Below I provide an outline of doryline biology and diversity. More information on the natural history of each lineage can be found under individual genus accounts.

The 28 genera of the Dorylinae recognized here form a well-supported monophyletic group that is in turn a part of a more inclusive formicoid clade (Brady et al. 2006, Moreau et al. 2006). The formicoids hold almost 90% of all extant ant species. Within this group, the Dorylinae is sister to all other lineages. These include the subfamilies Aneuretinae, Dolichoderinae, Ectatomminae, Formicinae, Heteroponerinae, Myrmeciinae, Myrmicinae, and Pseudomyrmecinae (Ward 2014). Recent dating analyses place the origins of the crown group (that is all extant species including their common ancestor) dorylines at 70–100 million years and indicate that the subfamily has undergone a rapid radiation early in its history (Brady et al. 2014, Borowiec, in prep.). The rapid initial diversification can be perhaps attributed to specialization on other social insects as prey. This adaptation is currently not unique to this clade, but doryline ants were likely the first that evolved it. It is probably a fair generalization to say that dorylines use preformed cavities for nesting and move their colonies often relative to other ants, particularly in the case of the ‘true army ants’ (see below).

Dorylines occur on all continents except Antarctica but are the most prominent in tropical regions of the world. A few species range into the warm temperate zone, as far as the state of New Jersey in northeastern United States and western Turkey and the Dodecanese in the Mediterranean. In the southern hemisphere, they reach southern Australia and Tasmania, South Africa, and at least as far as Chubut province in southern Argentina.

Within the subfamily, a group of genera sometimes termed ‘the true army ants’ (or ‘AenEcDo’ army ants; Kronauer 2009) can be distinguished. The true army ants include the genera Aenictus, Aenictogiton, Cheliomyrmex, Dorylus, Eciton, Labidus, Neivamyrmex, and Nomamyrmex. These ants possess what has been termed the ‘army ant syndrome’: a set of morphological and behavioral characteristics that are common to all species of the group (Brady 2003). The syndrome includes collective foraging, frequent colony relocations, and specialized queen morphology. The true army ants always forage in highly coordinated swarms with no scouts or leading foragers. There are no permanent nests and colonies move periodically, a behavior that is thought to be an adaptation to local depletion of resources. The queens are also highly specialized, wingless and with large abdomens (dubbed ‘dichthadiigynes’), capable of producing large numbers of eggs. Colony foundation in army ants is also unusual. Young queens are unable to start a new colony independently as in most ant species, and instead they rely on a retinue of workers departing with them from their parent nest. An extensive body of literature exists concerning army ant biology and excellent, readable overviews of the subject have been published (Gotwald 1995, Kronauer 2009).

The true army ants currently account for the majority of doryline diversity. More than 45% of the described species of Dorylinae are classified in just two true army ant genera: Aenictus (Figure 9, occurring in the Old World) and Neivamyrmex (Figure 37, New World). As noted above, the army ants are where the largest colony sizes and most conspicuous foraging behaviors evolved. Several species became adapted to taking a variety of invertebrate prey in addition to the more standard social insect diet and during foraging they form the trails that are such a memorable sight to most visitors to the tropics. These formidable columns of ants are observed in the genera Eciton (Figure 24) and Labidus (Figure 28), in the New World and Dorylus (Figure 20) in Sub-Saharan Africa and Southeast Asia. In addition to conspicuous predators with varied diets, Dorylus and Eciton contain species that retain the ancestral condition of preying upon other ants. The latter specialization seems to be a rule for most other species and genera (Aenictus in the Old World, Neivamyrmex and Nomamyrmex (Figure 42) in the New World) of the true army ants for which we have foraging observations.

The sexual dimorphism of the true army ants is remarkable, even compared to other ants, and has contributed to a complex early taxonomic history and the establishment of a ‘dual taxonomy’ in which descriptions of new species were often based on males unassociated with any females (see History of taxonomic and phylogenetic research below).

Aenictogiton, an African lineage now recognized as the sister group to Dorylus and thus in phylogenetic terms nested within the true army ant clade (Brady et al. 2014), deserves a separate mention. The genus has been originally described from a male (Emery 1901b) and its worker remained a mystery for over one hundred years, until the first workers were collected in Uganda in 2006. Since then more workers have been uncovered and their conspecificity with Aenictogiton males from the same area was confirmed through molecular phylogenetics (unpublished data). The present study is the first to formally describe the worker caste of this elusive genus. The worker morphology of Aenictogiton is reminiscent of certain species of the distantly-related Leptanilloides and suggests that these ants are strictly subterranean. This is the only lineage of the true army ants for which no behavioral observations have ever been published.

In addition to the true army ants, the subfamily Dorylinae comprises a variety of forms that prior to the study of Brady et al. (2014) had been classified in the erstwhile subfamilies Cerapachyinae and Leptanilloidinae. The relationships among these lineages are still not entirely understood, but it is now certain that they do not represent a monophyletic group (Borowiec, in prep.). Many of the genera discussed below have been treated as a single genus Cerapachys prior to this revision (Brown 1975). The scarce information on foraging habits indicates that most species prey on other ants or on termites. It is unclear if any species of this assemblage evolved the complete suite of the army ant syndrome, although collective foraging and specialized queens similar to those found in true army ants are known in some lineages.

In much of the myrmecological literature, the true army ants and other ants classified here as dorylines (sometimes referred to as ‘non-army ant dorylines’; all other genera of the former subfamilies Cerapachyinae and Leptanilloidinae) have not been universally recognized as close relatives. As a result of this, the research on the taxonomy and phylogeny of these assemblages has followed largely separate paths. Because of this I keep these histories separate in the account that follows, concluding with a review of how phylogenetic considerations brought these groups together. The summary presented here focuses on the classification at the genus level and above. Remarks on species-level taxonomy can be found under the individual genus accounts.

The first taxon that would eventually be included in the Dorylinae is Dorylus helvolus, which was described from a male by Linnaeus in 1764 as Vespa helvola. The convoluted taxonomic history of this species and the true army ants in general has been vividly described by Gotwald (1982, 1995). Male army ants are very distinctive insects and early entomologists were confused about the identity of these forms because of the absence of worker-male associations. It was not until 1849 when Thomas Savage, a missionary in Africa, was able to observe the strange males together with workers in the field, that this first ‘army ant mystery’ was solved (Savage 1849).

The descriptive work on the true army ants continued with unfortunate proliferation of taxa described based on workers, males, or even gynes unassociated with other sexes or castes. This practice has been especially common in Aenictus and Dorylus but has impacted most true army ant lineages. The result of this approach is a ‘dual taxonomy’ with many forms known from only the worker caste or from males, but not both. In the genus Aenictus, for example, about 50 out of the 180 currently recognized species are known only from the male and no species is known from worker, gyne, and male.

Morphologically distinct lineages of the true army ants exist in the Old and New World and the two regions have not shared a genus-level taxon since at least 1840 (Shuckard 1840a). The affinity of the these ants, however, has long been recognized and the family-level taxon Dorylidae was first erected by Leach (1815) to hold the then-known males of Old World Dorylus and New World Labidus. The scope and rank of the Dorylinae have been in a flux throughout of most of its history. In Emery’s landmark Genera Insectorum (1910) treatment the Dorylinae was a subfamily of the Formicidae consisting of three tribes: ‘Ecitini’ (correct spelling Ecitonini) to accommodate the New World forms and the genus Aenictus, the Dorylini for Dorylus of the Old World, and the Leptanillini for the ants, now considered unrelated, that independently acquired queen morphologies similar to those of certain dorylines. Ecitonines were later treated as a subfamily (Brown 1973, Snelling 1981) and most subsequent authors followed suit (e.g. Gotwald 1982, 1995). Aenictus was first treated as separate from Ecitonini by Borgmeier (1955) who placed them in Aenictini, later given subfamily status by Bolton (1990b). The higher classification of the true army ant clade thus followed a trend of proliferation of subfamilies that culminated in the recognition of four subfamilies: Aenictinae, Aenictogitoninae, Dorylinae, and Ecitoninae (Baroni Urbani et al. 1992). This was reversed only recently with a synonymization of all these names plus Cerapachyinae and Leptanilloidinae under Dorylinae (Brady et al. 2014).

At the genus level, the current classification of the New World army ants was firmly established by Borgmeier in his monographic studies on these taxa (Borgmeier 1953, 1955), where he treated all five genera recognized in the present study. In the Old World, the generic classification has also been stable, with Aenictus and Dorylus both recognized in Emery’s Genera Insectorum (1910). Dorylus has traditionally been divided into a number of subgenera, including the non-monophyletic Anomma to accommodate the above-surface generalist foragers (Kronauer et al. 2007). Because the Old World genus Aenictogiton was known only from males, it has not always been associated with the army ants, but its generic concept has not changed since its original description (Emery 1901b).

