A revision of the ant genus Mystrium in the Malagasy region with description of six new species and remarks on Amblyopone and Stigmatomma (Hymenoptera, Formicidae, Amblyoponinae)

Abstract The genus Mystrium is revised for the Malagasy region. Six species, Mystrium barrybressleri sp. n., Mystrium labyrinth sp. n., Mystrium eques sp. n., Mystrium mirror sp. n., Mystrium shadow sp. n., and Mystrium janovitzi sp. n. are described as new. Two existing names, Mystrium fallax Forel and Mystrium stadelmanni Forel, are synonymized with Mystrium voeltzkowi Forel and Mystrium mysticum Roger, respectively. All recognized species, including species outside of the Malagasy region, are assigned to one of the three newly proposed species groups. The associations between existing names and males are reexamined, and males of eight of the ten Malagasy species are described or redescribed. The taxonomic history of Mystrium highlights the importance of using unique identifiers when designating type specimens and the use of deposited vouchers in phylogenetic and ecological studies. Keys to species for workers, queens, and males are provided. Furthermore, a neotype for Mystrium mysticum is designated, as well as lectotypes for Mystrium camillae Emery, Mystrium rogeri Forel, Mystrium fallax Forel, Mystrium oberthueri Forel, Mystrium stadelmanni Forel, and Mystrium voeltzkowi Forel. Stigmatomma gingivale (Brown) is reassigned to Amblyopone as comb. rev. and Amblyopone awa Xu & Chu, Amblyopone kangba Xu & Chu, Amblyopone meiliana Xu & Chu, and Amblyopone zomae Xu & Chu are transferred to the genus Stigmatomma as comb. n.


Introduction
The genus Mystrium Roger, 1862 was originally described from Madagascar, and ten species names were regarded as valid prior to this revision (Fig. 1). The Malagasy region has an incomparably abundant and rich Mystrium fauna with over half of the species endemic to the Malagasy region. After Roger established Mystrium with a single species, Mystrium mysticum Roger, 1862, Forel described five additional species in the Malagasy region between 1895 and 1899 (Forel 1895;1897;1899). The first record of Mystrium outside the Malagasy region was M. camillae Emery, 1889 from Myanmar, and the second was M. silvestrii Santschi, 1914from Cameroon. Later Menozzi (1929 revised the genus Mystrium, including all six Malagasy species, M. camillae, and M. silvestrii, and provided a key for workers to all of these species. This revision (Menozzi 1929) was the last taxonomic work on the Malagasy Mystrium. A subspecies was later established under Mystrium camillae, Mystrium camillae javana Karavaiev, 1925; however, this subspecies was synonymized with Mystrium camillae by Brown (1960). Outside the Malagasy region, Mystrium oculatum Xu, 1998, was described from China, and M. leonie Bihn & Verhaagh, 2007and M. maren Bihn & Verhaagh, 2007 from West Papua, Indonesia. Mystrium oculatum was synonymized with M. camillae by Bihn and Verhaagh (2007).
Recent research in Madagascar demonstrated that the previous taxonomic works were insufficient to diagnose the species boundaries of Mystrium in the Malagasy region. Specimens collected from a single colony showed remarkable morphological variations within each existing species, and these variations rendered most published species-level diagnostic characters insufficient to delimit species. A reassessment of species boundaries was also needed in light of recent ecological and phylogenetic studies. Ecological studies reported various reproductive systems across Mystrium species, including typical winged queens, short-winged queens (Molet et al. 2009), small ergatoid queens (Bouchet et al. 2013;Molet et al. 2007a) (Fig. 1B) and intercaste anomalies intermediate between queen and worker (Molet et al. 2009;Molet et al. 2012). Ergatoid queens, short-winged queens and the intercastes were not described by Menozzi (1929). The genus Mystrium was included as an evolutionarily interesting lineage in most recent molecular phylogenetic papers on ants (Brady et al. 2006;Kück et al. 2011;Moreau and Bell 2013;Moreau et al. 2006;Ouellette et al. 2006;Saux et al. 2004), but no functional key was available for establishing identification of the ecological and molecular vouchers. The existing taxonomic system of Mystrium was inadequate to incorporate new scientific knowledge into existing species limits. The taxonomic system of the males was in even worse condition. Of the previous male descriptions for four of the ten valid species, only a male of Mystrium voeltzkowi Forel, 1897 has been precisely associated with conspecific workers. The other male descriptions, for M. mysticum, M. rogeri, and M. oberthueri, were vague, insufficient, and as we conclude, wrongly associated.
In this study, we revise the species-level taxonomy for the genus Mystrium in the Malagasy region. Comprehensive morphological and distributional examinations using recently accumulated material enabled us to recognize ten species in this region; six species are new, and two existing names, Mystrium fallax and M. stadelmanni, are synonymized with M. voeltzkowi and M. mysticum, respectively. We propose a new species-group system, and all recognized species, including extralimital species, are assigned to one of the three species groups. Both workers and queens of existing Malagasy species are redescribed with new diagnostic characters. Associations between existing names and males are reexamined, and the males of eight of the ten Malagasy species of Mystrium are described or redescribed. Identification keys to species for workers, queens, and males are provided. Furthermore, a neotype for M. mysticum is designated, as well as lectotypes for M. camillae, M. rogeri, M. fallax, M. oberthueri, M. stadelmanni, and M. voeltzkowi. Though we do not include species outside of the Malagasy region in this revision, we provide taxonomic remarks on M. camillae to better define species in the camillae group from the Malagasy region.
In addition to the species-level taxonomy of Mystrium, we also propose new genuslevel diagnostic characters of Mystrium. In our previous work (Yoshimura and Fisher 2012), we proposed that the elongated mandible in the genus Amblyopone Erichson, 1842 evolved separately from that in the XMAS (Xymmer Santschi, 1914+ Mystrium + Adetomyrma Ward, 1994+ Stigmatomma Roger, 1859 clade; however, we did not propose any diagnostic characters to distinguish between Amblyopone and XMAS clade genera. In this study, based on the comparison between Mystrium, the other genera in the XMAS clade (Yoshimura and Fisher 2012), and Amblyopone, we consider the presence of the dentiform setae on the labrum as a diagnostic character in workers to distinguish the genus Amblyopone from genera in the XMAS clade.   used through the manuscript in Snodgrass (1957) to indicate the distal secondary division of the paramere, it is the more appropriate term. In the same work, Snodgrass used the term harpago only once and did not explain the origin of the term. In this study, we use simplified names for wing veins. This simplification is not in conflict with the precise system used in our previous studies (Yoshimura and Fisher 2012). The simplified vein names are shown in Figure 3I, and the precise names were given in parentheses on figure 10c in Yoshimura and Fisher (2012). Terminology for caste names. We classify females into three phenotypes: winged gyne, large wingless phenotype, and small phenotype with or without vestigial wings. Some species lack the winged gyne, and small phenotype acts as a reproductive cast; some species have intermediates between winged gyne and large phenotype. The presence or absence of each caste, and the role of each phenotype are consistent within species groups defined below, but vary among the species groups (Table 1). In this study, we use "queen" in reference only to the winged gyne; intercaste for intermediates between winged gyne and large wingless phenotype; major worker for the large wingless phenotype and minor worker for the small phenotype when the small phenotype does not function as a reproductive; worker for the large wingless phenotype and ergatoid queen for small phenotype when the small phenotype functions as a reproductive. The term female is used as a collective noun to indicate all female phenotypes. Males often display remarkable intra-specific variation in size and shape; however, no clear morphological division among those variations is recognized. Relationships between caste and species groups are given in Table 1.
Terminology for geographical regions. Terminology and definition for geographical regions follows Fisher (2010).  (Fig. 4A). 5. Labrum lacking small dentiform setae arranged horizontally. 6. Palpal formula 4,3. 7. Distinct extension present on the distal edge of second labial palpomere (Fig. 4B). 8. Mandible linear without a distinct basal angle. 9. Masticatory margin of mandible with two rows of projections: true mandibular teeth on dorsal row and a series of basal denticles on the ventral row; basal ventral denticles larger than true mandibular teeth except for apical teeth. 10. At midlength of mandible, row of basal denticles arranged on ventral edge of mandibular shaft, distant from mandibular teeth (Fig. 4C).
11. Mandible twisted inward so that apical tooth is located ventrally (Fig. 4D). 12. No basal large projection present on basal portion of mandibular inner margin. 13. Ridge present on apical portion of mandible, which is originally dividing ventral and inner surfaces of mandible, inserting dorsally to the apical tooth (Fig. 5A vs. 5B: Stigmatomma mystriops Brown, 1960, see comments).
16. One or two stout spines present on posterior portion of abdominal sternum VII (Fig. 5C).
17. Body surface with spatulate to squamose setae in some workers of all species. Diagnosis of male. The diagnostic characters uniquely observed in Mystrium within the subfamily Amblyoponinae are given in italics.
11. Abdominal segment IV with tergosternal fusion. 12. Pretergite of abdominal segment IV distinctly differentiated from posttergite, with the cinctus between them.
