroeder 2012). Palm-feeding butterflies are not known to sequester noxious secondary compounds from their larval hosts, and naïve avian predators readily consumed adult E. hypermnestra in laboratory trials (S.-H. Yen, unpublished results), supporting the hypothesis that Elymnias are Batesian and not Müllerian mimics.
This checklist enumerates and verifies all current combinations and synonyms, and provides original literature citations, type localities, repositories of type specimens, photographs of specimens and genitalia, and maps of each subspecies' geographical range and type locality. Integrative taxonomic practice employing multi-locus molecular phylogenetics in concert with data from wing and genitalic characters has informed taxonomic decisions to retain or revise contentious classification and nomenclature. This checklist is meant to clarify taxonomic problems in the genus and aid biologists interested in studying butterfly biodiversity, but will also serve as a framework for future studies on the phylogeny, biogeography, wing pattern evolution, and speciation of this fascinating radiation of Batesian mimetic butterfly species.

Examination of original literature and type specimens
The taxonomic changes we propose are based on examinations of hundreds of specimens in dozens of museums, quantification of wing and genitalic characters including over 100 dissections of males and females (Wei et al. in prep.), and a multi-locus molecular phylogenetic analysis based on six genetic markers from over 200 specimens including nearly every species that we recognize (Lohman et al. in prep.).
Verification of type specimens was based on information provided in the original literature as well as critical review of the collection of specific authors, especially Fruhstorfer. All taxonomic treatments proposed in the present study, including the availability of infrasubspecific taxa, follow regulations and suggestions of the latest version of ICZN (1999).
All publications with original descriptions of new taxa or describing new taxonomic acts were consulted to verify the status and collection localities of type material. Geographical information was obtained directly from specimen labels and from literature to provide accurate locality data and minimize misinterpretation of geographical localities caused by misidentified or mislabeled specimens.
All images of specimens photographed in various museum collections are used here with permission from each museum. Except for the photographs provided by KUTH (Department of Entomology, Kasetsart University), David J. Lohman, and the Museum of Comparative Zoology, Harvard University, all the other photographs were taken by Chia-Hsuan Wei and Shen-Horn Yen.
The following abbreviations are used to specify the repository of type material. Specimens, including type specimens, were borrowed and/or photographed from many of these institutions and private collections.

Abbreviations of specimen repositories
DNPFIC Forest Insect Collection, Department of National Parks, Wildlife and Plant Conservation, Thailand time since divergence from their sister taxon. We regard subspecies as phenotypically distinctive geographic variants and do not expect them to be monophyletic or reproductively isolated from other subspecies (Braby et al. 2012). However, a subspecies should be differentiable from other conspecific subspecies using morphology, genetic data, or other characteristics. Given the within-species morphological variability known from this and other mimetic butterfly taxa (Punnett 1915), we adopted the following procedures for evaluating the validity of prior taxonomic hypotheses: (1) We reconstructed phylogenies using both morphological (Wei et al. in prep.) and molecular data (Lohman et al. in prep.), and used these as guides for interpreting relationships among species and taxonomic boundaries within species complexes; (2) In these analyses, we endeavored to include specimens from the type localities (or the surrounding area-at least the same island group) of the nominotypical subspecies to substantiate taxonomic boundaries of geographically widespread species with multiple subspecies (e.g., agondas, casiphone, hypermnestra, nesaea, and panthera); (3) We considered the geological history of a species' range (Hall 2001;Hall and Smyth 2008;Sathiamurthy and Voris 2006), particularly for taxa that are rare in museum collections and not readily available for morphological or molecular study. For example, present-day Sulawesi comprises multiple terranes, some of which originated in different biogeographical subregions, that collided in the Miocene (Lohman et al. 2011;Stelbrink et al. 2012). If two or more subspecies of the same species are described from a large and geologically complex island such as Sulawesi or New Guinea and we had limited material for phylogenetic study, then we generally retained the landmass's different subspecies for lack of evidence to synonymize them; (4) When genetic and/or ecological data suggested that different names had been applied to different mimetic forms, sexes, or seasonal forms, we synonymized these taxa.
For convenience, we have divided the genus into 15 monophyletic species groups (Lohman et al. in prep.) named after each clade's oldest named species. Because of the uniformity of genitalia and extreme intraspecific variability in wing patterns, there are few if any morphological synapomorphies that can be used to discriminate these species groups. They have been circumscribed based on relatedness as inferred by a multilocus molecular phylogeny.