The taxonomic history of non-army ant dorylines began with Alfred Russell Wallace collecting a worker specimen later described by Fredrick Smith of the British Museum (Smith 1857) as Cerapachys antennatus. Subsequently Roger (1861, 1862) introduced two new names: Syscia and Ooceraea. Roger considered the former among his ‘Ponera-like ants’ but he gave no discussion of the affinities of the latter. The small postpetiole and large abdominal segment IV characteristic for Ooceraea led Mayr (1865) to include the genus in Myrmicinae. Several other generic names were added later, including Sphinctomyrmex, Cylindromyrmex, Lioponera, Parasyscia, Acanthostichus, and Simopone. Finally, Forel (1893) recognized their affinity and placed all of the mentioned above (including Syscia and Ooceraea) under his tribe ‘Cerapachysii’ within Ponerinae. Wheeler (1902) proposed a classification where he recognized a subfamily Cerapachyinae separate from Dorylinae and Ponerinae, and with three tribes: ‘Acanthostichii’, ‘Cerapachyi’, and ‘Cylindromyrmii’.

The general trend in genus-level taxonomy can be summarized as a progressive addition of new names coupled with increasing number of Cerapachys synonyms. This lumping of names with Cerapachys began with Emery (Emery 1902), who published an early reclassification of the genus. He recognized five subgenera: Cerapachyss. str., Parasyscia, Ooceraea, Syscia, and Cysias. His classification was largely based on the number of antennal segments and abdominal morphology. In the same work he also erected a new genus, Phyracaces, and considered Lioponera as separate from Cerapachys. Emery’s Genera Insectorum (Emery 1911) treatment of the Ponerinae stabilized the classification of cerapachyines under three tribes and with Sphinctomyrmex (with two subgenera), Cerapachys (four subgenera), Phyracaces, Lioponera, Acanthostichus (two subgenera), Cylindromyrmex, and Simopone as constituent genera. Although this classification remained largely unchallenged for years to come, new genus-level names continued to be added, including Procerapachys, Nothosphinctus, Zasphinctus, Chrysapace, Hypocylindromyrmex, Metacylindromyrmex, Aethiopopone, and Neophyracaces. Most of these were introduced as subgenera and in general have not been widely used. Brown’s landmark revision (Brown 1975) established a generic classification that has not been overturned until the present treatment. Brown recognized only four genera and no valid subgenera in his Cerapachyini: Cerapachys (with nine synonyms), Simopone, Sphinctomyrmex (four synonyms), and Leptanilloides. He also recognized Acanthostichus, Ctenopyga, and Cylindromyrmex (three synonyms) as good genera but relegated them to two other tribes, the Acanthostichini and Cylindromyrmecini. Subsequent changes to this scheme included synonymy of Ctenopyga under Acanthostichus (MacKay 1996) and the classification and taxonomy of Leptanilloides and its close relatives (see below). Brown’s treatment of Cerapachys was very broad and although he recognized the morphological disparity of different forms, he preferred informal species groups to subgenera. Much later Xu (2000) described a very distinctive ant from Yunnan, China, for which he introduced a new generic name, Yunodorylus. This short-lived genus was soon also synonymized with Cerapachys (Bolton 2003). Such an all-encompassing definition of Cerapachys made it a polyphyletic taxon, as revealed through recent molecular phylogenetic research (Brady 2003, Brady et al. 2014; discussed below). The most recent advancement in genus-level taxonomy was the study of Bolton and Fisher (2012), focusing on ‘Simopone genus-group’, an assemblage of likely non-monophyletic taxa (Brady et al. 2014) characterized by similar morphologies. This work introduced two new non-army ant genera, Tanipone and Vicinopone.

In 1923 a genus of distinctive, minute and blind ants, Leptanilloides, was first described (Mann 1923) from Bolivia. This taxon was first classified as Dorylinae, Ecitonini and later transferred to its own subfamily (Baroni Urbani et al. 1992, Brandão et al. 1999a). Another genus, Asphinctanilloides, was added to the subfamily and molecular phylogenetics revealed that the male-based name Amyrmex also belonged to this lineage and not to Dolichoderinae where it had been long placed (Ward and Brady 2009). Following additions of multiple new species to Leptanilloides (Longino 2003, Donoso et al. 2006, Borowiec et al. 2011, Silva et al. 2013), Brady et al. (2014) synonymized Leptanilloidinae under Dorylinae. This present revision goes one step further and considers Asphinctanilloides and Amyrmex synonyms of Leptanilloides.

The study of doryline phylogeny and evolution is intertwined with taxonomic considerations, although a summary somewhat independent from the above account, which focuses primarily on the nomenclature, is possible. The work on relationships within the true army ants centers on the controversy of whether they evolved once (the monophyly hypothesis) or more times independently (the polyphyly hypotheses; Gotwald 1979). The affinity of the New World and Old World army ants has been recognized very early on because of similar overall morphologies of males, but this hypothesis of monophyletic origins later became a matter of contention. Certain authors suggested that these lineages were derived independently; perhaps even three times (Brown 1975, Gotwald 1979). This line of thought prompted separating the true army ants into separate subfamilies, as mentioned above. This view persisted until 1990s and in Bolton’s important morphological study (Bolton 1990b), the phylogenetic tree he drew suggested independent evolution of ecitonines and dorylines plus aenictines, with the clade comprising Dorylus and Aenictus more closely related to cerapachyines (considered monophyletic). Later cladistic work, however, supported the monophyletic hypothesis (Baroni Urbani et al. 1992, Brady and Ward 2005). Brady (2003) was the first molecular phylogenetic study to investigate this problem and recovered monophyletic true army ants. This result was recently repeated in a much more comprehensive molecular phylogenetic study, although statistical support was low (Brady et al. 2014). Genomic data, however, reveals that the monophyly hypothesis is likely driven by bias and strongly suggests that New World and Old World army ants evolved independently (see below; Borowiec, in prep.).

Early views on the relationship of the ‘cerapachyines’ to the true army ants were conflicting; these opposing perspectives were succinctly summarized by William Morton Wheeler (1902). He first presented Carlo Emery’s view, which represented an early recognition that these ants are related (Emery 1895b, 1901b), and then contrasted it with the views of Auguste Forel and his own, stating that Cerapachyinae is ‘the most archaic and generalized of existing Formicidae’ and stressing affinities to the Ponerinae. Emery’s early considerations recognize the similarities of doryline and cerapachyine males, which in both groups have retractable genitalia, forked subgenital plate (abdominal sternite IX) and no cerci. Forel (1899, 1901b) and Wheeler (1902), on the other hand, focus on the differences in ‘general habitus’, behavior, and colony sizes, pointing out that cerapachyines appear to have small colonies that are nothing like those of dorylines. Emery seems to have yielded under the pressure from his peers and included cerapachyines in his treatment of the Ponerinae for Genera Insectorum (Emery 1911). In his treatment of the Dorylinae for Genera Insectorum published a year prior (Emery 1910), he remarks: ‘All agree against me that is best to leave the groups of the type of Cerapachys, Acanthostichus, and Cylindromyrmex with the ponerines. I will not insist on being right against the unanimity of my colleagues in myrmecology’. At the same time, Emery managed to preserve an implication of doryline affinity by considering cerapachyines as a part of ‘sectio Prodorylinae’.

Wheeler (1920) later pointed out that Prodorylinae and Cerapachyinae are essentially equivalent and that the latter should be used. At the same time he changed his mind and recognized that cerapachyines and true army ants are related, following his observation of raiding behavior of Lioponera (then Phyracaces) and studies on larval morphology. Despite this, however, the early views of Forel and Wheeler prevailed, and cerapachyine ants were either considered as members of the Ponerinae or a separate group with affinities to the latter. This was where Brown (1975) placed them in his study of the group, although he discusses Emery’s original ideas at some length. It wasn’t until 1990 when Bolton (1990a, 1990b) returned to the ideas of cerapachyine-doryline affinity and, through a careful morphological investigation, convincingly showed that the then recognized subfamilies Aenictinae, Cerapachyinae (including Leptanilloides; see below), Dorylinae, and Ecitoninae form a monophyletic group.