18. Serrate denticles present on basal portion of ventral margin of aedeagus in lateral view.
24. Position of cu-a on forewing variable, close to or far from junction between media and cubitus.
25. Radius present on hindwing. 26. 1rs-m present on hindwing. 27. Media on hindwing present apical to 1rs-m. Comments on generic diagnosis. The workers of the genus Mystrium can be distinguished easily from those of the other amblyoponine genera by their characteristic head shape and mandibles (Fig. 1). Mystrium has a wide head with posterior margin strongly concave, mandible linear and longer than head with rounded apex in full-face view, and two separated rows of "teeth" on the masticatory margin. In this study, we propose further detailed characters as the result of our comparative study. All Mystrium females share six characters unique to this genus which distinguish it from the other amblyoponine genera: spatulate setae on clypeus mesal of the mandible insertion (character 4: Fig. 4A), a distinct extension on the distal edge of the second segment on the labial palp (character 6: Fig. 4B), mandibular teeth distant from basal denticles (character 10: Fig. 4C), twisted mandible (character 11: Fig. 4D), a ridge on the apical portion of the mandible inserting dorsally to the apical tooth (character 13: 5A), and one or two lateral spines on abdominal sternum VII (character 16: 5C). Character 10, mandible teeth and basal denticles arranged in two rows (Fig. 4C), was used in Bolton (1994) to separate Mystrium from Amblyopone and Stigmatomma, both of which were in Amblyopone at that time. For character 11, Keller (2011: character 29) described the mandibular type as "torqued." Although all amblyoponine genera share this character state, the mandible in Mystrium is unique: in females the mandible is further twisted so that the apical tooth of queens and minors is directed ventrally (Fig. 4D). For character 13, the ridge (= hump in Gronenberg et al. 1998) on the mandible functions as a pivot when the mandibles snap (Gronenberg et al. 1998). We confirmed that all females have this ridge, although it is uncertain if all females can snap their mandibles regardless of the shape of the apical tooth (see description for each species group). For character 16, we have confirmed that some species of Stigmatomma have three to nine lateral spines on abdominal sternum VII, while neither Adetomyrma nor Xymmer have this spine. Brown (1960) described Stigmatomma mystriops (Brown, 1960) as a species that has specialized mandibular characters similar to Mystrium, e.g. teeth arranged in two separated rows, the presence of a ventral ridge on the subapical teeth, and a small apical tooth; however, these characters differ distinctly between S. mystriops and Mystrium species. The mandibular teeth and basal denticles are consistently separated and these two rows are almost parallel, except for the apical portion of the mandible in Mystrium, while the separation exists only on the basal portion of the mandible in S. mystriops. The ridge in Mystrium is the extension of a carina, which usually divides the lateral and ventral surfaces of the mandible, and the extension of the carina in Mystrium is inserted dorsally to the apical tooth (Fig. 5A). However, the ridge in S. mystriops is developed independently from another carina along the ventral one mentioned above, which in S. mystriops continues to the ventral margin of the apical tooth the same as a usual mandible of ants (Fig. 5B). The apical tooth in S. mystriops is directed mesally, not ventrally as in Mystrium. In addition to the Mystrium-like characters (Brown 1960), we found two pairs of long setae on the anterior clypeal margin in S. mystriops. These strange characters observed in S. mystriops may be part of a specialized snapping mandible system.
In addition to characters unique to Mystrium, we discuss two generic characters (characters 5 and 9) that are useful for distinguishing amblyoponine genera. Neither character 5 nor 9 distinguish Mystrium from the other XMAS genera; however, both of these characters do distinguish Mystrium from Amblyopone.
Character 5 is provisionally proposed here to separate Amblyopone from genera in the XMAS clade. In the XMAS clade, the basal part of the labrum is lacking small dentiform setae (Fig. 6A). However, in Amblyopone, small dentiform setae are present on the labrum (Fig. 6B). Mandible character 9 proposed in Yoshimura and Fisher (2012) is unique to the XMAS clade. In our previous work, we described the unique evolution of the mandible in the XMAS clade. We presented as evidence the two-layered "teeth" on the masticatory surface of the mandible, which consists of true mandibular teeth on the dorsal row and apically extended basal denticles on the ventral row (Fig. 6E). We now update the mandibular character description because we have since observed in a few species of Amblyopone two-layered mandible dentition, such as in Amblyopone au02 (QMT152688). Even though two-layered mandible dentition was found in Amblyopone, as we described in character 9, the principal true mandibular teeth on the A dentiform labral seta is absent B dentiform labral setae are present C, E the masticatory margin is the main functional part of the mandible in XMAS clade species D, F the whole mandibular shaft is the functional part of the mandible in Amblyopone. C, D Photos by Alex Wild. dorsal row are developed (Fig. 6F) larger than the basal denticles on the ventral row in Amblyopone, while the basal denticles are usually larger than true mandibular teeth on the masticatory margin with the exception of apical teeth in the XMAS clade (Fig. 6E).
The differences in mandibular characters between species in the XMAS clade and those in Amblyopone are associated with the parts of the mandible used for catching their prey. We assume that the masticatory margin is the main functional part of the mandible in XMAS clade species (Figs 6C, 6E), while the whole mandibular shaft is used for this purpose in Amblyopone (Figs 6D, 6F). The presence or absence of dentiform setae on the labrum (character 5: Figs 6B vs. 6A) supports our assumption. Amblyopone species probably "hold" prey using teeth along the whole shaft of the mandible (Fig. 6F), conical setae on the flat clypeus, and dentiform setae on the labrum (Fig. 6B); while species of Stigmatomma (in XMAS clade) "pinch" prey using their masticatory margin (distal part) of the mandible (Fig. 6E), and support it using anterior clypeal conical setae strongly extended anteriorly (Fig. 6A). When the mandibles of both groups elongated under these two use conditions, their morphology adapted differently. The mandibular shaft in Amblyopone became wider and stouter (Fig. 6F), while the basal denticles in XMAS clade became larger than true teeth and distinctly directed basally (Fig. 6E). Brown (1960) was the first to propose that the dentiform setae on the labrum functioned as a structure to grip active prey. He attributed this character to Amblyoponini but our review suggest that dentiform setae on the basal part of the labrum (as in Fig.  6B) is restricted to three genera: Amblyopone, Onychomyrmex and Apomyrma. Ward (1994) reported that conical setae are present on the labrum in Apomyrma stygia Brown, Gotwald & Lévieux, 1971. Saux et al. (2004 described the presence of these setae in Amblyopone australis Erichson, 1842, A. hackeri Wheeler, 1927, A. longidens Forel, 1910, and Stigmatomma gingivale (Brown, 1960. Keller (2011) reported the presence of these setae in Amblyopone australis, A. mercovichi Brown, 1960, Onychomyrmex doddi Wheeler, 1916, and Apomyrma stygia, but their absence in Concoctio concenta Brown, 1974, andMyopopone castanea (Smith, 1860). Ward (1994) showed in his figure 9 labral setae in Onychomyrmex which is currently identified as Onychomyrmex au01 (PSW10030, CASENT0172779) though in Saux et al. (2004), this taxon is referred to as an undescribed species of Amblyopone. In addition to the above species, we found the presence of dentiform setae on the labrum in Amblyopone michaelseni Forel (CASENT0100441), Amblyopone au01 (CASENT0434465), and Amblyopone au02 (QMT152688). In Apomyrma stygia (CASENT0000077), the dentiform setae are present and distributed not only along the basal portion but also along the whole surface of the labrum; in addition, the conical setae on the anterior clypeal margin are absent. We also confirm the absence of dentiform setae on the labrum in Prionopelta, Concoctio and Bannapone.
Because of the presence of the dentiform labral setae, Stigmatomma gingivale is retransferred to Amblyopone as A. gingivalis Brown, 1960 comb. rev., although Yoshimura and transferred this species from Amblyopone to Stigmatomma. The dentiform labral setae in A. gingivalis are more abundant than those of A. australis, and covering on the distinct basal convexity on the labrum. In addition to the presence of the dentiform labral setae, true mandibular teeth larger than the basal ventral denticles are also present in A. gingivalis and provide an additional supportive character for this generic transfer.
The male diagnostic characters used in this revision follow those in Yoshimura and Fisher (2012). All worker diagnostic characters are applicable to both alate and ergatoid queens.

Species groups in Mystrium and list of species
In this study, we divide the species of Mystrium into the three species groups listed below. The subdivisions of the genus are useful because of the difficulty of specieslevel identification.
camillae species group: Alate queen present (Table 1). Dorsal surface of pronotum covered with regular reticulation in worker. Second maxillary palpomere shorter than the third in the Malagasy species. Labrum in females gently convex but without median longitudinal carina. Two lateral spines present on abdominal sternum VII in females. Whole vertex sculptured and not smooth in worker. Petiole in dorsal view in females relatively wide (PtI>210 in queen; PtI>180 in worker). Known from the Afrotropical, Malagasy, Indomalaya, and Australian regions. mysticum species group: Alate queen present (Table 1). Dorsal surface of pronotum covered with somewhat irregular striae, and central part of pronotal dorsum covered with fine striae in worker. Second maxillary palpomere longer than the third. Labrum in females with longitudinal carina and strongly convex on its central portion. Single lateral spine present on abdominal sternum VII in females. Ventral area of vertex in females not sculptured, smooth. Petiole in dorsal view in females relatively long (PtI<189 in queen; PtI<183 in worker). Known only from the Malagasy region.
voeltzkowi species group: Only ergatoid queen, which is relatively smaller than worker, present; alate queen absent (Table 1). Dorsal surface of pronotum covered with somewhat irregular longitudinal striae in worker. Second maxillary palpomere longer than the third. Labrum in females gently convex but without median longitudinal carina. Two lateral spines present on abdominal sternum VII in females. Whole vertex in females sculptured and not smooth. Petiole in dorsal view relatively long (PtI<195 in queen; PtI<180 in worker). Known only from the Malagasy region.
A list of Mystrium species is given below. Sexes and castes known in each species are given in brackets as: w, worker; aq, alate queen; eq, ergatoid queen; m, male. In addition to the species listed below, we recognize two undetermined morphospecies in the collection and provide notes for these taxa in the remarks of the camillae species group. The associations between the species names and the species codes used in previous papers are also discussed in the remarks for each species.

Males
Males of labyrinth and eques are unknown.
1 Lateral ocellus relatively large and close to eye; distance between lateral ocellus and eye less than 1.2× maximum diameter of lateral ocellus (Fig. 26A).
Eye large, occupying about 0.75× of head length (Fig. 26C    Basoventral expansion of aedeagus long, its distal portion relatively narrow and sharp (Fig. 29A)

Review of species
Due to the remarkable intra-specific morphological variation in each species, most specific characters are distinct only when individuals belonging to the same caste and in the same body size range are compared. The definition of each caste is given in each speciesgroup description. Characters unique to each species group in the Malagasy region are given in italics.

camillae species group
The camillae species group has two species in the Malagasy region, Mystrium labyrinth sp. n. and Mystrium barrybressleri sp. n., in addition to the species known from outside of the Africa. In addition to the four described extralimital species, we recognize two more undescribed morphospecies, one from Indonesia (Mystrium id01: CASENT0102169) and one from Mozambique (Mystrium mz01: CASENT0318957) in the species group. In this study we include minimal discussion of the species outside the Malagasy region because the material currently available is insufficient for an in-depth analysis. The camillae species group is based on the distinctly reticulate head and pronotal dorsum (Fig. 13A).