Distribution maps and type localities
A variety of sources were used to infer the distribution maps that we provide, including museum data, taxonomic and other publications (Aoki et al. 1982;Braby 2000;Ek-Amnuay 2012;Hanafusa 2001;Inayoshi 2017;Monastyrskii 2004;Okubo 2010;Parsons 2000;Saito and Koshi 2012;Suzuki 2006;Tateishi 2001;Treadaway and Schroeder 2012;Uémura and Kitamura 2001;Vane-Wright and de Jong 2003). The majority of Elymnias taxa were described when most of South and Southeast Asia were colonized by European countries, and many of the type locality names given in the species descriptions have changed since colonial times. Therefore, in addition to the original type locality names given in the species description, we have attempted to provide the modern locality names in parentheses. In the text below, we do not attempt to use present knowledge of the taxon's distribution to infer the precise location where the type was collected. However, when designating type localities on the distribution maps, we have attempted to use knowledge of the taxon's current distribution and other information to indicate the type locality as precisely as possible. Nonetheless, many type localities are imprecise and cannot be localized because many labels simply list the island where the specimen was found (e.g., New Guinea) rather than a precise locality. Elymniopsis has often been regarded as a distinct genus since its establishment in 1907 by Fruhstorfer, and most references of Afrotropical butterflies list it as a genus of its own (e.g., Larsen 2005). Hemming (1943) first synonymized this genus with Elymnias and the opinion was followed by Gardiner (2010) and further supported by the phylogenetic studies by Peña et al. (2006). In our molecular phylogenetic study (Lohman et al. in prep.), this taxon is sister to all of the Asian species. However, aside from their wing patterns, which mimic various Acraea spp. (Nymphalidae: Heliconiinae), the morphological features of this species are not distinct from other Elymnias (Wei et al. in prep.), and we refrain from retaining the monotypic genus Elymniopsis.

ssp. euploeoides Talbot, 1932
Elymnias euploeoides Talbot, 1932 The distinctiveness of esaca and vasudeva has never been doubted and they have been treated as distinct species in all prior studies. The former is distributed throughout most of the Greater Sunda Islands, the Philippines, and the Thai-Malay peninsula, and the range of the latter encompasses northeast India, Myanmar, northern Laos, northern Vietnam and southwest China. The wings of male esaca are shorter and more attenuate than vasudeva, and have black ground coloration with a metallic submarginal band in some specimens. The male of vasudeva is not dramatically different from the female in wing shape or color pattern. However, our molecular phylogenetic analysis (Lohman et al. in prep.) reveals that both specimens of vasudeva (from China and India) are nested within a clade of three esaca specimens from Java, Mindanao, and peninsular Malaysia. This paraphyletic relationship suggests that the two species should be synonymized. However, we regard both species as valid because: 1) wing color and pattern are strongly dimorphic in esaca, but more or less monomorphic in vasudeva; 2) the wing shape of esaca males differs from females, and this is not true of vasudeva; and 3) the two taxa are parapatric. 7 "Battak Mountains" is the type locality for many butterfly and other animal taxa, but the name appears on no recent map of Sumatra. It seems to refer to the mountainous region historically inhabited by the Batak ethnic groups in northeast Sumatra. From the description in de Nicéville and Martin (1895), "Battak Mountains" seems to refer to the portion of the Barisan Mountain Range running along the western edge of North Sumatra Province, including the peaks surrounding Lake Toba.
The subspecies esaca taeniola is synonymized with esaca borneensis because there are no consistent morphological differences between them and no obvious biogeographical barriers within the island Borneo that would restrict gene flow and maintain subspecific differences. 9 oberthuri was originally described as a species by Fruhstorfer (1902b) and subsequently downgraded to be a subspecies of vasudeva (Fruhstorfer 1907). We synonymize it with esaca andersonii because these two names seem to represent opposite sexes of the same subspecies confined to the Thai-Malay Peninsula.