Modern research on the phylogeny of the Dorylinae began with Bolton’s two influential works, as already signaled above (Bolton 1990a, 1990b). Dissections of abdominal sclerites and the morphology of the mesosoma led him to conclude that the ‘cerapachyines’ are distinct from the Ponerinae and he then resurrected Emery’s ideas about their close relationship with the true army ants. These views held to scrutiny with a more formal cladistic analysis of morphological characters conducted by Baroni Urbani et al. (1992), who also included Aenictogiton in their considerations. This landmark study was the first to focus on a formal reconstruction of relationships among all ant subfamilies using a matrix of morphological characters analyzed with maximum parsimony. It recognized a monophyletic grouping of six subfamilies that now constitute the Dorylinae, in subsequent literature referred to as the ‘doryline section’ or the ‘dorylomorph group’. Later research by Seán Brady and Phil Ward (Brady 2003, Brady and Ward 2005) focused on the phylogeny of this clade, with particular emphasis on the true army ants. The 2005 study confirmed a number of synapomorphies for the doryline group of subfamilies originally pointed out by Bolton (1990b), and cemented it as a monophyletic group that can be recognized using morphology as well as molecules. The speciose and morphologically diverse ‘Cerapachys’ was not extensively sampled in these phylogenies, but the several species that were included already suggested that the genus was not monophyletic.

The current limits of the subfamily were not established until the molecular study of Brady et al. published in 2014. As explained above, prior to that, the genera of this group had been classified in as many as six subfamilies. The affinities of those subfamilies had not always been recognized, but a close relationship has since been convincingly demonstrated through a series of independent morphological (Bolton 1990b, Baroni Urbani et al. 1992, Brady and Ward 2005) and molecular phylogenetic studies (Brady 2003, Brady et al. 2006).

The above mentioned study of Brady and colleagues (Brady et al. 2014) was a major advancement of our understanding of doryline phylogeny. These authors sampled representatives of all of Brown’s Cerapachys species groups and 26 out of the 27 extant doryline genera treated here. This work showed that a number of well-circumscribed monophyletic lineages exists within what was called Cerapachys. The phylogeny also shows that these lineages taken together do not form a clade. General patterns recovered in this phylogeny show the importance of geography, with most of the strongly supported monophyletic groupings comprising either Old World or New World lineages. There is moderate support for the monophyly of the true army ants, as well as for a major split between the New World genera (Cheliomyrmex, Eciton, Labidus, Neivamyrmex, Nomamyrmex) and Old World army ants (Aenictus, Aenictogiton, Dorylus), a result also recovered in an earlier molecular phylogeny (Brady 2003). The phylogeny contains many short internodes and a time-calibrated analysis suggests that most of the lineages recognized here as genera diversified in the first 20 million years of the group’s evolution. This apparently rapid initial divergence was perhaps spurred by the transition to preying on brood of other ants (Brady et al. 2014). Following synonymization of the six subfamilies, Brady et al. (2014) did not propose a tribal classification within the group, reflecting the lack of understanding of many relationships.

Recent advances in DNA sequencing techniques provide orders of magnitude more data than has been used in phylogeny reconstruction in previous studies (Brady 2003, Brady et al. 2014), and this approach has been applied to the doryline phylogeny (Borowiec, in prep.). The results reveal that army ant monophyly is likely an artifact driven by systematic bias resulting from model mis-specification. The new findings also show an even stronger signal of broad biogeographic patterns. Even the genomic data, however, do not provide confidence in all relationships at the backbone of the doryline tree. The phylogeny derived from this unpublished genomic data is summarized in Figure 1. All the genera recognized here appear monophyletic and a few well-supported clades that include more than one genus are recovered. These clades include (1) a clade of Chrysapace, Cerapachys, and Yunodorylus, (2) a clade comprising Eusphinctus, Ooceraea, and Syscia, (3) a well-defined clade that was already strongly supported by earlier analyses (Brady et al. 2014), containing Lioponera, Lividopone, Parasyscia, and Zasphinctus, (4) a clade uniting all New World dorylines except for the Central and North American species of Syscia, and (5) Old World army ants clade, which comprises Aenictogiton, Aenictus, and Dorylus. The genera Eburopone, Simopone, Tanipone, and Vicinopone cannot be placed with confidence at any particular position on the doryline tree.

Morphological characters uniting morphologically disparate genera in clade (1) are not obvious, although all these ants currently occur in the Indomalayan region (at least one Chrysapace is known from Eocene Baltic amber). The members of the three genera in clade (2) are mostly soil- and leaf litter-dwelling species and are characterized by small or entirely absent eyes in the worker and reduced antennal segment count in both females (from 12 to 11 or 9) and males (from 13 to 12 or 11). This antennal count reduction is universal across species of this group, although independent reductions occurred elsewhere in the Dorylinae. Clade (3) is well-defined and phylogenetic relationships within it are also very well-supported. The members of this group share the loss of the forewing costal vein, a trait which is present in many other dorylines. Genomic data also strongly suggest a ‘New World Clade’ (4). Within that large clade, the termite hunters Acanthostichus and Cylindromyrmex form are sister genera, and New World army ants (equivalent of the former Ecitoninae) form another well-supported monophyletic group. It is unclear which genus is sister to New World army ants but the data suggest either Leptanilloides or Sphinctomyrmex. Morphological synapomorphies for both the large New World clade and the clade uniting New World army ants with Leptanilloides and Sphinctomyrmex remain elusive. The New World army ants, or Eciton genus-group, which by themselves undoubtedly form a clade, however, possess a highly derived morphology and are well-differentiated from other dorylines. Many of the characters found in New World army ants appear to be independently derived in the Old World army ants clade (clade 5), which comprises Aenictogiton, Aenictus, and Dorylus. Within that group Aenictogiton is sister to Dorylus, forming a clade from which Aenictus apparently diverged a long time ago (Borowiec, in prep.).

A pattern emerges when combining morphological scrutiny of doryline ants with the molecular phylogenetic results: most groupings for which a strong signal of monophyly was recovered in the molecular phylogenies are also easily diagnosed using morphology. This is especially true for all the genera recognized here, but less so for most above-genus groupings. The present revision builds upon this fact and resurrects a number of names hitherto treated as synonyms of Cerapachys or Sphinctomyrmex to better reflect evolutionary relationships. It also introduces two new generic names for taxa previously placed in Cerapachys, and it synonymizes others. The total number of recognized doryline genera is raised from 19 (Brady et al. 2014) to 28. Species previously classified under Cerapachys are here treated in 9 genera, and species of the former Sphinctomyrmex are placed in three genera. I do not propose a tribal classification for the subfamily. As explained above, several lineages cannot be conclusively placed on the backbone of the doryline tree. This means that in order to effectively apply tribal names to monophyletic groups one would need to erect at least four tribes with only one genus each (Eburopone, Simopone, Tanipone, and Vicinopone) and then either erect a single tribe for the New World clade (Figure 1) or further split that group into multiple monogeneric tribes to preserve the distinct ‘Ecitonini’. Here I chose an alternative approach and abandon tribal classification altogether, instead referring to clades grouping multiple genera with informal names such as ‘Old World army ants’ or ‘Eciton genus-group’.

It is worth stating here that the division of these taxa into multiple genera is not motivated solely by the need for monophyletic groupings in modern classifications, but also by the fact that the species hitherto classified under Cerapachys exhibit unusual and confusing variation in morphological characters that have traditionally been used to delimit genera in other ant groups. These characters include the number of palpal segments, number and development of tibial spurs, the development of abdominal segment III (postpetiole), as well as other features. I believe that the revised classification not only better reflects the phylogeny, but also targets for assignment of names those clades that are most morphologically distinct. Reorganization of the ‘Cerapachys’ diversity into more manageable taxa will hopefully stimulate future species-level revisions and permit easy integration of newly discovered forms into this new framework.

The classification proposed here is outlined below. The general distribution and the number of valid described species (excluding subspecies) are given in parentheses. Biogeographic divisions follow those outlined by Cox (2001). Indomalayan region comprises Indian subcontinent and Southeast Asia west of Wallace’s line and the Australasian region includes Australia and Pacific islands east of Wallace’s line.