Although several exceptions have been observed outside of the Malagasy region, the second maxillary palpomere, which is shorter than the third, differentiates this species group from the other Mystrium species in the Malagasy region (Fig. 13C). All known queens have wings, developed eyes and complete ocelli as is usual among ant queens. Rarely, the wings are shrunken and incomplete, as observed in an individual of M. camillae collected in Indonesia (CASENT0009854), but the wing sclerites, e.g. the division of mesoscutum, mesoscutellum, and metanotum, are still fully developed (as in Fig. 7A). Two phenotypes were observed in conspecific workers of Mystrium barrybressleri, M. camillae, and Mystrium mz01 from Mozambique (Table 1): the mandible is stouter and the apical tooth is feebly sclerotized and curved inward in one phenotype (as in Fig. 15A), while the mandible is narrower and the apical tooth is moderately sclerotized and directed ventrally in the other (as in Fig. 15B). The differences between these two worker phenotypes in terms of mandible shape, body size, setae, and body color, are similar to those between workers and ergatoid queens in the voeltzkowi species group (see the voeltzkowi group description below for detail); however, these differences are relatively less pronounced. The small workers are also lack the wing buds, functional spermatheca, and developed ovaries, which were found in ergatoid queens in the voeltzkowi species group (Molet et al. 2009). The function of the two phenotypes in the camillae species group has not yet been clarified. All known queens in this group are alate, at least with complete wing sclerites. Hence in the present study, we regard these two phenotypes as a variation of the worker caste. We applied the term major worker to the phenotype with the curved apical tooth, and minor worker to the phenotype with the straight apical tooth.
Although Mystrium camillae is not found in the Malagasy region, we designate the lectotype for M. camillae and give some remarks below to discuss the diagnostic characters of M. barrybressleri more clearly. Posterolateral corner of head strongly to moderately expanding posteriorly. Posterior face of vertex forming slightly blunt angle with dorsal face on median line of head, so that declivity of vertex on lateral part slightly steeper than on median part. Ventral half of vertex sculptured. Eye small, only with several ommatidia. Anterior margin of clypeus convex with long, conical setae, of which median pair larger than adjacent pair. Genal tooth of head moderately developed, reaching about half of lateral lobe of clypeus. Masticatory margin of mandible slightly visible in full-face view, and dorsal surface on distal portion of mandible slightly wider than that on mandibular shaft. Second maxillary palpomere shorter than third. First flagellomere (third antennal segment) about 0.5× length of pedicel (second antennal segment). Central part of pronotal dorsum strongly to weakly, and regularly reticulate. Lateral surface of pronotum strongly and regularly reticulate on its posterior portion. Mesonotum distinct in dorsal view, its length slightly shorter than that of propodeum. Metanotal groove shallowly impressed and mesonotum higher than pronotum in lateral view. No carina present on dorsal edge of metapleural gland bulla. Petiole extremely wide in dorsal view (PtI>195).

Mystrium barrybressleri
Body color yellowish brown to reddish brown. Wings present and well developed. Wing sclerites fully developed even if wings have been removed. Posterolateral corner of head moderately expanding posteriorly; expansion weaker than that of workers. Posterior surface of vertex forming slightly blunt angle with its dorsal surface on median line of head, so that declivity of vertex on lateral part slightly steeper than on median part. Ventral half of vertex weakly sculptured, almost smooth. Eye well developed. Both anterior and lateral ocelli clearly present, median portion of lateral ocelli and posterior portion of anterior ocellus edged by blackish pigment. Anterior margin of clypeus convex with long conical setae, of which median pair larger than adjacent pair. Genal tooth of head moderately developed, reaching about half of lateral lobe of clypeus. Masticatory margin of mandible visible in full-face view, and width of dorsal surface on distal portion slightly wider than that on mandibular shaft. Spoon-shaped seta present on basal side of each basal denticle on masticatory margin of mandible. First flagellar segment on antenna short,  Eye moderately large, occupying 0.6× head length. Ocelli protruding from dorsal margin of head in full-face view. Dorsal margin of head in full-face view rounded. Both anterior and lateral ocelli small. Lateral ocellus small and distant from eye: distance between these more than 2× maximum diameter of lateral ocellus. Posterior half of vertex clearly differentiated from its dorsal half, its dorsal face distinctly shorter than its posterior face. Palpal formula 4,3. First segment of maxillary palp flattened and distinctly wider than second segment. Second maxillary palpomere shorter than third. Notauli weakly impressed on mesoscutum, but often unclear or absent. Petiole in dorsal view thin, its length 0.5× that of abdominal tergite III. Petiolar dorsum covered with fine punctures. Abdominal tergum VIII without deep punctures, almost smooth.
Distal portion of abdominal sternum IX smooth and not punctured. Basal ring quite short, not extending basally. Telomere slightly extending distally more than digitus. Bas- oventral expansion of aedeagus well developed basoventrally, distinctly longer than dorsal extension. Ventral margin of aedeagus gently curved ventrally in lateral view. Aedeagus moderately narrowing distally on its distal half and its distal portion narrowly rounded.
On forewing, cu-a usually located at junction of Media (M) and Cubitus (Cu), sometimes slightly basal from junction.
Body color reddish brown to blackish brown.      Etymology. The specific epithet is the patronym of Dr. Barry Lee Bressler, retired physicist, former adjunct professor of physics at Virginia Polytechnic Institute and State University, and amateur naturalist, in recognition of his interest in myrmecology and his support for research on ants.
Distribution. MADAGASCAR: as in Figure 42A. Additional material examined. In addition to the type material, specimens from the following localities were examined in this study: MADAGASCAR. Remarks. The workers of Mystrium barrybressleri can be distinguished from other Mystrium workers and queens in the Malagasy region by a combination of the following characters: mesosoma lacking wing-sclerites; first flagellomere (third segment of antenna) shorter than pedicel (second segment of antenna) (as in Fig. 10C); and second maxillary palpomere shorter than third (Fig. 13C). The wider petiole in dorsal view (Figs 35A, B) separates M. barrybressleri workers from the other Mystrium workers in the Malagasy region. A combination of smaller body size, small lateral ocelli distant from the eye (as in Fig. 26B), rounded posterior margin of head with protruding lateral ocelli in full-face view (Fig. 31A), cu-a on the forewing located at junction of Media (M) and Cubitus (Cu) (Fig. 31C), wide and short petiole in dorsal view, and fine punctation on abdominal tergite III separates M. barrybressleri males from other known Mystrium males in the Malagasy region.
The workers of Mystrium barrybressleri are most similar to those of Mystrium camillae Emery. Mystrium camillae appears to be a complex of several species (see discussion in M. camillae) but we did not have sufficient material to fully evaluate the castes and worker variation throughout the range of the species complex. Mystrium barrybressleri workers can be distinguished by the shape of the propodeum. In workers, the lateral portion of the propodeum is steeply and strongly convex in M. "camillae" (Fig. 33D), while the convexity is gentle and weak in M. barrybressleri (Fig. 33C). Also, in M. barrybressleri, minor workers have setae much narrower and simpler than those of major workers ( Fig. 35A: major vs. 35B: minor), while in M. "camillae" minor and major workers have similar setae.
There are additional differences with M. camillae (sensu stricto, as defined by the syntype series from Myanmar). The pair of spatulate setae on the anteromedial portion of clypeus is distinctly longer in M. barrybressleri (Fig. 34C) than those in the syntype series of M. camillae (Fig. 34D). The queen of M. barrybressleri can be distinguished by the presence of simple setae on the pronotum, while in M. camillae sensu stricto, the setae are clavate. Additional diagnostic characters can be presented once the species limits are further refined for M. camillae and related species.
Mystrium barrybressleri is found from the rainforests and montane rainforests, and the distribution pattern of this species is similar to that of a forest species Tetramorium andrei Forel (Hita Garcia and Fisher 2012: fig. 141). No remarkable difference was observed between specimens from NW Ambaravaranala (-18.26667°, 45.40667°) and those from the other localities, even the former forest is currently isolated from the others.  (Emery 1889), the queen is dealate, and was not collected from the same colony as any of the syntype workers.
According to the revision of Mystrium in the Indo-Australian region (Bihn and Verhaagh 2007), Mystrium camillae is widely distributed in the Indomalaya, and Australian regions: from Australia to Brunei, China, India, Indonesia, Malaysia, Myanmar, Papua New Guinea, the Philippines and Singapore. We find that specimens currently determined as M. camillae display remarkable morphological variation, some of which appears not to be intra-specific but rather due to differences among species. For example, we found a small-sized queen with vestigial wings in Indonesian material (CASENT0009854), workers with longer setae on the anteromedial portion of the clypeus in specimens from New Guinea, a large queen with simple setae on the pronotal dorsum in specimens from China (CASENT0275389), and a strange yellow male from Australia (CASENT0172083).
We have found that clarifying species boundaries in the genus Mystrium is much more complicated than would be expected from previous studies. The differences among some sibling species similar to M. camillae could be recognizable only in a particular sex, caste, or phenotype, as in the case of M. camillae and M. barrybressleri. A reexamination of the species boundaries of Mystrium camillae based on a detailed comparative study using comprehensive colony samples from each local region will be necessary. Posterolateral corner of head moderately expanding posteriorly. Posterior face of vertex forming almost right angle with its dorsal face on median line of head, so that declivity of vertex on lateral part as steep as that on median part. Ventral half of vertex sculptured. Eye relatively larger than that of M. barrybressleri. Anterior margin of clypeus convex with long conical setae, of which median pair larger than adjacent pair. Genal tooth of head relatively long, as long as lateral lobe of clypeus. Masticatory margin of mandible in full-face view slightly visible on its basal half, invisible on its distal half. Width of dorsal surface of mandible almost identical from mandibular shaft to distal portion. Second maxillary palpomere shorter than third. First flagellomere (third antennal segment) as long as pedicel (second antennal segment). Central part of pronotal dorsum and lateral surface of pronotum strongly and regularly reticulate. Mesonotum often not differentiated and indistinct from propodeum in dorsal view, its length as long as that of propodeum. Metanotal groove indistinct in lateral view, but mesonotum slightly higher than pronotum. Short, but distinct ridge present on dorsal edge of metapleural gland bulla. Petiole wide in dorsal view, but narrower than that of M. barrybressleri (PtI<185 Wings present, well developed. Wing sclerites fully developed even if wings have dropped off. Posterolateral corner of head moderately expanding posteriorly; expansion weaker than that of workers. Posterior face of vertex forming almost right angle with its dorsal face on median line of head, so that declivity of vertex on lateral part as steep as that on median part. Ventral half of vertex sculptured. Eye well developed. Both anterior and lateral ocelli clearly present, median portion of lateral ocelli and posterior portion of anterior ocellus edged by blackish pigment. Anterior margin of clypeus convex, with long conical setae, of which median pair larger than adjacent pair. Genal tooth of head distinctly developed, reaching slightly posterior of lateral lobe of clypeus. Masticatory margin of mandible almost invisible in full-face view, and dorsal surface on distal portion as wide as that on mandibular shaft. Spatulate seta present on basal side of each basal denticle on masticatory margin of mandible. First flagellar segment on antenna as long as pedicel. Setae on pronotum distinctly spatulate, widened distally. Propodeal declivity in lateral view slightly convex on dorsal part of metanotal gland bulla, making blunt angle with its dorsal margin. Petiole relatively long in dorsal view, 0.5× length of abdominal segment III.