ssp. nigricans Tateishi, 2001
Elymnias esaca nigricans Tateishi, 2001  11 Staudinger (1889) described albofasciata based on specimens from Palawan. However, the locality of a "type specimen" deposited in ZMHB is labeled as "Tanyong Malim, Malacca". We consider the type to be either mislabeled or simply not a type of this subspecies. Staudinger (1889) indicated that he compared specimens of dara from Malacca (specimen provided by Künstler) and Palawan. The mistake in labeling is probably caused by historical confusion of the type locality since its publication. 12 Moore (1894) described patnoides as a distinct species, but Fruhstorfer (1907)    it, but not its senior homonym, is included in a relevant adopted Part of the List of Available Names in Zoology.." However, as already stated by Lamas (2010), hypermnestra Linnaeus does not fulfill any of the three conditions specified above, because: 1) hypermnestra Linnaeus has not been maintained as a nomen protectum and does not fulfill the conditions specified in Article 23.9; 2) hypermnestra Linnaeus has not been conserved by the Commission under Article 81; and 3) no part of the List of Available Names in Zoology has been adopted yet for Lepidoptera. Meanwhile, hypermnestra Scopoli has been used numerous times as a valid name after 1899 [see Article 23.9.1.1], and therefore does not qualify as a nomen oblitum. Consequently, if use of the younger homonym [hypermnestra Linnaeus] is to be maintained, the case needs to be submitted to the International Commission for Zoological Nomenclature for a ruling under the plenary power (Article 81). In addition, if Linnaeus's hypermnestra is eventually considered invalid by the Commission, the other earliest available name for this species is Papilio undularis Drury, 1773, which is now used to represent the subspecies of northeast India, and the valid subspecific name for the population of Java would be protogenia Cramer, 1779. 18 Roepke's atrata has rarely been mentioned in previous literature, and, having examined the original description, we regard it as a junior synonym of the nominotypical subspecies, which is also from Java. , there is another female form with whitish hindwings (the forms obfuscata and paraleuca) that possibly mimics either D. melanippus or D. affinis in Thailand and Vietnam. Since the current subspecies classification has been adopted by local guidebooks and other publications for so long (Corbet et al. 1992;Ek-Amnuay 2012;Monastyrskii 2005;Pinratana 1988), we do not propose any nomenclatural change prior to a thorough phylogenetic/population genetics study based dense sampling of the entire region is completed. 22 Moore's hainana was described from specimens from Hainan Island, China, and the name has been applied to the Taiwanese population since the late 19 th century. Since Hainan is between China's Guangxi Province and northern Vietnam, where septentrionalis and tonkiniana were described, respectively, and because examination of dozens of specimens evince no consistent morphological differences among these subspecies, we synonymize these three names and regard hainana as the valid name. 23 Penang is a small island in the Andaman Sea lying just off the western coast of peninsular Malaysia.
The strait that separates this small island (293 km 2 ) from the peninsula is only 2-8 km wide, yet seems to form a dispersal barrier between the peninsular population (commonly known as agina, but herein changed to beatrice, see discussion below) and insular discrepans. The female type specimen of discrepans seems to be a morphologically intermediate form between orange, Danaus-mimicking tinctoria and dark, Euploea-mimicking phenotypes. This phenotype has not been documented from the mainland. We therefore retain the name discrepans because of the taxon's distinctive female wing patterns; further studies will ascertain whether this subspecies is genetically distinct from other hypermnestra subspecies.  (1871) already stated that the type locality of the "real nigrescens" is Sarawak (Borneo), and the female wing patterns of the populations in Borneo and Malay Peninsula are slightly different, it is not appropriate to apply the name nigrescens to the peninsular population. Fruhstorfer (1902) noticed that "nigrescens" sensu Distant was different from the Bornean one, so he proposed a nomen novum, beatrice, to refer to "nigrescens sensu Distant". The concept of Fruhstorfer's beatrice, however, is probably not monophyletic as he listed Perak, Lingga (Riau), Deli (North Sumatra), and Sumatra in the geographical range of beatrice, but our morphological study does not support lumping the Sumatran population with the peninsular Malaysian one. In the same publication, Fruhstorfer (1902a) described an aberration of nigrescens, namely agina, for populations in Singapore, Sumatra, and Perak. The name agina has been used much more frequently than beatrice to represent the population in the Malay Peninsula including Singapore (Corbet 1943;Lamas 2010;Pinratana 1988 trice and agina became available for representing the populations in Malay Peninsula and Singapore. Determination of their validity, therefore, depends on the priority. Considering the fact that Corbet's (1943) use of agina is much later than Fruhstorfer's (1902a) proposal of a nomen novum, we conclude that beatrice should be used to represent the populations in southern part of the Malay Peninsula, including Singapore. The syntype series of both names contain more than one subspecies, so designation of a lectotype for both names will be necessary to fix the concept and use the names. This work will be published elsewhere. 25 de Nicéville and Martin (1895) stated that they had "great difficulty in identifying satisfactorily the common species of Elymnias of the undularis group occurring in Sumatra". They decided to follow Distant's (1882a) concept of "nigrescens" but still noticed that the Sumatran population of Elymnias hypermnestra (as nigrescens or protogenia) had smaller wings and duller coloration. Fruhstorfer (1907) noticed the opinion of de Nicéville & Martin, and decided to give the Sumatran population a status as a color form, and name it decolorata. However, since it was originally published as an infrasubspecific taxon, the name is not available under the Code unless another author uses the name to represent a valid taxon. In 1982, Aoki and colleagues enumerated the subspecies of Elymnias hypermnestra that occur throughout its range. They became the first authors to use decolorata to represent the Sumatran population. According to the Code (Article 45.5.1), the authorship of decolorata should be attributed to Aoki et al. (1982) because they made it available for use for the first time.