Dorylinae Leach, 1815 (worldwide, 27 extant genera and 1 extinct genus, 685 described extant and 8 extinct species)

= Acanthostichini Emery, 1901a

= Aenictinae Emery, 1901a

= Aenictogitoninae Ashmead, 1905

= Cerapachyinae Forel, 1893a

= Cheliomyrmecini Wheeler, W. M., 1921

= Cylindromyrmecini Emery, 1901a

= Ecitoninae Forel, 1893a

= Eusphinctinae Clark, 1951

= Leptanilloidinae Baroni Urbani, Bolton & Ward, 1992

= Lioponerini Ashmead, 1905

Acanthostichus Mayr, 1887 (Nearctic, Neotropical, and Dominican amber, 23 extant and1 fossil species)

= Ctenopyga Ashmead, 1906

Aenictogiton Emery, 1901b (Afrotropical, 7 extant species)

Aenictus Shuckard, 1840b (Palearctic, Afrotropical, Indomalayan, and Australasian, 184 extant species)

= Paraenictus Wheeler, W. M., 1929

= Typhlatta Smith, 1857

Cerapachys Smith, F., 1857 (Indomalayan, 5 extant species)

= Ceratopachys Schulz, 1906

Cheliomyrmex Mayr, 1870 (Neotropical, 4 extant species)

Chrysapace Crawley, 1924a, gen. rev. (Malagasy, Indomalayan, and Baltic amber, 3 extant and 1 undescribed fossil species)

Cylindromyrmex Mayr, 1870 (Neotropical and Dominican amber, 10 extant and 3 fossil species)

= Holcoponera Cameron, 1891

= Hypocylindromyrmex Wheeler, W. M., 1924a

= Metacylindromyrmex Wheeler, W. M., 1924a

Dorylus Fabricius, 1793 (Palearctic, Afrotropical, and Indomalayan, 60 extant species)

= Alaopone Emery, 1881, syn. n.

= Anomma Shuckard, 1840c, syn. n.

= Cosmaecetes Spinola, 1851

= Dichthadia Gerstäcker, 1863, syn. n.

= Rhogmus Shuckard, 1840c, syn. n.

= Shuckardia Emery, 1895b

= Sphecomyrmex Schulz, 1906

= Sphegomyrmex Imhoff, 1852

= Typhlopone Westwood, 1839, syn. n.

Eburopone Borowiec, gen. n. (Afrotropical and Malagasy, 1 extant species)

Eciton Latreille, 1804 (Neotropical, 12 extant species)

= Camptognatha Grey, 1832

= Holopone Santschi, 1925

= Mayromyrmex Ashmead, 1905

Eusphinctus Emery, 1893a, gen. rev. (Indomalayan, 2 extant species)

Labidus Jurine, 1807 (Nearctic and Neotropical, 7 extant species)

= Nycteresia Roger, 1861

= Pseudodichthadia André, 1885

Leptanilloides Mann, 1923 (Nearctic and Neotropical, 19 extant species)

= Amyrmex Kusnezov, 1953, syn. n.

= Asphinctanilloides Brandão, Diniz, Agosti & Delabie, 1999, syn. n.

Lioponera Mayr, 1879, gen. rev. (Palearctic, Afrotropical, Malagasy, Indomalayan, and Australasian, 73 extant species)

= Neophyracaces Clark, 1941, syn. n.

= Phyracaces Emery, 1902, syn. n.

Lividopone Fisher and Bolton, 2016 (Malagasy, 1 extant species)

Neivamyrmex Borgmeier, 1940 (Nearctic, Neotropical, and Dominican amber, 127 extant and 1 fossil species)

= Acamatus Emery, 1894

= Woitkowskia Enzmann, 1952

Neocerapachys Borowiec, gen. n. (Neotropical, 2 extant species)

Nomamyrmex Borgmeier, 1936 (Nearctic and Neotropical, 2 extant species)

Ooceraea Roger, 1862, gen. rev. (Pantropical; native in Indomalayan and Australasian, 11 extant species)

= Cysias Emery, 1902, syn. n.

Parasyscia Emery, 1882, gen. rev. (Palearctic, Afrotropical, Malagasy, Indomalayan, and Australasian, 50 extant species)

†Procerapachys Wheeler, W. M., 1915b (Baltic amber, 3 fossil species)

Simopone Forel, 1891 (Afrotropical, Malagasy, Indomalayan, and Australasian, 39 extant species)

Sphinctomyrmex Mayr, 1866b (Neotropical, 3 extant species)

Syscia Roger, 1861, gen. rev. (Nearctic, Neotropical, and Indomalayan, 5 extant species)

Tanipone Bolton & Fisher, 2012 (Malagasy, 10 extant species)

Vicinopone Bolton and Fisher, 2012 (Afrotropical, 1 extant species)

Yunodorylus Xu, 2000b, gen. rev. (Indomalayan, 4 extant species)

Zasphinctus Wheeler, W. M., 1918, gen. rev. (Afrotropical and Australasian, 20 extant species)

= Aethiopopone Santschi, 1930, syn. n.

= Nothosphinctus Wheeler, W. M., 1918, syn. n.

The Dorylinae possess a number of characteristic morphological traits, which are discussed in detail below. A

Hymenoptera Anatomy Ontology

An illustration of morphological characters used in this revision is presented in Figures 2–4. Below I discuss these characters in hope that this will help to clarify the terminology and facilitate identification.

Doryline ants are characterized by their predation on other social insects, a condition that appears to be apomorphic for the group, but there is also a plethora of morphological characters that distinguish them from other ant lineages. Extensive work grounded in examination of morphology has been done to infer ant phylogeny (Gotwald 1969, Ward 1990, Baroni Urbani et al. 1992, Bolton 2003, Keller 2011), and the evolution of the doryline clade in particular (Bolton 1990a, 1990b, Brady and Ward 2005, Barry Bolton’s unpublished work, this study). As a result, multiple morphological characters have been identified as diagnostic or possibly synapomorphic for the Dorylinae. The list and discussions below are adapted and updated from the works cited above. Likely synapomorphies are in italics.

1. Lateral area of clypeus very narrow in full face view; distance from anterior clypeal margin to paraoculoclypeal sulcus greater than distance from anterior clypeal margin to frontoclypeal sulcus where antennae insert.

The clypeal area of the head capsule (Figure 2) is medially delimited by the so-called frontoclypeal sulcus that laterally extends from the midline of the head to the front of the antennal socket area. Lateral to the antennae, the boundary delimiting the clypeus from the rest of the head capsule has been referred to as the paraoculoclypeal sulcus (Keller 2011). In most ants, the distance from the anterior margin of the head capsule (including clypeus) to the frontoclypeal sulcus in the area where the antennae attach is greater than the distance from the margin to the paraoculoclypeal sulcus. In the dorylines the clypeus is very narrow and this condition is reversed. Certain Cylindromyrmex and Acanthostichus species have the clypeal sulci poorly visible and the medial area of clypeus is protruding forward over the mandibles, thus creating a considerable distance from the antennal socket area to the anterior margin of head capsule. The distance from the anterior margin of the torulo-posttorular complex (see below) to the anterior margin of head, however, is always short and thus the clypeus can be still considered laterally narrow in these species.

A relatively narrow clypeus is present in several ant subfamilies, but the condition described above appears to be restricted to the Dorylinae, Martialis, Leptanillinae, Amblyoponinae, and Proceratiinae. I suspect that a thorough study of the clypeal area in the dorylines may reveal new characters of diagnostic or phylogenetic utility within the subfamily.

2. Parafrontal ridges present, i.e. genae carinate laterally of antennal sockets.

Another feature characteristic and likely synapomorphic for the Dorylinae is the presence of often prominent ridges extending some distance back from the paraoculoclypeal sulcus (when visible), laterally to the antennal socket (Figure 2). This feature is present in most doryline species, although it has been reduced several times, for example in Acanthostichus (Figure 5) and Cylindromyrmex, Dorylus, and in some species of Aenictus, Simopone and Zasphinctus. This character was considered a synapomorphy of the Cerapachyinae (Brown 1975, Bolton 2003), even though it is well developed in many New World true army ants.

3. Torulo-posttorular complex present, often vertical, occasionally horizontal; antennal sockets exposed or (more rarely) partially concealed by torulo-posttorular complex in full-face view.

Keller (2011) provided much needed clarification of the nomenclature relating to the structures in the antennal socket area in ants. He differentiated frontal carinae from the medial arch of torulus (both structures that arise medially from the antennal socket) and recognized that the relative development and configuration of these structures vary considerably among ants. In the dorylines, the frontal carinae sensu Keller have been interpreted to be present only in their posterior part (called posttorular flanges) and fused to the torular lobe (an expansion of the above-mentioned medial arch of the torulus). This whole structure was named torulo-posttorular complex (Keller 2011). This complex in the dorylines can be vertical and thus forming lobes that in full-face view project towards the observer without obscuring the antennal socket area (Figure 2) or, more rarely (Acanthostichus, Cylindromyrmex, Simopone), horizontal or expanded laterally to conceal the antennal sockets. The distinction among the two conditions is not entirely clear-cut however, as multiple lineages or species of the dorylines possess poorly developed lateral expansions of the torulo-posttorular complex that may conceal parts of the medial area of antennal sockets.