Mystrium labyrinth
Body color blackish brown. Male unknown. Etymology. This species name is the English word labyrinth, inspired by the strong reticulation covering the body surface of the new species. The species epithet is a noun, and thus invariant.
Remarks. Mystrium labyrinth females can be distinguished easily from the other Mystrium females in the Malagasy region by a combination of the following characters: the pronotal dorsum covered with strong regular reticulation (as in Fig. 13A); the second maxillary palpomere shorter than the third (as in Fig. 13C); the long genal tooth of head nearly reaching anterior end of the lateral lobe of the clypeus (Fig. 14B); and the first flagellomere (third antennal segment) as long as the pedicel (second antennal segment) (Fig. 10D). The worker of M. labyrinth is similar to that of Mystrium silvestrii; however, M. labyrinth can be distinguished from M. silvestrii by an anterior clypeal margin with long, well-developed conical setae (weak in M. silvestrii and Mystrium mz01), and a moderately long petiole (Fig. 35C) (short and wide in M. silvestrii as in M. barrybressleri).
The queen described here is not associated with workers from the same colony (collected alone from -12.51444°, 49.18139°); however, the similarities between workers and queens are sufficient to determine their relationship. The male of this species is unknown.

mysticum species group
The mysticum species group is endemic to the Malagasy region, and consists of two species, Mystrium mysticum Roger, 1862 andMystrium rogeri Forel, 1899. Unique characters in this group are: a longitudinal carina on the central portion of the labrum in worker and queen (Fig. 8A); a single lateral spine on abdominal sternum VII in worker and queen (Fig. 8C); and unsculptured ventral area on the vertex in worker (Fig. 11A). The separation of the queens is usually straightforward in this species group except for intercastes, because queens have wings and completely developed wing-sclerites. The intercastes, which are the intermediate individuals between worker and queen having the lateral and mid ocelli and the wing sclerites, are sometimes found and all characters of this caste could be varied from absent to incompletely developed in variable degrees. These intercastes could be keyed out as either workers or queens; however, the description of the intercastes was included in the queen section.
The two phenotypes mentioned for the camillae species group are present in the workers of mysticum species group, similarly to the voeltzkowi species group (Table  1). One has a somewhat curved apical tooth (as in Fig. 15A), larger body size, and darker body color; and the other has a straight apical tooth, smaller body size, somewhat narrower body setae, and brighter body color. However, the division of the two phenotypes is weaker than that in the camillae species group, and no other distinct differences could be found between the two phenotypes except for the four characters mentioned above. The "small bright color" phenotype does not seem to be functional in reproduction as in the ergatoid queens of the voeltzkowi species group. In this study, we regard these two phenotypes as morphological variation observed in the worker caste, and we reference the larger one as a major worker and the smaller one as a minor worker when necessary. Posterolateral corner of head strongly expanding posteriorly. Posterior face of vertex forming slightly blunt angle with its dorsal face on median line of head, so that declivity of vertex on lateral part slightly steeper than on median part. Ventral half of vertex smooth and not sculptured. Eye small to moderately small. Anterior margin of clypeus convex and with short conical setae. Genal tooth of head undeveloped: anterolateral corner of head angulate. Masticatory margin of mandible widely visible in full-face view, difference in width of dorsal surface of mandible relatively large between mandibular shaft and distal portion. Second maxillary palpomere longer than third. First flagellomere (third antennal segment) as long as pedicel (second antennal segment). Shallow and fine longitudinal striae irregularly impressed on central part of pronotal dorsum, sometimes with shallow reticulation around them. On lateral surface of pronotum, shallow, fine longitudinal striae impressed, wide and shallow punctures usually arranged on central horizontal line. Mesonotum differentiated from propodeum in dorsal view, length as long as that of propodeum in large individuals, shorter than propodeum in small individuals. Metanotal groove shallowly impressed in lateral view, mesonotum as high as pronotum in large individuals, higher than pronotum in small individuals. Metanotum weakly developed in largest individuals. Metapleural gland bulla developed, so that propodeal declivity in lateral view weakly convex posteriorly on its ventral portion. Petiole in dorsal view moderately wide, and relatively wider than that of M. rogeri.

Mystrium mysticum
Body    in variable degrees in intercaste. Posterolateral corner of head strongly expanding posteriorly; expansion even stronger than that of workers. Posterior face of vertex forming slightly blunt angle with dorsal face on median line of head, so that declivity of vertex on lateral part slightly steeper than on median part. Ventral half of vertex smooth and not sculptured. Eye well developed. Both anterior and lateral ocelli clearly present and developed in most cases, lateral ones rarely smaller; ocelli varied from absent to developed in intercaste. Anterior margin of clypeus convex with small conical setae. Anterolateral portion of head angulate or with short spine. Masticatory margin of mandible widely visible in full-face view, and dorsal surface on distal portion distinctly wider that on mandibular shaft, difference in width much larger than that in M. rogeri. Mandibular teeth (a row of dentition on dorsal side) often lacking on mid-portion of mandibular shaft. A spatulate seta present on basal side of each basal denticle on masticatory margin of mandible. First flagellar segment on antenna as long as pedicel. Setae on pronotum almost simple, narrowing distally with strongly sharpened apex. Propodeal declivity in lateral view almost straight and forming right to slight obtuse angle with dorsal margin, ventral portion with small convexity by metapleural gland. Petiole relatively long in dorsal view, about 0.6× length of abdominal segment III.
Body color black. Eye relatively small, occupying about 0.5× length of head. Ocelli relatively distant from dorsal margin of head or just failing to reach dorsal margin in full-face view. Dorsal margin of head in full-face view rounded. Both anterior and lateral ocelli small. Distance between lateral ocellus and eye relatively long, about 3× longer than diameter of lateral ocellus. Posterior half of vertex clearly differentiated from dorsal half, its dorsal face almost as long as its posterior face. Palpal formula 4,3. First segment of maxillary palp flattened and distinctly wider than second segment. Second maxillary palpomere longer than third. Notauli clearly impressed on mesoscutum. Petiole in dorsal view thin, its length about 0.65× that of abdominal tergite III. Petiolar dorsum covered with rough, deep punctures. Abdominal tergum VIII roughly and deeply punctured.
Abdominal sternum IX punctured on its distal portion. Basal ring short, not extending basally. Telomere extending slightly further distally than digitus. Basoventral expansion of aedeagus well developed basoventrally, distinctly longer than dorsal extension. Ventral margin of aedeagus strongly curved ventrally in lateral view. Aedeagus weakly narrowing distally, its distal portion widely rounded.
On forewing, cu-a located at junction of Media (M) and Cubitus (Cu). Body color reddish brown to black. Distribution. MADAGASCAR and COMOROS: as in Figure 49A. Additional material examined. In addition to the type material, specimens from the following localities were examined in this study: MADAGASCAR. Antsiranana.   Remarks. Mystrium mysticum females can be distinguished easily from other Mystrium females by a combination of a distinct longitudinal carina on the central portion of the labrum (Fig. 8A) or a single lateral spine on abdominal sternum VII (Fig. 8C), and the convexity of the anterior margin of the clypeus (Fig. 9C). In addition to these diagnostic characters, a combination of unsculptured and smooth areas on the ventral portion of the vertex (Fig. 11A) and the convexity of the anterior margin of the clypeus (as in Fig. 9C) separates M. mysticum from the other Mystrium species in workers; in larger-sized alate queens (HW>2.00), simple setae on the pronotal dorsum (Fig. 9A) separate M. mysticum from the queens of other species. For males, the small eye and small ocelli that are separated from each other by more than 3× the maximum diameter of lateral ocellus (Fig. 26B), petiolar dorsum (Fig. 28C) and abdominal ter-gum VIII (Fig. 28A) with rough and deep punctures, and well-developed basoventral expansion of aedeagus (Fig. 29A) separate M. mysticum from the other Mystrium males in the Malagasy region.
Several diagnostic characters for M. mysticum have been given in previous papers, (e.g. Forel 1891; Menozzi 1929); however, our comparative study revealed most of those characters could not cover the whole range of the morphological variation observed in Mystrium species. Here we propose a new set of diagnostic characters to identify Mystrium mysticum. The members of the mysticum species group display remarkable morphological variation in body size, relative mandibular length, relative shape of petiolar node in dorsal view, and body color. The available material revealed that the morphological variation within M. mysticum is larger than the difference between M. mysticum and the species it most resembles, M. rogeri. For example, in order to separate M. rogeri from M. mysticum, Menozzi (1929) used the relatively wider shape of the petiolar node in dorsal view and the shape of setae (=hairs). However, in these two species the index of the petiolar node largely overlaps (PtI 153-183 vs. PtI 146-168) and the shape of setae varies gradually from spoon-shaped (small minor worker) to simple (large major worker) in single colony members. The separable characters for workers proposed in Forel (1895Forel ( , 1899 cannot distinguish species in the mysticum species group. In fact, his syntypes for Mystrium stadelmanni consisted of minor workers of M. mysticum and M. rogeri (see below). In addition to the above variation, Molet et al. (2012) reported the presence of intercaste "mosaic monsters" in M. rogeri, which are intermediate individuals having zero to three ocelli and mesosoma with incompletely developed wing sclerites. We confirmed the same intercastes in M. mysticum as well, and these individuals will be keyed out with either workers or queens in our new identification key.
The redescription of the male of Mystrium mysticum in this study may be the first actual description of the male of this species. Previous male descriptions for M. mysticum have been given in Forel (1891), Emery (1899), and Menozzi (1929). Of these descriptions, Forel (1891) and Menozzi (1929) are assumed to share the same specimen in ZMHB, Berlin. Although we have not examined the actual male specimen in ZMHB, the character description of the specimen does not fit the males in the mysticum species group. For example, the color of the abdomen is light brown in Forel (1891), and the eye occupies almost the entire lateral margin of the head in Menozzi (1929). Rather, those characters fit males in the voeltzkowi species group. We could not find useful information in Emery's description (Emery 1899); however, he did not mention any distinct differences from known males at that point. We have already confirmed that the male in the type series of M. rogeri is actually a male of M. oberthueri, so only males of the voeltzkowi species group, i.e. M. voeltzkowi and M. oberthueri were known in 1899. Hence, we consider all male descriptions for M. mysticum in previous studies unreliable.