ssp. timorensis Fruhstorfer, 1907
Elymnias nigrescens timorensis Fruhstorfer, 1907 26 Although caudata was originally proposed as a species of its own, some authors (Gupta 2007;Wynter-Blyth 1957) treated it as a subspecies of hypermnestra (or undularis) due to the similarity in the Danausmimicking females. Our morphological and molecular studies demonstrate that caudata is a distinct, monophyletic taxon that is sister to hypermnestra. 27 Swinhoe's merula is based on a single male type collected from Sri Lanka. Having examined the type deposited in the Natural History Museum, London, we are convinced that merula should be a synonym of hypermnestra. Lamas (pers. comm.) suggests synonymizing merula with the Sri Lankan fraterna; however, the male of Sri Lankan fraterna is quite different from merula and we cannot at present conclude that synonymizing it with this subspecies is warranted. We presume that the single specimen of merula was accidentally introduced with imported palms or is an aberration; we treat this name as incertae sedis. 28 The true identity of Godart's leucocyma has been problematic since its description in 1819. Godart specified "Java" as the source of the specimen but gave a vague description without any figure. Doubleday (1844) suggested that northern India (near the border with Myanmar) might be the source of the specimen. Moore (1878a) mentioned the name leucocyma in his checklist without providing any further information. In 1882, Marshall & de Nicéville recognized the validity of leucocyma and synonymized malelas with it. However, Moore (1894) considered leucocyma to be the name that should be validated rather malelas. Fruhstorfer (1907), based on Godart's simple description, doubted that the origin of specimen was Java or northern India, and suggested placing leucocyma closer to hypermnestra.
There are more than 2 species with color patterns similar to leucocyma (viz. forewing with metallic blue sheen and hindwing with undulate margin) in Java and northern India, so we cannot specify the use of this name until more evidence becomes available. Moreover, Hewitson's figure (1861: pl. 9, fig. 34) of leucocyma was a misidentification of hewitsoni and has no relevance to this problem. 29 Our molecular phylogenetic analysis confirms that detanii and nepheronides represent opposite sexes of the same species as Araya (2016) (1889) placed parce as a subspecies of panthera, but morphological and molecular evidence suggest that parce does not belong to the panthera-group; it is more closely allied to harterti. 32 Fabricius stated that the type locality of the nominotypical panthera is Tranquebar (Tharangambadi, Tamil Nadu) in southern India, but the current distribution of this species in India seems to be restricted to the north. It is necessary to confirm the actual distribution of the species in India to verify whether the type locality falls in the actual distribution range or is simply a port from which the specimen was exported during the colonial period.  33 The subspecies dusara and dulcibella were both described from Java, suggesting the names might be synonymous. However, many lepidopteran species have distinctive populations western and eastern parts of the island (Aoki et al. 1982;Tsukada and Nishiyama 1982;Yata and Morishita 1985), and we therefore retain these two subspecies as valid. 34 Butler's lutescens was proposed as a full species based on specimens from Borneo (collected by Lowe), Malacca, Singapore, and Penang (from Roberts' collection), and according to Butler (1867: 404), lutescens was similar to dusara. This taxon, however, has been synonymized with the Indian nominotypical subspecies for long with no clear reason. In the present study, we revalidate the name and use it to represent the population in Borneo, as Borneo is the first locality mentioned in Butler's original description. The other conspecific taxa described from Borneo, such as labuana, lacrima, defasciata, alfredi, and pantherina are therefore newly synonymized with lutescens in the present study. It is not clear whether panthera from peninsular Malaysia is genetically distinct from Bornean populations, so we do not further revise the plethora of subspecific names associated with panthera.  