The vertical configuration of the torulo-posttorular complex is characteristic and probably synapomorphic to the Dorylinae. A similar morphology is present in some proceratiines, in particular in Probolomyrmex. Similar configuration exists in the Leptanillinae and in Apomyrma (Amblyoponinae). A horizontally expanded torulo-posttorular complex is also found in the Amblyoponinae (Myopopone).

4. Stipes of maxilla sharply divided into proximal and distal faces, with proximal face extending beyond inner margin of stipes; prementum concealed when mouthparts closed.

The major sclerite of the maxilla, the stipes, is in dorylines sharply divided into a raised proximal face and sunken distal face on its outer surface (Gotwald 1969). The distal face of the stipes accommodates the labrum when the mouthparts are closed and the proximal face is medially expanded beyond the inner margin of the sclerite. When the mouthparts are closed, this condition causes the prementum (the major sclerite of the labium) to be concealed behind the labrum and the medial extension of the proximal face of the stipes. In most worker ants the stipes is not divided into two distinct faces and prementum is visible when mouthparts are closed, although a carina dividing the stipes is present in numerous taxa. Aenictus is an exception among dorylines, as the maxilla in this genus is not divided into two faces and the prementum is visible with mouthparts fully closed. The prementum can also be seen in Cheliomyrmex, where the division of the stipital surface is weakly marked. The division of maxilla into two faces is a likely synapomorphy of the Dorylinae, secondarily lost in Aenictus.

A similar condition has apparently independently evolved in some Amblyoponinae where the prementum is also concealed (Keller 2011).

5. Eyes frequently reduced or absent.

Eyes are poorly developed in most species of the doryline workers, although large and multifaceted eyes are present in a number of lineages. A few speciose lineages lack the eyes completely and without exception in any of the species (e.g. Aenictus, Dorylus), while many other genera are either blind in most species or with very small eyes present (e.g. Acanthostichus, Eciton, Neivamyrmex, Syscia, Ooceraea). Large worker eyes can be present in some or all species of Cerapachys, Chrysapace, Cylindromyrmex, Lioponera, Lividopone, Simopone, Tanipone, and Vicinopone. Many of the species with large eyes are arboreal, but others are surface-foragers or their natural history is unknown.

6. Mesosoma internally with fused meso- and metafurcal arms, externally corresponding to an endophragmal pit.

The mesosoma of the Dorylinae worker ants possess a pit in the cuticle, located anteriorly to the propodeal spiracle and near mesometapleural Pronotomesopleural suture (where present; Figures 2 and 14A). This pit, known as the mesosomal endophragmal pit, interiorly corresponds to a junction between cuticular projections that serve as muscle attachments (apodemes) and the lateral wall of the mesosoma. The endophragmal pit may not be discernable with light microscopy in species where it is not surrounded by a cuticular concavity, especially in smaller species. In most ants, the apodemes of the second and third thoracic sternites are separate and bifurcated, and called mesofurca and metafurca, respectively. The mesofurca and metafurca thus form two U- or V-shaped structures inside the mesosoma, with their bases directed posteriorly, and the opening of the ‘V’ directed anteriorly. Some distance from their base, the arms of the mesofurca are additionally connected by a transverse bar of cuticle, the mesofurcal bridge. A variation of this scheme occurs when metafurcal arms first diverge and then fuse together; in some ants the mesofurcal arms point sideways and the mesofurcal bridge forms a triangle but mesofurca and metafurca still separate. In the dorylines, this condition appears to be much modified, with the arms of the mesofurca directed laterally instead of anteriorly and fused to the anteriorly-pointing arms of the metafurca. The laterally projecting fused arms of mesofurca and metafurca then attach to the lateral wall of the mesosoma. In larger species this manifests itself as the mesosomal endophragmal pit. As a result of this modification, in the dorylines the ordinarily transverse mesofurcal bridge points backwards to reach the transverse mesofurcal arms. The character system of mesosomal apodemes has not been methodically investigated in ants but Vilhelmsen et al. (2010) studied these structures in other aculeates.

7. Metapleural gland orifice concealed beneath a ventrally directed cuticular flap or flange.

The metapleural gland is a feature found exclusively in ants and is believed to aid in colony sanitation (Yek and Mueller 2011). The gland is internally located in a cuticular chamber at the junction of the metathorax and the propodeum and it opens externally on either side of the mesosoma at a variable distances from the propodeal declivity and above the hind coxae. This metapleural gland orifice and cuticle surrounding it exhibit various modifications in the Formicidae. In the dorylines, the orifice is not visible in dorsal or lateral view because it is overhung by a dorsal flap of cuticle (Figure 2). A similar flap has been reported in Simopelta in the Ponerinae (Keller 2011). A dorsal flange also occurs in the Leptanillinae, but the orifice can still be seen in lateral view.

I suspect that a careful study focused on the structures surrounding metapleural gland orifice would reveal additional genus-level diagnostic characters in the Dorylinae.

8. Helcium sternite bulging ventrally and articulated on the inner wall of the tergite.

The helcium is a term used for the presclerites of abdominal segment III. The relative development and place of articulation of the helcial sternite varies in ants. The dorylines exhibit a rare condition where the sternite is well developed and bulging ventrally, not obscured by the tergite in lateral view. The sternite is also articulated to the tergite some distance up on the inner wall of the latter, so that the tergite overlaps the sternite. A ventrally bulging helcial sternite appears to occur only in some Proceratiinae, Tatuidris, and in the Myrmicinae. In the myrmicine ants, however, the articulation of the sternite and the tergite is along the lateral margins, and thus the sclerites are not overlapping. In all other ants the helcial sternite is flat or only slightly convex, not readily visible in lateral view.

9. Abdominal segment III with complete tergosternal fusion.

The degree of fusion of sternites to the tergites of abdominal segments varies in ants (Kusnezov 1955). Bolton (1990b) examined numerous representatives of the Dorylinae (‘the doryline section’) and concluded that in general, the tergite of the abdominal segment III is fused to the sternite in workers and gynes of these ants, although the condition varies in males. Keller (2011) refined this character system by distinguishing whether it is presclerites (helcium) or postsclerites, or both, that are fused. The complete fusion in dorylines is also present in most poneroid ants (most Amblyoponinae, Ponerinae, Proceratiinae) and possibly in Martialis. In the formicoid clade, the fusion occurs in Ectatomminae and in Heteroponerinae. In other subfamilies, either the postsclerites or both pre- and postsclerites of abdominal segment III are unfused.

10. Abdominal segment IV without tergosternal fusion.

Complete fusion of tergites and sternites of abdominal segment IV is rare in ants and apparently restricted to Agroecomyrmecinae, Ponerinae, and Proceratiinae. In other subfamilies the sclerites are unfused or only presclerites exhibit fusion (Bolton 1990b, 2003).

11. Spiracles of abdominal segments V–VII shifted posteriorly on each segment, not concealed by the posterior margin of the preceding tergite and visible without distension or dissection.

This is a character that is a likely synapomorphy of the subfamily and does not appear to occur in any other ants. In most ants the spiracles of abdominal segments I (the propodeum) through IV are visible, but those of abdominal segments V, VI, and VII are ordinarily concealed by the posttergites of their respective preceding segments. These spiracles cannot be thus seen without distension or dissection of the gaster. In the dorylines, however, the spiracles are shifted posteriorly on the posttergites and visible in specimens without any manipulation (Figures 2, 5A). The exposed spiracles are obvious even in species where the size of the distal abdominal segments has been substantially reduced, such as in Ooceraea.

12. Pygidium modified: either large and with dorsum flattened and armed with teeth or spines, or reduced to a narrow V-shaped sclerite.

In general, the pygidium (last visible abdominal tergite) is derived in the dorylines, departing from a condition of a large, evenly rounded sclerite that is presumed to be plesiomorphic for ants. The degree and nature of the modification varies, however. In many genera previously classified under the Cerapachyinae the pygidium has a flattened medial area and is armed with thick, specialized setae that are thought to have sensory function (Hölldobler 1982). Occasionally the tip of pygidium is also forked, with a prong of variable length along each side of the sting. In Dorylus the pygidium also has a flattened disc but it is never armed with numerous modified setae, instead possessing two to four cuticular spines that are tipped with one thick seta each. In Aenictus, Eciton and other New World army ants, as well as in Leptanilloides, the pygidium is modified into a very narrow transverse sclerite that is not armed with any modified spines or setae. A pair or two of thick setae can be, however, present in some New World army ants. Worker Aenictogiton is an exception among the Dorylinae, as the pygidium in this genus is large and simple, evenly rounded and without armament of cuticular projections or spine-like setae.