Clarification of the male-worker association in Mystrium mysticum in this study marks a great advance; however, some problems remain, especially in the level of confidence in diagnosing males. External characters, e.g. strong and rough punctures on the petiolar dorsum (Fig. 28C) and abdominal tergite VIII (Fig. 28A), distinguish the mysticum species group from the other Malagasy species. However, a specific character is neither possible nor practical to distinguish Mystrium mysticum males from M. rogeri males. In fact, although we proposed the aedeagal character to separate M. mysticum and M. rogeri (Fig. 29), this difference cannot be confirmed without dissection. Unexpected remarkable variation could exist in the genital character as well as the other external characters. Further and more thorough taxonomic examinations of males are necessary to find diagnostic characters able to separate these two species.
The neotype of Mystrium mysticum was designated in this study because we concluded that the original type specimens of this "mysterious species" are lost. Our study revealed that no one examined any of the original type specimens after the publication of the original description by Roger (1862). According to that paper, the description was made based on two queens, and at least one of these was alate. The two queens must have been deposited in the museum in Paris; however, we could not find these queens in the MNHN collection. Although the depository of Mystrium mysticum was given again in the original description of M. camillae Emery, 1889, Emery did not provide any further information. Santschi (1914) discussed the differences between M. silvestrii and M. mysticum when he described M. silvestrii as a new species; however, he only provided information about workers, not queens. Forel described five species in the genus Mystrium from 1895 to 1899, and he noted in the latest description for M. rogeri Forel, 1899 that he never examined the actual syntype of M. mysticum. In spite of these difficulties, we still conclude that the species having queens with simple setae is M. mysticum for the following four reasons: (1) we recognize only two species in the mysticum species group from a large number of Mystrium specimens accumulated from the whole range of the Malagasy region; (2) when Forel (1899) described M. rogeri only queens with simple setae were known from Madagascar and were constantly identified as M. mysticum; (3) as Forel (1899) mentioned, Roger (1862) did not describe any specialized character for the queen's setae; and (4) the queens of M. rogeri have spatulate setae as Forel (1899) expected. In fact, the oldest queen specimen with spatulate setae, which is found in MNHN, was collected after Forel's description (collection date is 1919).
A queen was chosen as the neotype because the original syntypes were two queens. The queen caste also offers a greater number of characters to distinguish this taxon from others in this species group. The location of the neotype was not based on a proximity to the original syntype series, but on the availably of a large collection series that included workers, males and queens.
Mystrium stadelmanni Forel, 1895 is synonymized with Mystrium mysticum in this study. M. stadelmanni is described as a species having a longer petiole, narrower mesosoma, and less constricted abdomen compared with M. mysticum; however, our observations reveal that the boundary of this species is doubtful. None of the diagnostic characters above represent more than intraspecific variation observed in a series of workers from a single colony of M. mysticum. The fact that syntypes of M. stadelmanni consist of two minor workers of M. mysticum and one minor worker of M. rogeri certi-fies that Forel's determination (Forel 1895) was not sufficient to clarify proper species boundaries in this group. In addition to the above, we should note one more important point. Forel (1895) determined specimens in ZMHB to be the new species M. stadelmanni based on comparisons with workers of M. rogeri that he probably misidentified as M. mysticum. Forel believed his specimens, which were given by Emery (Forel 1892), were M. mysticum by that time (Forel 1892;1895); however, he described M. rogeri in 1899 by using workers from the same collection. Our conclusion of a misidentification of M. rogeri is supported by his description of a "longer petiolar node in M. stadelmanni" which is correct if Forel (1895) (1895) is determined to be the first actual description for the M. mysticum worker.
Updating species boundaries and providing easier identification tools for Mystrium mysticum will help reorganize and connect existing ecological and phylogenetic information to proper species names in Mystrium. This taxonomic revision permits us to update identifications in recent ecological publications (Molet et al. 2009;Molet et al. 2007a;Molet et al. 2012), changing some of their identification from M. rogeri to M. mysticum: for example, five colonies (BLF00519, BLF04466, BLF04770, BLF06144, and BLF10994) out of 16 colonies listed as M. rogeri in Table 1 of Molet et al. (Molet et al. 2007a) and out of material for Molet et al. (2009) are those of M. mysticum; a queen of M. rogeri in Figure  2 of Molet et al. (2009) and in Figure 1 and supplemental Figure A1 of Molet et al. (2012) is that of M. mysticum. A worker of M. mysticum was referred to as Mystrium 'red' in Figure  2 of Molet et al. (2009), as was supplemental Figure A1 of Molet et al. (2012) (see also in M. mirror). On the other hand, "M. mysticum" in the material of three recent phylogenetic papers (Brady et al. 2006;Kück et al. 2011;Moreau and Bell 2013;Ouellette et al. 2006) are M. voeltzkowi, not M. mysticum. "M. mysticum" in Molet et al. (2009   .6 (10 specimens measured).
Posterolateral corner of head strongly expanding posteriorly. Posterior face of vertex forming slightly blunt angle with dorsal face on median line of head, so that declivity of vertex on lateral part slightly steeper than on median part. Ventral half of vertex smooth and not sculptured. Eye small to moderately small. Anterior margin of clypeus straight or weakly concave and with small conical setae. Genal tooth of head undeveloped, but anterolateral corner of head angulate. Masticatory margin of mandible narrowly visible in full-face view, and difference in width of dorsal surface of mandible relatively small between mandibular shaft and distal portion. Second maxillary palpomere longer than third. First flagellomere (third antennal segment) as long as pedicel (second antennal segment). Shallow and fine longitudinal striae irregularly impressed on central part of pronotal dorsum, and sometimes shallowly reticulated around those striae. On lateral surface of pronotum, deep, fine longitudinal striae impressed and wide and deep punctures arranged on central horizontal line. Mesonotum differentiated from propodeum in dorsal view, length as long as that of propodeum in large individuals, shorter than propodeum in small individuals. Metanotal groove shallowly impressed, mesonotum in lateral view as high as pronotum in large individuals, higher than pronotum in small individuals. Metanotum weakly developed in largest individuals. Metapleural gland bulla developed, so that propodeal declivity in lateral view weakly convex posteriorly on ventral portion. Petiole in dorsal view moderately narrow, relatively narrower than that of M. mysticum.
Body color brown to black. Wings usually present and well developed; but lacking in intercaste. Wing sclerites fully developed even if wings have dropped off in queen; developed to undeveloped in variable degrees in intercaste. Posterolateral corner of head strongly expanding posteriorly, expansion not differentiated from that of workers. Posterior face of vertex forming slightly blunt angle with dorsal face on median line of head, so that declivity of vertex on lateral part slightly steeper than on median part. Ventral half of vertex smooth and not sculptured. Eye well developed. Both anterior and lateral ocelli clearly present in queen; ocelli varied from absent to developed in intercaste. Anterior margin of clypeus straight or weakly concave with small conical setae. Anterolateral portion of head weakly to strongly angulate or with short spine. Masticatory margin of mandible narrowly, slightly visible in full-face view, dorsal surface on distal portion moderately wider than on mandibular shaft, difference in width much smaller than that in M. mysticum. Teeth always present on whole mandibular shaft. Spatulate seta present on basal side of each basal denticle on masticatory margin of mandible. First flagellar segment on antenna as long as pedicel. Setae on pronotum spatulate, widened distally with sharp or blunt apex. Propodeal declivity in lateral view almost straight and forming right to slightly obtuse angle with its dorsal margin, of which ventral portion with small convexity by metapleural gland. Eye relatively small, occupying about half of head length. Ocelli relatively distant from dorsal margin of head in full-face view, or just failing to reach dorsal margin. Dorsal margin of head in full-face view rounded. Both anterior and lateral ocelli small. Distance between lateral ocellus and eye long, 3× longer than diameter of lateral ocellus. Posterior half of vertex clearly differentiated from dorsal half, dorsal face almost as long as posterior face. Palpal formula 4,3. First segment of maxillary palp flattened and distinctly wider than second segment. Second maxillary palpomere longer than third. Notauli clearly impressed on mesoscutum. Petiole in dorsal view thin, its length about 0.5-0.65× that of abdominal tergite III. Petiolar dorsum covered with rough, deep punctures. Abdominal tergum VIII roughly and deeply punctured.
Abdominal sternum IX punctured on its distal portion. Basal ring short, not extending basally. Telomere extending slightly further distally than digitus. Basoventral expansion of aedeagus less developed basoventrally, as long as dorsal extension. Ventral margin of aedeagus gently curved ventrally in lateral view. Aedeagus weakly narrowing distally, its distal portion widely rounded.
On forewing, cu-a located at junction of Media (M) and Cubitus (Cu). Body color reddish brown to black. Distribution. MADAGASCAR and COMOROS: as in Figure 49B. Additional material examined. In addition to the type material, specimens from the following localities were examined in this study: MADAGASCAR. Diego-Suarez.  (Fig. 8A) or a single lateral spine on abdominal sternum VII (Fig. 8C), and the straight anterior margin of the clypeus (Fig. 9D). In addition to these diagnostic characters, a combination of unsculptured and smooth areas on the ventral portion of the vertex (as in Fig. 11A) and the straight anterior clypeal margin separates M. rogeri from the other Mystrium species workers; and a larger-sized alate queen (HW>1.99) with spatulate setae on the pronotal dorsum (Fig. 9B) separate M. rogeri from the queens of other species. For males, the small eye and small ocelli (as in Fig. 26B) that are separated from each other by a distance of more than 3× the maximum diameter of lateral ocellus, petiolar dorsum (as in Fig. 28C) and abdominal tergum VIII (as in Fig. 28A) with rough and deep punctures, and less developed basoventral expansion of aedeagus (Fig. 29B) separate M. rogeri from the other Mystrium males in the Malagasy region.
We propose a new set of diagnostic characters in this study to correctly identify Mystrium rogeri. The diagnostic characters listed above are constant in females regardless of their body size. Because of the remarkable morphological variation in each species, most diagnostic characters for M. rogeri proposed in previous papers (e.g. Emery 1899; Menozzi 1929) cannot separate M. rogeri from the other Mystrium species, especially from M. mysticum. Mystrium rogeri also displays distinct variation in body size, color, relative mandible length, and shape of the petiole. The difference in the petiolar shape (Emery 1899;Menozzi 1929) cannot be used for complete separation between M. rogeri and M. mysticum (PtI 153-183 vs. PtI 146-168). As mentioned in their discussion of M. mysticum, Molet et al. (2012) reported "mosaic monster" intermediates between queen and worker. The existence of this intercaste is not unique to M. rogeri, but rather a character of the mysticum species group that we confirmed in M. mysticum as well. In this study, the intercastes are keyed out with either of the workers or queens of each species.
As we discussed in Mystrium mysticum, further studies are necessary to better separate the males of M. rogeri and M. mysticum. Here we propose only an aedeagal character to separate these very similar males (Fig. 29).