38 We examined many nesaea from different regions of Borneo and found no consistent difference among them. Frustorfer's coelifrons is therefore treated as a junior synonym of hypereides. The relationship between the Malayan lioneli with hypereides can be addressed in future phylogenetic studies. 39 Lamas (pers. comm.) regards casiphone and kamara as different species, so erinyes, exclusa, and lombokiana are therefore regarded as subspecies of kamara. Our phylogenetic study based on morphology and DNA sequence data, however, show that kamara is conspecific with casiphone; kamara seems to represent a sexually dimorphic, mimetic form in which both males and females differ from the sexually dimorphic mimetic forms of casiphone. We therefore associate all subspecies previously included under kamara with casiphone. 40 Geyer, when describing casiphone, did not specify the geographical provenance of his specimen. Westwood (1851) and Fruhstorfer (1907) suspected that Java was possibly the origin of Geyer's casiphone. Having compared the original drawing of casiphone and the specimens from Java, Sumatra, Bali and Lombok, we conclude that the color pattern of the Javanese population matches well with Geyer's figure. 41 de Nicéville's erinyes was originally described as a full species, and then downgraded to be a subspecies of casiphone by Fruhstorfer (1907) or kamara (Corbet et al. 1992). Since we now consider this subspecies should be associated with casiphone, Fruhstorfer's combination should be revived. 42 Since we regard kamara as a junior synonym of casiphone, the name lombokiana, originally described as a subspecies of kamara, is treated as a new junior synonym of praetextata in the present study. 43 Martin's djilantik and de Nicéville's exclusa were previously placed under casiphone and kamara, respectively. We regard them as different color forms. 44 We examined many specimens from northeast India to Vietnam and found no consistent morphological differences to support the current subspecies classification. We therefore synonymize ivena and nilamba with malelas.
Elymnias malelas nilamba Fruhstorfer, 1911 45 Distant's saueri was originally described as a full species and then placed under casiphone as a subspecies by Fruhstorfer (1907) due to the similarity in wing pattern. Recently, Araya and Saito (2014) separated them into two morphologically defined species. Our phylogenetic analysis reveals that these taxa are not even sister to each other. The sister species of saueri is saola, and we therefore affirm the species status of saueri. 46 Semper (1887) described kochi as a full species, while Fruhstorfer downgraded it to be a subspecies of beza due to the similarity of their wing patterns. Treadaway and Schroeder (2012) considered kochi a full species, and our phylogenetic study places kochi as the sister group of kanekoi from Negros, so its status as a full species is upheld.

ssp. amoena Tsukada & Nishiyama, 1979
Elymnias amoena Tsukada & Nishiyama, 1979 Fig. 14N Tijdschr. Ent. 4 (5/6): 159, pl. 8, fig. 3. Elymnias vitellia f. basium Fruhstorfer, 1907. unavailable  The species agondas has long been considered a highly variable species, the females of which mimic various Taenaris spp. throughout its range in New Guinea and the surrounding islands. Our phylogenetic analysis, however, reveals that thryallis, which was currently placed as a subspecies of cybele, is nested within agondas, and the branch support for this relationship is strong. We therefore conclude that both agondas and cybele are not monophyletic species as currently circumscribed. We sink thryallis into agondas, thus forming a monophyletic species, but further clarification of the subspecific nomenclature is difficult because of the myriad described taxa, vague descriptions of many type localities (frequently "New Guinea"), and our lack of access to material of several rare "subspecies" of agondas, cybele, and other members of the species group. 