13. Sting apparatus with furcula fused to base of sting or absent in most species.

The furcula is a Y- or wishbone-shaped sclerite flexibly attached at the base of the sting (Snodgrass 1984), present in most ants but apparently fused to the sting base in all Dorylinae examined thus far except Leptanilloides (Brandão et al. 1999a). In the Aneuretinae, Dolichoderinae, and Formicinae the furcula is also reduced or fused and in Simopelta, a ponerine genus with army ant-like habits, this sclerite is also fused to the sting base. The furcula serves as the attachment for muscles responsible for protraction of the sting (Hermann and Chao 1983). These muscles connect directly to the anterior region of the sting bulb in the dorylines where furcula was observed to be fused with the sting (Hermann and Blum 1967, Hermann 1969). It is unclear whether the fusion of furcula and sting is a synapomorphy of the Dorylinae that has reverted to the unfused state in Leptanilloides or whether the fusion evolved more than once in the subfamily. A comprehensive and comparative study of the doryline sting is lacking. The sting apparatus appears generally functional and capable of piercing human skin, as in Eciton, but more rarely it can be non-functional as a weapon, as in Dorylus (Hermann 1969, Bolton 1990b). The sting been described in varying detail in the army ant genera Aenictus, Cheliomyrmex, Dorylus, Eciton, Labidus, Neivamyrmex, and Nomamyrmex (Hermann and Blum 1967, Hermann 1969) and in a few non-army ant species including Leptanilloides (Brandão et al. 1999a), Acanthostichus, Lioponera, Syscia, and Zasphinctus (Hermann 1969).

14. Metacoxal cavities fully closed, without a Pronotomesopleural suture in the broad annulus.

The morphology of the cuticle surrounding the sockets where hind coxae articulate (the coxal cavities) varies among ants. The primitive condition is presumably one where the cavities are not fully surrounded by the cuticle (the annulus) and the cavities are connected to the petiolar foramen. This condition is present in Myrmeciinae, Aneuretinae, Platythyrea in the Ponerinae, and Ectatomminae. A modification of this state occurs where the annulus surrounds the cavities, so that the cuticle is continuous around the openings, although a Pronotomesopleural suture can be discerned where the outgrowths of the cuticle closing the foramen meet. This state is present in some Amblyoponinae, most Ponerinae, some Heteroponerinae, Paraponera, and some Proceratiinae. Finally, the cuticle can be entirely fused and no Pronotomesopleural suture is visible in the cuticle surrounding coxal cavities. This is the condition observed in the Dorylinae. This character state is common among the subfamilies of the formicoid clade, including Dolichoderinae, Formicinae, Myrmicinae, Pseudomyrmecinae, and some Heteroponerinae (Bolton 2003). The cavities are also fully closed and without a Pronotomesopleural suture in some taxa outside the formicoids, namely in the Leptanillinae, Martialis, some Amblyoponinae, Agroecomyrmecinae, few Ponerinae, and some Proceratiinae.

15. Metatibial gland present, located distally on the ventral surface of hind tibia.

Multiple ant lineages have been found to possess a glandular structure on the ventral (flexor) surface of their hind tibiae (Hölldobler et al. 1996). Given our current understanding of ant phylogeny (e.g. Brady et al. 2006), it is likely that these metatibial glands have evolved more than once. Bolton (1990b) recognized the presence of this gland as characteristic of most dorylines and described the variations in its external manifestation within the subfamily. In some taxa the presence of metatibial gland can be detected only with histological examination, and externally its visibility in very small-bodied species can sometimes be confirmed only through scanning electron microscopy (Borowiec and Longino 2011). The gland is externally visible in most doryline genera and is conspicuous in most true army ants except Nomamyrmex, where it is entirely absent, at least externally. Other genera apparently lacking obvious metatibial glands are Chrysapace, Eusphinctus, Lividopone, Simopone, Sphinctomyrmex, Tanipone, and Vicinopone. The metatibial gland is also apparently absent in a few species of Acanthostichus, Lioponera, Neivamyrmex, and Zasphinctus. Most Simopone possess a well-developed gland on the inner surface of hind basitarsus but no gland on the tibia. Syscia possess both the metatibial and metabasitarsal glands, the latter of different appearance and apparently independent origin from the one found in Simopone.

Glands on the hind tibiae also occur in the Ponerinae (Schmidt and Shattuck 2014), as well as in a few Amblyoponinae and Myrmicinae, although in all these cases the glandular surfaces are positioned differently than in the Dorylinae, suggesting convergence.

As mentioned above, several Dorylinae genera are lacking externally visible metatibial glands and there has been no comprehensive histological study of all the lineages. Given this, coupled with unresolved relationships among most lineages of the subfamily, it is somewhat uncertain whether this character is a synapomorphy of the whole clade or if it is primitively absent from early-branching lineages.

Despite its presence in some of the better-studied species (true army ants, O. biroi), the function of the gland in the Dorylinae is unknown (Billen 2009).

The worker Dorylinae can be thus easily recognized through a combination of metapleural gland orifice concealed by a dorsal cuticular flap, large and convex sternite of the helcium, and exposed abdominal spiracles of segments V–VII.

1. Abdominal sternite IX (hypopygium) modified, bidentate to biaculeate.

The appearance of the abdominal sternite IX of the male is distinctive in the Dorylinae. A simple sclerite with convex or medially tapered posterior margin appears to be the plesiomorphic condition, present in most ants. In the dorylines the hypopygium often has a convex posterior margin and is laterally drawn into two processes, ranging from blunt triangular denticles to long, parallel prongs. Occasionally further modifications, including folds, excisions, and additional teeth are also present on the hypopygium. In some lineages the male hypopygium is useful for species identification. Leptanilloides is again the exception, and the sclerite in this lineage is relatively simple, sometimes medially convex or concave. Outside of the Dorylinae, a biaculeate male hypopygium is present in at least two genera, Paraponera and Nothomyrmecia.

2. Cerci absent from male genitalia.

The cerci (also called pygostyles) are paired sensory structures articulating with the last abdominal tergite. Most male ants have cerci but their loss has occurred several times independently, in Leptanillinae, Martialis, some Amblyoponinae, and some Proceratiinae (Boudinot 2015). The Dorylinae is the largest clade of ants where male cerci appear to be absent from all species. This character was important for the early recognition of a close relationship of Acanthostichus and other ‘cerapachyines’ with the true army ants (Emery 1895).

3. Genitalia completely retractile.

The doryline males are able to retract their genital capsule into the abdomen. In other ant taxa, even those that also lack cerci, the genitalia cannot be completely retracted. The genitalia of true army ants are always well-concealed in dead specimens and not visible without dissection. The genitalia of other, particularly smaller dorylines, however, can be partially visible in dead males. The small-sized males of Leptanilloides may be an exception among the dorylines, as the known specimens have exerted genitalia, with most of the genital capsule visible without artificial distension or dissection of the abdomen. The abdomen of some Leptanilloides species also appears too small to allow for full retraction of the genital capsule.

4. Jugal lobe absent from hindwing.

The jugal lobe is a basally located projection of the wing membrane. Bolton (2003) recognized that its presence varies in the hind wing of alate ants. He cites this character as highly polymorphic in the poneromorph subfamilies, absent in the Leptanillinae and the formicoid subfamilies.

The male Dorylinae can be thus recognized by the lack cerci, almost universally bispinose hypopygium, and retractable genital capsule. The last two characters do not apply to Leptanilloides, but the cerci are lacking in this genus, too. Leptanilloides also has extremely reduced tegulae or is lacking them entirely, a loss perhaps unique among male ants.

Doryline gynes share many worker characteristics and can be recognized by the same putative synapomorphies as the worker, except perhaps for the highly specialized ‘dichthadiigyne’ queens of the true army ants. The latter may be difficult to distinguish from convergently evolved specialized queens of Leptanilla (Leptanillinae; Baroni Urbani 1977), Onychomyrmex (Amblyoponinae; Wheeler 1916) or Simopelta (Ponerinae; Gotwald and Brown 1967). Perhaps the best single character to identify dichthadiigynes as Dorylinae is the presence of posteriorly shifted abdominal spiracles V-VII, visible on the gaster without distension or dissection of abdominal sclerites. For further discussion of gyne morphology see ‘Characters used to describe gyne morphology’ below.