The lectotype for Mystrium rogeri is designated. M. rogeri was described by Forel in 1899 using at least two workers and one male collected in Madagascar, and we confirm one worker and one male in his collection. Although we could not find any information to identify which type specimens were used in his original description (Forel 1899), Forel mentioned that the workers he determined as M. rogeri were the same specimens he determined as Mystrium mysticum (Forel 1892), and which Emery had sent him. According to the description in Forel (1892), the worker of M. rogeri is smaller and stouter than the worker of M. mysticum; however, we could not deduce how he determined the boundary of "M. mysticum" in the worker caste. To further complicate the story, when Forel described M. rogeri in 1899, he probably determined workers of actual M. mysticum as Mystrium stadelmanni (see remarks for Mystrium mysticum). Moreover, once he determined Emery's material as M. rogeri, the worker caste of M. mysticum should have been considered "unknown" once again. In addition to the misidentification of the worker caste, his description did not provide any details to identify which male specimens he studied. We also confirm Forel's suspicions (Forel 1899) that the male described as M. rogeri is not conspecific with the worker. We verify that the male specimen in his collection, which was probably used for the description of M. rogeri, is Mystrium oberthueri.
Mystrium rogeri is one of the most common Mystrium species used in recent ecological and phylogenetic studies (Molet et al. 2009;Molet et al. 2007a;Molet et al. 2012;Moreau et al. 2006;Saux et al. 2004). However, the identification of their voucher material is now reassessed: part of the material identified as M. rogeri in Molet et al. (Molet et al. 2009;Molet et al. 2007a;Molet et al. 2012) is not M. rogeri (see remarks under M. mysticum).

voeltzkowi species group
The voeltzkowi species group is endemic to the Malagasy region, and consists of six species: Mystrium eques sp. n., Mystrium janovitzi sp. n., Mystrium mirror sp. n., Mystrium oberthueri Forel, 1897, Mystrium shadow sp. n., and Mystrium voeltzkowi Forel, 1897. The complete lack of alate queens and the presence of ergatoid queens that are relatively smaller than conspecific workers are unique characters in this group.
In this study, we define ergatoid queens as the individuals having a mandible with an apical tooth that is moderately sclerotized and directed ventrally ( Fig. 15B; Table 1). On the other hand, the workers are defined as individuals having a mandible with an apical tooth that is hardly sclerotized and curved inward (Fig. 15A). This definition for the ergatoid queen in the voeltzkowi species group is assumed on the basis of previous ecological reports by Molet et al. (Molet et al. 2007a;Molet et al. 2007b), although we did not confirm whether the "ergatoid queens" defined here are functional queens in all species. The morphological differences between workers and ergatoid queens in voeltzkowi species group are distinctly pronounced (as in Fig. 50F vs. 53F), which are shown in Molet et al. (2009) as well, compared with the differences observed between major and minor workers (as in Fig. 43B vs. 43D) in camillae and mysticum species groups (see also description of each group above). The body size of the ergatoid queens is usually smaller than that of workers in the same colony; however, the ranges largely overlap in species level. Both castes in some species display remarkable morphological variation in body size and shape, setae, or body sculpture, while the differences among species are often slight. The identification of specimens unassociated with the other castes or sexes is often problematic. Most specific differences are observed when individuals within the same body size range are compared. Posterolateral corner of head strongly to moderately expanding posteriorly. Posterior face of vertex forming almost a right angle with dorsal face on median line of head, so that declivity of vertex on lateral part as steep as that on median part. Whole region of vertex finely striated. Eye relatively smaller than that of M. oberthueri. Anterior margin of clypeus strongly convex with long conical setae. Genal tooth of head relatively long, as long as lateral lobe of clypeus. Masticatory margin of mandible almost invisible in full-face view. Width of dorsal surface of mandible on distal portion slightly wider than that on mandibular shaft. Second maxillary palpomere longer than third. First flagellomere (third antennal segment) as long as pedicel (second antennal segment). Strong, deep and thick longitudinal striae regularly impressed on whole central part of pronotal dorsum. Strong, deep, and thick longitudinal striae impressed on lateral surface of pronotum. Mesonotum differentiated from propodeum in dorsal view, length slightly shorter than that of propodeum. Metanotal groove shallowly and gently  impressed, and mesonotum higher than pronotum in lateral view. Metapleural gland bulla moderately developed, and propodeal declivity in lateral view almost straight. Petiole widened on posterior 1/4 and gently narrowed anteriorly in dorsal view, anterior margin straight to gently rounded and sometimes edged by thin striae.

Mystrium eques
Body color blackish brown. Appendages brighter, and four distal segments of antennal club yellowish.   Wings vestigial and forming small but distinct appendages. Wing sclerites undeveloped. Posterolateral corner of head strongly expanding posteriorly, expansion relatively weaker than that of workers. Posterior face of vertex forming almost a right angle with dorsal face on median line of head, so that declivity of vertex on lateral part as steep as on median part. Ventral half of vertex sculptured. Eye small but distinct. Ocelli absent. Anterior margin of clypeus distinctly convex with long to short conical  setae. Anterolateral portion of head with short spine. Masticatory margin of mandible almost invisible in full-face view. Dorsal surface of mandible on distal portion slightly wider than that on mandibular shaft. Spatulate or narrow spoon-shaped seta present on basal side of each basal denticle on masticatory margin of mandible. First flagellar segment on antenna about 1.0-1.2× length of pedicel. Setae on pronotum distinctly spatulate, widened distally with sharp or blunt apex. Metapleural gland bulla moderately developed, not expanding to dorsum of propodeal spiracle, so that propodeal declivity in lateral view weakly convex and rounded posteriorly on its ventral 1/3. Petiole relatively long in dorsal view, about 1.0-0.8× length of abdominal segment III.
Body color blackish to reddish brown. Male unknown.
Etymology. This species name is the Latin word eques, inspired by the distinctly developed clypeus (shield) on the new species. The species epithet is a noun and invariant.
Distribution. MADAGASCAR: as in Figure 56A. Additional material examined. In addition to the type material, specimens from the following localities were examined in this study: MADAGASCAR. Toamasina.  Remarks. The worker of Mystrium eques is easily distinguished from those of other Mystrium species by a combination of the following characters: strong and deep longitudinal sculpture on the dorsal and lateral surface of pronotum (as in Fig. 17B); anteromedial portion of the clypeus strongly convex anteriorly (Fig. 18B); and sharper angle between the posterior and dorsal faces of the vertex on the median line (as in Fig. 16A). The queen of M. eques is differentiated from other Mystrium queens by having vestigial wings reduced to small appendages (as in Fig. 25D) with undeveloped wing sclerites (ergatoid), posterior face of the vertex forming approximately a right angle with its dorsal face on the median line (Fig. 21A), and a strongly convex anterior clypeal margin (Fig. 22A).
The worker of M. eques is similar to that of M. oberthueri, however, the strongly convex anterior clypeal margin separates M. eques (Fig. 18B) from M. oberthueri (Fig.  18A), in which the anterior margin of the clypeus is straight. The male of M. eques is not yet known. Because of the similarity between the workers and queens of M. eques and M. oberthueri, we assume the male of M. eques is also similar to that of M. oberthueri. The most easily distinguishable character between workers of M. eques and M. oberthueri, the anterior margin of the clypeus (Figs 18A, B), is unfortunately not useful for males because the anterior clypeal margin and the conical setae on it are less developed in males. Mystrium eques (Fig. 56A) has a sympatric distribution with M. oberthueri (Fig. 56D), so collection locality does not separate these two similar species either. The current male specimens identified as M. oberthueri may contain males of M. eques. Further studies are necessary to resolve this issue.    (10 specimens measured).

Mystrium janovitzi
Posterolateral corner of head moderately expanding posteriorly. Posterior face of vertex forming almost right angle with dorsal face on median line of head, so that declivity of vertex on median part even steeper than on lateral part. Ventral half of vertex sculptured. Eye usually large, but sometimes small. Anterior margin of clypeus straight with small, extremely low conical setae. Genal tooth of head undeveloped, anterolateral corner of head angled. Masticatory margin of mandible almost invisible in full-face view, and width of dorsal surface of mandible almost identical from mandibular shaft to distal portion. Second maxillary palpomere longer than third. First flagellomere (third antennal segment) about as long as pedicel (second antennal segment). Shallow, rough, and fragmented longitudinal reticulation irregularly impressed along the entire pronotal dorsum. Shallow, rough, and fragmented longitudinal reticulation irregularly impressed on lateral surface of pronotum. Mesonotum often not differentiated from propodeum in dorsal view, length as long as that of propodeum. Metanotal groove indistinct in lateral view. Metapleural gland bulla moderately developed, and propodeal declivity in lateral view almost straight though sometimes weakly rounded. Petiole relatively long (PtI 138.7-160.3), with slightly widened posterior 1/4 to 1/3 in dorsal view, anterior margin gently rounded or almost straight, and never edged by striae.
Wings vestigial and reduced to small appendages, which are sometimes indistinct. Wing sclerites undeveloped. Posterolateral corner of head moderately expanding posteriorly, expansion not differentiated from that of workers. Posterior face of vertex forming almost right or even slightly acute angle with dorsal face on median line of head, so that declivity of vertex on median part even steeper than on lateral part. Ventral half of vertex sculptured. Eye varied from well developed to small. Ocelli absent. Anterior margin of clypeus straight with extremely short and small conical setae. Genal tooth of head absent, corner usually unangled, sometimes angled into a small, short spine. Masticatory margin of mandible almost invisible in full-face view, and dorsal surface on distal portion as wide as that on mandibular shaft. Spatulate seta present on basal side of each basal denticle on masticatory margin of mandible. First flagellar segment on antenna as long as pedicel. Setae on pronotum usually narrowing distally, or sometimes widened distally, with strongly sharpened apex. Metapleural gland bulla less developed, not expanding to either dorsum of propodeal spiracle or propodeal declivity margin in lateral view, so that posterior margin of propodeum in lateral view almost straight. Petiole relatively long in dorsal view, about 0. Eye moderately large, occupying 0.6× head length. Ocelli reaching to dorsal margin of head in full-face view, or just failing to reach dorsal margin. Dorsal margin of head in full-face view rounded. Both anterior and lateral ocelli small. Lateral ocellus small and distant from eye: distance between these more than 2× maximum diameter of lateral ocellus. Posterior face of vertex rising steeply from neck, dorsal face slightly shorter than posterior face. Palpal formula 4,3. First segment of maxillary palp flattened and distinctly wider than second segment. Second maxillary palpomere longer than third. Notauli deeply to shallowly impressed on mesoscutum, but often unclear. Petiole in dorsal view thin, length 0.8-0.65× that of abdominal tergite III. Petiolar dorsum covered with shallow, sparse, irregular punctures, or almost smooth. Abdominal tergum VIII without deep punctures, almost smooth.