51 The true identity of agondas is mysterious. Boisduval stated that the source of the type specimen was the Solomon Islands (Boisduval 1832: 138, pl. 3, fig. 5). However, no Elymnias are known from the Solomon Islands at present (Tennent 2002), and the easternmost point in the range of this species is Woodlark Island in Milne Bay Province, Papua New Guinea-approximately 350 km from the Solomon Islands across the Solomon Sea. Although the circumscription of "Solomon Islands" has changed throughout history, an area with that name has never included Woodlark Island. Bougainville Island can be included in the Solomon Islands, but Elymnias agondas has never been found there. The original drawing of agondas is a male, and since the highly variable wing pattern of male agondas is not a reliable diagnostic character, we tentatively retain the nominotypical agondas as a taxon with questionable geographical provenance, but the nominal subspecies should not be applied to any population until the true collection locality of the type can be discerned. Fruhstorfer, 1904 According to our phylogenetic analysis, cumaea is not a monophyletic group and is part of a radiation on Sulawesi that includes hicetas and hewitsoni. The subspecies phrikonis is not allied with other cumaea subspecies, but comprises the sister group to a clade including cybele, vitellia, holofernes and agondas. We therefore elevate phrikonis to the species level. Fruhstorfer (1907) described relicina from Sanana (Sula Besi), and we consider this name should be synonymized with phrikonis as they are not different from each other morphologically. 55 Nominotypical cumaea is only reported from Halmahera. We have not been able to obtain specimens for inclusion in the molecular phylogeny so we have no indication of how many subspecies previously associated with cumaea should be retained. 56 Having examined specimens of thyone, we conclude that this taxon should not be regarded as a junior synonym of cumaea cumaea, but we tentatively place thyone as a subspecies of cumaea pending molecular data for inclusion in a phylogenetic study. 57 In our phylogenetic study, a specimen from North Sulawesi identified as toliana is sister to hicetas, and this pair is sister to bornemanni, which we regard as a full species. Since we have no genetic data from nominotypical cumaea, we are unsure were to place this taxon and tentatively retain toliana as a subspecies of cumaea, even though this arrangement makes cumaea polyphyletic. 58 Having examined the type specimens of bonthainensis and hicetas, we failed to find any distinguishing morphological characters. These two names are therefore synonymized. 59 Martin described rarior as a subspecies of hicetas without a clear indication of its type locality in Sulawesi. In our phylogenetic study, two specimens identified as rarior are paraphyletic with regard to hicetas and butona. Since the type locality of rarior is unclear, we retain rarior as a subspecies of hicetas. Elymnias cumaea resplendens Martin, 1929. TL: Celebes (Indonesia: Sulawesi). TS: NHMT. Mitt. münchn. ent. Ges. 19: 162. 60 The situation of bornemanni is similar to that of phrikonis. It is not closely related to other cumaea subspecies, so we treat it as a full species. 61 phrikonis has been regarded as a subspecies of cumaea. Our phylogenetic analysis, however, places it as the sister to a clade consisting of cybele, holofernes, umbratilis, vitellia, and agondas. We therefore elevate phrikonis to full species status. Since relicina was described from the same locality as phrikonis we synonymize these two names in the present study. 62 Elymnias cybele umbratilis was originally described based on five syntypes from Biak and synonymized with thryallis by subsequent authors. Our phylogenetic analysis, however, reveals that umbratilis is a distinct taxon, not closely related to any subspecies of cybele, and sister to holofernes. We therefore revive this taxon and give it full species status. 63 Martin (1929) placed resplendens with cumaea, but our analysis suggests that resplendens is closely related to hewitsoni and deserves full species status.

Discussion
Wing patterns of Elymnias butterflies appear to be highly evolvable, which facilitates Batesian mimetic resemblance to a variety of phenotypically dissimilar model species.