Number of antennal segments. The ant antenna includes only three ‘true’ segments (scape, pedicel, and funiculus), that is metameric structures connected to other segments via muscles, with funiculus further subdivided into secondary structures. Together, the scape, pedicel, and funicular segments are thus sometimes referred to as antennomeres. In the taxonomic literature concerning ants, however, use of ‘antennal segments’ instead of ‘antennomeres’ is widespread and I follow this convention here. In the Dorylinae, the plesiomorphic condition is 12-segmented antennae (Figure 12A) but a number of lineages underwent reduction, either to 11 segments (Eusphinctus, many Ooceraea, few Parasyscia, Simopone, many Syscia, some Yunodorylus, some Zasphinctus), 10 segments (most Aenictus, at least one Ooceraea; Figure 9A), or 8–9 segments (few Aenictus, some Ooceraea, some Syscia). In Dorylus the number of antennal segments may vary among individuals from the same colony (Eguchi et al. 2014).

Relative size of the apical antennal segment. The size of the last (apical or terminal) antennal segment varies widely within the Dorylinae, both in length and width relative to other segments. The apical segment ranges from small in Simopone (Figure 49A) to swollen, much wider than the penultimate segment and longer than several preceding segments together in some Parasyscia (Figure 48). This character may potentially aid in genus-level identification but high intrageneric variation and continuous nature of this trait make it less suitable for a dichotomous key.

Cuticular apron of the clypeus. Many species in various genera possess a semi-translucent to opaque cuticular projection, or lamella, that arises from the clypeus and closes the gap between mandibles and the head capsule. This trait varies among and within genera.

Lateroclypeal teeth. Many dorylines possess cuticular projections that are arising from lateral portions of the clypeus and overhang mandibles (Figure 26B). These projections can be of various sizes and shapes, most commonly finger-like or triangular. This trait varies among and within genera.

Parafrontal ridges. As discussed above under worker diagnostic character 2, this is one of distinguishing characters of the Dorylinae worker. A few lineages such as Acanthostichus and Dorylus seem to lack this trait completely, although reductions of various degree occurred in several genera (see also above under Diagnostic characters of the worker).

Torulo-posttorular complex. Another defining feature of the Dorylinae, this character is discussed above under worker diagnostic character 3.

Antennal scrobes. Depressions of the cuticle that receive retracted antennal scapes are uncommon in the Dorylinae, well developed only in Cylindromyrmex and some species of Simopone. Although not considered scrobes here, feebly marked depressions that apparently receive antennal scape can be seen in certain species of Parasyscia and Lividopone.

Labrum shape. Most dorylines have a labrum that is notched medially on its distal (non-articulated) margin, although in a few (Aenictogiton, some Dorylus, Leptanilloides) the margin is evenly rounded across.

Proximal face of stipes. The stipites concealing the prementum are another trait of the subfamily that is discussed above under worker diagnostic character 4.

Number of maxillary palp segments. This character varies from the plesiomorphic number of six segments in Tanipone and most Simopone to only one segment in Aenictogiton. Although the palpal segment count is apparently constant throughout some genera, its reduction appears to often correlate with small size. As such, and because this character is often impossible to see without dissection, it is of limited utility for genus-level identification.

Number of labial palp segments. The discussion regarding maxillary palp segmentation applies to labial palps as well. The labial palps are never composed of more than four segments and as a rule have fewer segments than maxillary palps of the same individual. Exceptions to this rule are the New World army ants and Acanthostichus, where the labial palps are longer than maxillary palps, with three and two segments, respectively. In most Dorylus both maxillary palps are 2- or 1-segmented and labial palps are 2-segmented but the labial palps are more slender and longer than the maxillary palps. Taken together, the number of maxillary and labial palp segments is sometimes expressed as ‘palp formula’ which simply gives the two numbers separated by a comma. For example, palp formula 4,3 means the maxillary palps are 4-segmented and labial palps are 3-segmented.

Mandible shape and dentition. Shape of the mandibles varies across the Dorylinae, with many species retaining plesiomorphic triangular mandibles with well-differentiated basal and masticatory margins and numerous denticles of uniform shape on the latter (Figure 18B). Some dorylines, however, possess falcate mandibles where the distinction between basal and masticatory margins is blurred and a few large, one, or no teeth are present in addition to a pointed mandibular apex (Figure 14B). Derived mandibles are characteristic of lineages with pronounced worker size polymorphism, being especially conspicuous in army ant genera such as Dorylus or Eciton.

Eye size. In general, eyes are small or absent in dorylines, although exceptions do occur, as discussed under worker diagnostic character 5 above.

Presence of ocelli. The ocelli are rare in worker dorylines, although Chrysapace, Simopone (Figure 49B), and Tanipone species always possess them. Ocelli also occur in workers of some species of Cylindromyrmex and Lioponera (Figure 2). Because ergatoid (wingless and worker-like) queens and intercastes are common in the Dorylinae, presence of ocelli can be mistakenly inferred for the worker caste when examining a very worker-like gyne.

Head capsule above occipital foramen. Dorylines vary in the degree of differentiation between dorsal (or frontal) and posterior (or occipital) faces of the head. Most species have a distinct posterior surface of the head just anterior to the attachment with the mesosoma. In Simopone and Vicinopone (Figure 57A) there is no posterior surface and the head is seemingly ‘hanging’ by the dorsal face immediately anterior to occipital carina (see below). Some Aenictus and Neivamyrmex army ants have a very gradual transition to the posterior face, thus achieving similar appearance (Figures 11, 39).

Carina on ventrolateral surface of head. Ventrally on the head of most dorylines, a carina surrounding occipital foramen can be found (see below). Additionally, in some species there may be additional ridges that originate at the lateral corners of that ventral occipital carina and run partways or down the length of ventral head surface towards mandibular insertions (Figure 2). These ridges do not appear to be universally present in any genus except perhaps Neocerapachys but can be found in many Lioponera, Lividopone, and Parasyscia.

Carina surrounding occipital foramen. Many dorylines possess a carina around the occipital foramen. In some species, this carina joins ventrally to separate the area immediately anterior to the occipital foramen from the rest of the head capsule. This character is variable within and among genera.

Pronotal flange delimited by a ridge. The sloping surface of the mesosoma immediately behind the occipital articulation is known as the pronotal flange. This area can be evenly rounding into the pronotal dorsum, also called the pronotal neck, or separated from it by a variously developed cuticular margin. This feature can be consistently present within certain genera like in Cerapachys or Eburopone (Figure 22A), while in others, such as Ooceraea or Parasyscia, it occurs only in some species. In general, this trait tends to be reduced in small species.

Promesonotal connection. The connection between the first mesosomal notum, the pronotum, and the rest of the mesosoma is variously developed in ants. The pronotum is dorsally adjacent to the mesonotum and laterally to the mesopleuron. The connection can be fully articulated as in most Formicinae where a well-developed Pronotomesopleural suture is present, or rigidly fused as in the Myrmicinae where the entire mesosoma forms a single rigid block and usually there is no trace of Pronotomesopleural suture dorsally. Among the dorylines both of the above mentioned conditions can be found, along with intergradations. A completely unfused and mobile connection is present only in certain Leptanilloides, while in Dorylus, for example, a conspicuous Pronotomesopleural suture is present but the connection is not mobile. In others, only lateral portions of the Pronotomesopleural suture are unfused (see next character below) or, as is the case in Parasyscia, the connection is fully fused with no trace of Pronotomesopleural suture.

Pronotomesopleural suture. In the Dorylinae, the promesonotum and mesopleuron are often linked by a Pronotomesopleural suture (sometimes termed the 'promesopleural' Pronotomesopleural suture). This character is linked to the preceding one but it is treated separately because in many genera the Pronotomesopleural suture may be completely fused dorsally, at the same time being unfused laterally on the mesosoma (Figure 12A). A completely fused pronotomesopleural Pronotomesopleural suture is characteristic of Lividopone, Parasyscia, and Zasphinctus. In the army ants the lateral wall of the mesosoma may be dorsoventrally compressed around this area. As a result, the Pronotomesopleural suture is generally short, directed more posteriorly rather than dorsally as in other dorylines, or entirely fused.

Mesometapleural groove. Directly posterior to the pronotomesopleural Pronotomesopleural suture (or the area where the Pronotomesopleural suture would be found) is the mesopleuron. This area can be delimited posteriorly from the succeeding sclerite, the metapleuron, by a groove (Figure 12A) or it can be elevated relative to the metapleuron so that a cuticular ridge is formed at the boundary. This character is somewhat difficult to quantify but most dorylines show a division between the meso- and metapleura. The separation tends to be least pronounced in Aenictus and most New World army ants.