Distal portion of abdominal sternum IX smooth and not punctured. Basal ring long, distinctly extending basally. Telomere distinctly extending farther distally than digitus. Basoventral expansion of aedeagus moderately developed basoventrally, longer than dorsal extension. Ventral margin of aedeagus gently curved ventrally in lateral view. Aedeagus moderately narrowing distally on distal half, distal portion widely rounded.
On forewing, cu-a located far basal from junction of Media (M) and Cubitus (Cu). Body color reddish brown to black. Etymology. The specific epithet is the patronym of Dr. Tyler Janovitz in recognition of his interest in myrmecology and his support for research on ants.
Distribution. MADAGASCAR: as in Figure 56B. Additional material examined. In addition to the type material, specimens from the following localities were examined in this study: MADAGASCAR. Antsiranana. Ambohitra (Joffreville) (-12.53333°, 49.16667°), rainforest, 1100 m alt.; Parc National Montagne d'Ambre, 3.6 km 235° SW Joffreville (-12.53444°, 49.1795°), montane rainforest, 925 m alt.; Rés. Ankarana, 7 km SE Matsaborimanga (-12.9°, 49.11667°), rainforst, 150 m alt.;Forêt de Binara, 9. Remarks. The worker of Mystrium janovitzi is distinguished easily from that of the other Mystrium species by the combination of the following characters: the labrum lacking central longitudinal carina (as in Fig. 8B); low clypeal conical setae (Fig.  17C); the posterior face of the vertex forming a right angle with its dorsal face (as in Fig. 16A); wide pronotum covered with fine sculpture that is neither longitudinal nor reticulate (Fig. 17A). This combination of diagnostic characters for the worker is also useful for distinguishing ergatoid queens of M. janovitzi. The ergatoid queen of M. janovitzi is similar to conspecific workers in its sculpture and body color, but differs in having a narrower mandible with a straight apical tooth (as in Fig. 15B) and distinct vestigial wings (as in Fig. 25D). For the male, small ocelli not strongly protruding from posterior margin of head in full-face view (Fig. 31B), rounded posterior margin of head in full-face view, smooth abdominal tergum VIII (Fig. 28B), and vertex strongly rising from neck (Fig. 32B)   .
Posterolateral corner of head strongly expanding posteriorly. Posterior face of vertex forming a blunt angle with its dorsal face on median line of head, so that declivity of vertex on lateral part distinctly steeper than on median part. Ventral half of vertex sculptured. Eye developed, relatively larger than that of M. voeltzkowi. Anterior margin of clypeus straight to weakly convex with moderately long conical setae. Genal tooth of head weakly developed, reaching or slightly exceeding basal line of lateral lobe of clypeus. Masticatory surface of mandible in full-face view visible on basal half and invisible on distal half, width of dorsal surface of mandible almost identical from mandibular shaft to distal portion. Second maxillary palpomere longer than third. First flagellomere (third antennal segment) about 1.0-1.3× length of pedicel (second antennal segment). Pronotal dorsum covered with strong and longitudinal striae, center deeply impressed. Shallow but thick longitudinal striae impressed on lateral surface of pronotum. Mesonotum differentiated from propodeum in dorsal view, length shorter than that of propodeum. Metanotal groove shallowly and gently impressed, mesonotum higher than pronotum in lateral view. Metapleural gland bulla moderately developed, propodeal declivity in lateral view almost straight. Petiole gently narrowing from anterior 1/3 in dorsal view, anterior margin straight to gently rounded and not edged by striae.
Body color reddish brown to dark brown. Four distal segments of antennal club brighter.
Wings usually vestigial and reduced to quite small appendages; sometimes completely absent. Wing sclerites undeveloped. Posterolateral corner of head strongly to weakly expanding posteriorly, expansion relatively weaker than that of workers. Vertex usually thin, forming blunt angle between posterior and dorsal faces on median line of head, so that declivity of vertex on lateral part distinctly steeper than on median part. Ventral half of vertex sculptured, or not differentiated from dorsal region. Eye moderate and distinct. Ocelli absent. Anterior margin of clypeus straight to weakly convex with small conical setae. Genal tooth of head absent and not angled, or angled into small, short spine. Masticatory margin of mandible almost invisible in full-face view, and dorsal surface on distal portion as wide as that on mandibular shaft. Spatulate seta present on basal side of each basal denticle on masticatory margin of mandible. First flagellar segment on antenna moderately long, about 1.1-1.2× length of pedicel. Setae on pronotum almost simple, narrowing distally with strongly sharpened apex. Meta-pleural gland bulla moderately developed and not expanding dorsally to propodeal spiracle, so that propodeal declivity in lateral view weakly convex and rounded posteriorly on its ventral 1/3. Petiole relatively long in dorsal view, about 0.6-0.8× length of abdominal segment III.
Body color yellowish brown, brown or reddish brown. Eye quite large, occupying about 0.75× of head length. Ocelli protruding from dorsal margin of head in full-face view. Dorsal margin of head in full-face view rounded. Both anterior and lateral ocelli large. Distance between lateral ocellus and eye equal to or shorter than diameter of lateral ocellus. Posterior face of vertex clearly differentiated from dorsal face, dorsal face distinctly shorter than posterior face. Palpal formula 4,3. First segment of maxillary palp flattened and distinctly wider than second segment. Second maxillary palpomere longer than third. Notauli shallowly and weakly impressed on mesoscutum, but often unclear. Petiole in dorsal view thin, length 0.55-0.65× that of abdominal tergite III. Petiolar dorsum covered with fine punctures. Abdominal tergum VIII without deep punctures, almost smooth.
Distal portion of abdominal sternum IX smooth and not punctured. Basal ring short, not extending basally. Telomere distinctly extending distally farther than digitus. Basoventral expansion of aedeagus well developed basoventrally, distinctly longer than dorsal extension. Ventral margin of aedeagus almost straight in lateral view. Aedeagus distinctly narrowing distally, distal portion relatively sharp.
On forewing, cu-a located at junction of Media (M) and Cubitus (Cu), or slightly to far basal from junction.
Body color yellowish to reddish brown. Etymology. This species name is the English word mirror, inspired by the remarkable variation displayed in this species. This species might confuse an observer as a magic mirror would, which reflects different views to the observer. The species epithet is a noun and invariant.
Distribution. MADAGASCAR: as in Figure 56C. Additional material examined. In addition to the type material, specimens from the following localities were examined in this study: MADAGASCAR.  (Fig. 24B) separates M. mirror from M. voeltzkowi, and pronotal setae simply narrowing distally with a sharp apex (Fig. 25A) separate it from M. shadow. For the males, a large eye occupying almost 75% of the lateral margin of the head in full-face view (Fig. 26C), large lateral ocelli protruding from the dorsal margin of the head in full-face view (Fig. 38E), and a shorter distance between the eye and lateral ocellus (Fig. 26A) (Fig. 27).
The distribution range of Mystrium mirror (Fig. 56C) is allopatric with that of M. voeltzkowi (Fig. 56F), but M. mirror has a wider distribution than M. voeltzkowi. Most of the specimens of M. mirror were collected from tropical dry forests throughout most of western Madagascar (a few were collected in northern Madagascar), while most M. voeltzkowi were collected from rainforest in the northern part of the country. This clear separation in distribution range and preferred habitat may be useful to roughly distinguish M. mirror from M. voeltzkowi.
We confirm that Mystrium mirror was previously used as material in one phylogenetic study (Ouellette et al. 2006) and two studies of evolutionary biology (Molet et al. 2009;Molet et al. 2012). In Ouellette et al. (2006), the specimen referred to as M. mysticum3 (CASENT0500395) is M. mirror. An ergatoid queen of M. mirror was referred to as M. 'red' in figure 2 in Molet et al. (2009) and supplemental figure A1 in Molet et al. (2012). Prior to this study, the species boundaries among species in the genus Mystrium were too weak and ambiguous to provide proper taxonomic names for use in phylogenetic or ecological studies. However, the ecological data for Mystrium 'red' (Molet et al. 2009;Molet et al. 2007a;Molet et al. 2012) is still considered that of M. voeltzkowi even though an image of M. mirror was used in their figures. In fact, all colonies used as material in Molet et al. (2007a), which we could reexamine in the CASC collection, were identified as M. voeltzkowi (see also M. voeltzkowi). Posterolateral corner of head moderately expanding posteriorly. Posterior face of vertex forming almost right angle with dorsal face on median line of head, so that declivity of vertex on lateral part as steep as on median part. Whole region of vertex finely striated. Eye moderately small. Anterior margin of clypeus straight or weakly convex with long conical setae. Genal tooth of head relatively long, as long as lateral lobe of clypeus. Masticatory surface of mandible almost invisible in full-face view, and width of dorsal surface of mandible almost identical from mandibular shaft to distal portion. Second maxillary palpomere longer than third. First flagellomere (third antennal segment) about 1.2-1.5× length of pedicel (second antennal segment). Strong, deep longitudinal striae regularly impressed on whole central part of pronotal dorsum. Strong, deep longitudinal striae impressed on lateral surface of pronotum. Mesonotum differentiated from propodeum in dorsal view, length slightly shorter than that of propodeum. Metanotal groove shallowly and gently impressed in lateral view. Metapleural gland bulla moderately developed, and propodeal declivity in lateral view almost straight. Petiole widened on posterior 1/3 and gently narrowing anteriorly in dorsal view, anterior margin straight to gently rounded and often edged by thick striae. Wings vestigial and reduced to small but distinct appendages. Wing sclerites undeveloped. Posterolateral corner of head moderately expanding posteriorly, expansion not differentiated from that in workers. Posterior face of vertex forming almost a right angle with dorsal face on median line of head, so that declivity of vertex on lateral part as steep as on median part. Ventral half of vertex sculptured. Eye small but distinct. Ocelli absent. Anterior margin of clypeus almost straight with long to moderate conical setae. Anterolateral portion of head with short spine. Masticatory margin of mandible almost invisible in full-face view, and dorsal surface on distal portion as wide as that on mandibular shaft. Spatulate seta present on basal side of each basal denticle on masticatory margin of mandible. First flagellomere (third antennal segment) long, about 1.2-1.5× length of pedicel (second antennal segment). Setae on pronotum distinctly spatulate, widened distally with sharp or blunt apex. Metapleural gland bulla moderately developed, not expanding dorsally to propodeal spiracle, so that propodeal declivity in lateral view weakly convex and rounded posteriorly on its ventral 1/3. Petiole relatively long in dorsal view, about 0.8× length of abdominal segment III. Eye moderately large, occupying 0.6× of head length. Ocelli relatively distant from dorsal margin of head in full-face view. Dorsal margin of head in full-face view straight. Both anterior and lateral ocelli small. Lateral ocellus small and distant from eye: distance between these more than 1.5× maximum diameter of lateral ocellus. Posterior face of vertex not clearly differentiated from dorsal face, so that vertex almost continuously rounded. Palpal formula 4,3. First maxillary palpomere flattened and distinctly wider than second segment. Second maxillary palpomere longer than third. Notauli clearly impressed on mesoscutum. Petiole in dorsal view thin, 0.8× as long as abdominal tergite III. Petiolar dorsum covered with shallow, irregular punctures. Abdominal tergum VIII without deep punctures, almost smooth.