Many Elymnias are found on islands in the Indo-Australian Archipelago, and the isolation provided by islands seems to provide the opportunity for divergence and local adaptation, facilitating resemblance to different model species in different locales. The remarkable capacity for phenotypic evolution of wing patterns has resulted in sexually dimorphic mimicry, convergence of distantly related taxa on similar wing patterns, and marked phenotypic divergence among conspecific popuations. These phenomena have previously confounded attempts to produce an accurate taxonomic framework because few if any morphological characters are taxonomically or phylogenetically informative. Wing veination, male genitalia, and female genitalia are remarkably uniform among species of Elymnias; only slight variation in male genitalia might be useful for discriminating some species. Species delimitation and diagnosis in Elymnias has therefore traditionally relied almost entirely on wing patterns. Our molecular phylogeny, which uses genetic markers presumed to be unrelated to wing phenotypes, has detected multiple instances of similar wing patterns in non-sister Elymnias lineages that mimic the same, widespread model species. This similarity seems to be the result of convergent evolution, and we have therefore split these taxa into two or more monophyletic lineages (e.g., E. cumaea and E. cybele have each been split into four and three different species, respectively). On the other hand, some Elymnias species-like other mimetic butterfly taxa (Kunte et al. 2014;Merrill et al. 2015;Thompson and Timmermans 2014)-are polymorphic, with single species expressing different mimetic phenotypes in allopatric populations where they mimic different models. We have identified several instances of one nominal species nested within another, and synonymize these taxa under a single species name (e.g., E. cottonis into E. hypermnestra and E. cybele thryallis into E. agondas). Strong dimorphism caused many early workers to describe males and females as separate species, most of which have been synonymized. In this paper we confirmed Araya's (2016) conclusion of synonymizing E. detanii, known only from males, into E. nepheronides, known only from females; this rare species is known only from the Indonesian islands of Flores and Sumbawa. Similarly, E. vasudeva oberthurii has been sunk into E. esaca andersonii, as these apparently represent different sexes of the same species.
Females of several Elymias species, including E. agondas, E. hypermnestra, and E. esaca, are morphologically variable across their range. Rather than recognize every wing pattern variant as a different subspecies, we have synonymized many subspecies into geographically cohesive taxa, for example, within the islands of Borneo or New Guinea.
Much of the mismatch between Elymnias' previous taxonomic framework and its evolutionary history is due to rapid evolutionary change. This resulted in morphologically-delimited nominal species that were polyphyletic. In these cases, our molecular phylogenetic results make delimiting species relatively straightforward. However, there are several cases that are not as clear-cut. For example, we elected to retain E. esaca and E. vasudeva as distinct species despite their paraphyletic relationship because of marked, species-specific morphological differences in these two parapatrically distributed taxa. Population genetic theory predicts incomplete lineage sorting of genetic loci to persist for some time after speciation, resulting in paraphyletic species; the probability of reciprocal monophyly increases with time since divergence (Avise and Ball 1990). Thus, requiring all species to be monophyletic would underestimate true species diversity (Hickerson et al. 2006), particularly in recently diverged species (Knowles and Carstens 2007) such as esaca and vasudeva. However, we decided to sink E. kamara into E. casiphone despite their morphological differences because both taxa are wholly sympatric and because morphologically intermediate specimens are known. We included four specimens of E. c. casiphone and four of E. "kamara" from Java, Bali, and Lombok in our molecular phylogeny, and the topology of all genetic loci individually and together clearly indicated these taxa were conspecific. We suspect that a genetic switch is responsible for the distinct pair of E. casiphone casiphone male and female phenotypes (which mimic Euploea mulciber males and females) and the different, sexually dimorphic forms of E. casiphone kamara, which mimic other Euploea species.
Our molecular phylogeny identifies several examples of allopatrically or parapatrically distributed populations that form distinct, monophyletic sister groups: E. sansoni sansoni on Negros and E. sansoni aklanensis on Panay; E. patna from India and E. patna from peninsular Malaysia; E. vitellia vitellia from Seram and E. vitellia viminalis from Buru; and E. hypermnestra from Java and the Lesser Sundas and E. hypermnestra from everywhere else. These monophyletic sister lineages would likely be considered different species under a strict phylogenetic species concept, and, in most cases, preliminary Bayesian species delimitation analyses with the program Bayesian Phylogenetics and Phylogeography (BPP; Yang and Rannala 2010) suggest the sister lineages are different species. However, we refrain from splitting these species because we regard the geographic sampling of our phylogenetic work as too sparse, consider the degree of phylogenetic distance between the lineages to be too small, or otherwise fail to find convincing evidence that reciprocal monophyly is the result of anything more than geographical isolation. In addition, a recent simulation study suggests that programs such as BPP delimit population structure, not species (Sukumaran and Knowles 2017). Further work may find convincing evidence to split one or more of these pairs into two species.
Although there is one African and several mainland Asian species, most of Elymnias' diversity is found on the islands of the Indo-Australian Archipelago. Islands are considered laboratories for the study of evolution because they promote isolation and divergence while simplifying the task of delimitating populations and other taxa. Evolutionary study of this taxon provides an excellent opportunity to study the role of archipelagoes in diversification, and the evolutionary genetics of evolutionary novelty and speciation.