Mystrium oberthueri
Distal portion of abdominal sternum IX smooth and not punctured. Basal ring moderately long, expanding basoventrally. Telomere distinctly extending distally farther than digitus. Basoventral expansion of aedeagus moderately developed basoventrally, longer than dorsal extension. Ventral margin of aedeagus gently convex in lateral view. Aedeagus moderately narrowing distally and distal portion rounded.
On forewing, cu-a located far basal from junction of Media (M) and Cubitus (Cu). Body color reddish brown to black. Distribution. MADAGASCAR: as in Figure 56D. Additional material examined. In addition to the type material, specimens from the following localities were examined in this study: MADAGASCAR. Remarks. The worker of Mystrium oberthueri is easily distinguished from that of other Mystrium species by the combination of the following characters: strong, deep, and thick longitudinal sculpture on the dorsal and lateral surface of pronotum (Fig.  17B); straight anteromedial margin of the clypeus (Fig. 18A); a sharper angle between the dorsal and posterior faces of the vertex on the median line of the head (Fig. 16A). The queen of M. oberthueri is differentiated from other queens of Mystrium species by the vestigial wings reduced to small appendages with undeveloped wing sclerites (ergatoid: as in Fig. 25D), the posterior face of the vertex forming approximately a right angle with its dorsal face on the median line of the head (as in Fig. 21A), straight anterior clypeal margin (Fig. 22B), and moderately developed clypeal conical setae (Fig. 23B). The male of M. oberthueri can be separated from the other known Mystrium males by its smooth abdominal tergum VIII (as in Fig. 28), small lateral ocelli situated distant from posterior margin of head in full-face view, and straight dorsal margin of head in full-face view (Fig. 30A). We should note that the current male character diagnoses may not be able to distinguish between M. oberthueri and M. eques, which is not yet known from the male caste and may be quite similar to M. oberthueri.
The lectotype for Mystrium oberthueri is designated. According to Forel (1897), a single major worker and multiple ergatoid queens (as minor workers) were used for the original description of M. oberthueri. We confirmed one worker and one ergatoid queen in the Forel collection in MHNG and one ergatoid queen in MCZC, all of which are labeled with the location "Ste. Marie Madagascar." The single worker was chosen as the new representative specimen.
Posterolateral corner of head strongly expanding posteriorly. Posterior face of vertex forming blunt angle with dorsal face on median line of head, so that declivity of vertex on lateral part distinctly steeper than on median part. Ventral half of vertex sculptured. Eye moderately small. Anterior margin of clypeus strongly convex with long conical setae. Genal tooth of head moderately developed, reaching to about half of lateral lobe of clypeus. Masticatory surface of mandible in full-face view slightly visible on basal half and invisible on distal half, width of dorsal surface of mandible almost identical from mandibular shaft to distal portion. Second maxillary palpomere longer than third. First flagellomere (third antennal segment) about 1.0-1.3× length of pedicel (second antennal segment). Pronotal dorsum covered with strong longitudinal striae, which are waved and irregular, but median stria always deeply, straightly and clearly impressed. Strong, deep, and irregular longitudinal striae impressed on lateral surface of pronotum. Mesonotum differentiated from propodeum in dorsal view, length shorter than that of propodeum. Metanotal groove shallowly and gently impressed, mesonotum higher than pronotum in lateral view. Metapleural gland bulla moderately developed, and propodeal declivity in lateral view almost straight but sometimes weakly rounded. Petiole relatively short (PtI ), gently narrowing from anterior 1/3 in dorsal view, anterior margin straight to gently rounded and not edged by striae.
Body color reddish brown to black. Four distal segments of antennal club brighter. Wings vestigial and reduced to small but distinct appendages. Wing sclerites undeveloped. Posterolateral corner of head weakly expanding posteriorly, expansion distinctly weaker than that in workers. Posterior face of vertex forming blunt angle with dorsal face on median line of head, so that declivity of vertex on lateral part distinctly steeper than on median part. Ventral half of vertex sculptured. Eye small but distinct. Ocelli absent. Anterior margin of clypeus widely convex with short conical setae, sometimes with median notch. Genal tooth of head with short spine. Masticatory margin of mandible almost invisible in full-face view, dorsal surface on distal portion as wide as that on mandibular shaft. Spatulate seta present on basal side of each basal denticle on masticatory margin of mandible. First flagellomere (third antennal segment) moderately long, about 1.5-1.2× length of pedicel (second antennal segment). Setae on pronotum widened distally with sharp apex. Metapleural gland bulla moderately developed, not expanding to dorsum of propodeal spiracle nor propodeal declivity margin in lateral view, so that posterior margin of propodeum in lateral view almost straight. Petiole relatively long in dorsal view, about 0.5-0.8× length of abdominal segment III.
Body color reddish to blackish brown.
19A); blunter angle between the dorsal and posterior faces of the vertex on the median line of the head (as in Fig. 16B); the straight posterior declivity of the propodeum (as in Fig.  20). The workers of M. shadow are quite similar to those of M. mirror and M. voeltzkowi, and the differences among the workers of the three species are even slighter in small-sized individuals. When workers in the same body size range are compared, M. shadow can be distinguished from M. mirror by the central pair of conical setae that is larger in size than the adjacent pair (Fig. 19A), and from M. voeltzkowi by a straight declivity of the propodeum (as in Fig. 20). The queen of M. shadow can be distinguished from the other Mystrium queens by a combination of the mesosoma without developed wing sclerites but with vestigial wing appendages (ergatoid: Fig. 25D), posterior face of the vertex forming a blunt angle with its dorsal face on the median line of the head (Fig. 21B), spatulate setae on dorsum of head and pronotum (Fig. 25B), and the metapleural gland bulla that is moderately developed and not expanding dorsally to the propodeal spiracle (as in Fig. 24B). The ergatoid queens similar to those of M. shadow are those of M. voeltzkowi and M. mirror; however, a less-developed metapleural gland bulla distinguishes M. shadow from M. voeltzkowi (as in Fig. 24), and spatulate setae on the pronotal dorsum (Fig. 25B) and the presence of the genal tooth of the head (Fig. 54E) distinguish M. shadow from M. mirror. The male of M. shadow is not that similar to that of M. voeltzkowi or M. mirror. The male of M. shadow can be easily separated from the latter two males by the small ocelli (as in Fig.  26B) and small eye (as in Fig. 26D). On the other hand, the male of M. shadow is similar to that of M. janovitzi. The shape of the vertex is the character that separates these two similar males: the posterior face of the vertex in posterior view is not clearly differentiated from its dorsal face in M. shadow (Fig. 32A), while the posterior face of the vertex is clearly divided from its dorsal face on the median line of the head in M. janovitzi (Fig. 32B). This taxonomic revision permits us to update identifications in a recent ecological publication Molet et al. (2009). Material identified as "Mystrium mysticum" in their paper is M. shadow, not M. mysticum. Posterolateral corner of head strongly expanding posteriorly. Posterior face of vertex forming blunt angle with dorsal face on median line of the head, so that declivity of vertex on lateral part distinctly steeper than on median part. Ventral half of vertex sculptured. Eye moderately small. Anterior margin of clypeus straight to weakly convex with moderately long conical setae. Genal tooth of head moderately developed, reaching about half of lateral lobe of clypeus. Masticatory surface of mandible in full-face view visible on basal half and invisible on distal half, width of dorsal surface of mandible almost identical from mandibular shaft to distal portion. Second maxillary palpomere longer than third. First flagellomere (third antennal segment) about 1.0-1.3× length of pedicel (second antennal segment). Pronotal dorsum covered with strong to weak longitudinal striae often waved and irregular, but median stria always deeply, straightly and clearly impressed. Shallow, fine longitudinal striae irregularly impressed on lateral surface of pronotum. Mesonotum not differentiated from propodeum in dorsal view in small individuals but differentiated in large individuals, its length shorter than that of propodeum. Metanotal groove indistinct in small individuals and shallowly and gently impressed in large individuals; mesonotum higher than pronotum in lateral view. Metapleural gland bulla strongly developed, so that propodeal declivity in lateral view convex posteriorly on ventral side. Petiole widened on posterior 1/4-1/2, gently narrowed anteriorly in dorsal view, anterior margin straight to gently rounded and not edged by striae.
Wings absent or vestigial and reduced to quite small appendages. Wing sclerites undeveloped. Posterolateral corner of head gently expanding posteriorly, expansion distinctly weaker than that in workers. Vertex usually thin, forming blunt angle between dorsal and posterior faces on median line of head, so that declivity of vertex on lateral part distinctly steeper than median part. Ventral half of vertex sculptured, not differentiated from dorsal region. Eye small but distinct. Ocelli absent. Anterior margin of clypeus straight to weakly convex with short conical setae. Genal tooth of head absent, usually even not angular, but angular in smallest individuals. Masticatory margin of mandible almost invisible in full-face view, dorsal surface on distal portion from the same colony of types of M. voeltzkowi. The mistake was due to the remarkable polymorphism in this species. Once the original descriptions of M. fallax and M. voeltzkowi were published, the association between the large black worker (Fig. 50F) and the small reddish individuals (Fig. 53F) was not clarified for more than 110 years. Now accumulated material clearly shows the small reddish individuals are conspecific with the black large worker, and an ecological study (Molet et al. 2007a) revealed these small reddish individuals are ergatoid queens of this strange species (Fig. 1B).
The distribution of Mystrium voeltzkowi is limited to Mayotte and the northern part of Madagascar (Fig. 56F). The collection localities of M. mirror are in tropical dry forest, gallery forest, and spiny forest (Fig. 56C), while those of M. voeltzkowi are in tropical dry forest, montane rainforest, littoral rainforest, and rainforest (Fig. 56F). The distinct development of the metapleural grand bulla observed in M. voeltzkowi (Figs 20F, 24A) may be related to its nesting environment, which has higher moisture levels and requires stricter control of fungi.