Transverse groove dividing mesopleuron. The mesopleuron can be undivided or separated into two parts by a Pronotomesopleural suture or a cuticular ridge (Figure 2). If divided, the upper part is usually referred to as anepisternum and the lower as the katepisternum. This is another character that is rather variable in the dorylines, often within a genus. This division is generally absent from Aenictus and New World army ants.

Concavity surrounding pleural endophragmal pit. As explained in the discussion of the internal mesosomal structure above, an endophragmal pit is an impression in the cuticle that corresponds to invaginations of the cuticle. An examination of the pit itself often requires scanning electron microscope but the pit can be surrounded by a variously pronounced concavity of the cuticle, making it more easily discernable. When visible, the concavity is usually placed some distance anterior to and/or below the propodeal spiracle.

Margination of various body segments. This category encompasses characters that include dorsolateral margination of the mesosoma, petiole, or other segments. Several genera have some form of margination at the junction of the lateral and dorsal faces of the mesosoma or above the petiolar spiracle. The most characteristic margination of dorsolateral corners of the body is present in Lioponera, where it can range from being confined to the anterior half of abdominal segment II (petiole) to well-defined margins present across the posterior half of the head, most of mesosoma, posteriorly to abdominal segment IV (Figure 2).

Metanotal depression or groove on mesosoma. Dorsally the mesosoma may possess a groove or depression marking the division between the thorax and the propodeum (which is anatomically the first abdominal segment). Most dorylines have no such groove, but in Aenictus and New World army ants this distinction is usually pronounced (Figures 11, 14A).

Propodeal spiracle position. The position of propodeal spiracle on the lateral wall of the mesosoma can serve as a good character distinguishing army ants from other dorylines. When inspected in lateral view, the spiracle opening is almost always at or below the midheight of the mesosoma in the genera that are not considered army ants (Figure 12A) and well above in the army ants (i.e. in Aenictogiton, Aenictus, Cheliomyrmex, Dorylus, Eciton, Labidus, Neivamyrmex, or Nomamyrmex; Figure 20A). Rarely the spiracle is positioned slightly above the midheight in non-army ant dorylines but then the pygidium is usually armed with numerous peg-like setae (see below).

Shape and margination of propodeal declivity. The propodeum has a sloping or vertical posterior face that can be variously shaped and dorsally immarginate or with a distinct margin, bound by a carina. The most common shape of the doryline propodeum when viewed from behind is approximately rectangular. However, in Aenictus a more triangular shape is common. The dorsal margination appears to be more common in certain genera then in others, although there is often variability within genus.

Metapleural gland bulla. The metapleural gland is positioned at the posterior end of the metapleuron, on the lateral mesosoma directly above where the hind coxa articulates. The gland opens through the cuticle and has a chamber, or bulla, situated below the opening. The metapleural gland orifice is further discussed above under diagnostic characteristics. In the descriptions I indicate whether the gland bulla is visible through (Figure 5A) or obscured by the cuticle (Figure 16A).

Propodeal lobes. Propodeal lobes are projections of the cuticle on the propodeum, arising immediately lateral to the propodeal foramen, the opening in the mesosoma where abdominal segment II (petiole) articulates. These lobes are variously developed in the Dorylinae. They are completely absent in Aenictogiton, Dorylus, and Leptanilloides and absent or short in the Eciton genus-group. In other genera the lobes are well-developed and visible laterally as semicircular extensions of the propodeum projecting past the metapleural gland bulla (Figure 2).

Position of helcium. Two major portions can be distinguished in each abdominal segment starting with segment II (the petiole; the propodeum corresponds to segment I): presclerites and postsclerites. Presclerites form the portion that articulates with the preceding segment. Postsclerites constitute the part of the segment that is always exposed without dissection or extension of gastral sclerites. The helcium comprises the presclerites of abdominal segment III, that is its anterior portion that articulates with the petiole. The helcium can be positioned at or above the tergosternal Pronotomesopleural suture of segment III. The position of helcium can also be described relative to the midheight of postsclerites of abdominal segment III in lateral view. Axial helcium means positioned at the midheight, infraaxial below, and supraaxial above. The helcium can thus be positioned at the tergosternal Pronotomesopleural suture and at the same time be considered infraaxial if the Pronotomesopleural suture occurs below the midheight of the segment. In most dorylines the helcium is placed at the Pronotomesopleural suture and more or less axially. The most obvious departure from this state is when the helcium is in a supraaxial position and above the Pronotomesopleural suture (Figure 35A). This means that there is a narrow posterior face of the petiolar and narrow anterior face of the abdominal posttergite III. A supraaxial helcium also tends to have larger circumference than an axial helcium. This condition can be seen in Acanthostichus, Cerapachys, Lividopone, and at least one Yunodorylus. In some Leptanilloides the helcium is positioned supraaxially because abdominal sternite III is very large but in this case there is still a defined posterior face to the petiole and anterior face of the sternite III. For brevity, a helcium with small circumference is referred to as ‘narrow’ and a helcium with large circumference is referred to as ‘broad’ in the diagnoses.

Prora. Prora is used to describe a protrusion on the anterior face of abdominal poststernite III, below the helcium. In the dorylines this feature is variously developed, as a simple angle not delimited by carinae (Figure 7A) through a flat surface bounded by a U-shaped carina, to a broad lip-like structure (Figure 35A). This character tends to vary within genera.

Spiracle openings of abdominal segments IV–VI. Spiracles are visible on the gaster in the Dorylinae without dissection and their orifices can be variously shaped, from round to narrow and slit-shaped. Slit-shaped openings are rare and are found only in Eciton, Nomamyrmex, and some Neivamyrmex. Sometimes the spiracle opening on segment IV is more round than those of succeeding segments.

Girdling constriction of segment IV. The boundary between pre- and postsclerites of segment IV can be inconspicuous (Figure 58A) or marked with a constriction, also known as a cinctus (Figure 2). A simple transition is present only in some Yunodorylus while in all other genera there is a constriction. Because in specimens where the gastral sclerites are not expanded the constriction is near the articulation with the preceding segment it appears as if there is a constriction between the segments themselves.

Sculpturing of cinctus of abdominal segment IV. When the girdling constriction between pre- and postsclerites is present, it can take different forms. It can be a simple dip or a defined trench or gutter-like concavity. The constriction can also be smooth or sculptured, most often cross-ribbed with short lines (Figure 2). This character is sometimes variable within a genus, but in general the cross-ribbed cinctus is common among non-army ant dorylines and does not occur in true army ants.

Relative size of abdominal segment IV. This abdominal segment can form the bulk of the metasoma in certain species (Figure 9A) or, alternatively, be similar in size to segments III or V (Figure 7A). A large segment IV is associated with species where the waist is two segmented, or in other words where the segment III is well-differentiated from the rest of the metasoma. This condition is thus observed in all Aenictus, Eciton, Labidus, Neivamyrmex, Nomamyrmex, and Ooceraea. Other doryline genera have a generally smaller abdominal segment IV. Eburopone and Syscia present an intermediate condition where this abdominal segment is the largest but not always conspicuously so.

Anterior folding of abdominal tergite IV. In most dorylines the tergosternal Pronotomesopleural suture runs across the midheight of the segment in lateral view such that both poststernite and posttergite are visible across the entire length of the segment. In Syscia, however, the anterior portion of the Pronotomesopleural suture drops down in lateral view resulting in only the tergite being visible posterior to the cinctus (Figure 53A). This means that in ventral view the tergite can also be seen towards the anterior end of the postsclerite.

Girdling constrictions on tergites or sternites V and VI. Although most dorylines possess some form of constriction on abdominal segment IV, such constrictions can be also present posterior to that segment (Figure 51A). These can be conspicuous on both tergal and sternal portions and result in the characteristic metasoma of Aenictogiton, Eusphinctus, Sphinctomyrmex, and Zasphinctus. In other genera, the constrictions are less conspicuous but nevertheless present, as in Dorylus (Figure 20A) and certain Leptanilloides. The constrictions can also be present only on the sternites, as in some Acanthostichus or Cylindromyrmex.

Size and shape of pygidium. A modified pygidium, or tergite of abdominal segment VII, is likely a synapomorphy of the Dorylinae. See diagnostic characters above for a discussion.

Hypopygium. The hypopygium is the sternal portion of abdominal segment VII of workers. Occasionally, as in some species of Syscia and Ooceraea, the hypopygium can be lined with specialized thick setae similar to those on the pygidium.

Number and shape of tibial spurs. One or two multicellular articulated projections, known as spurs, may occur at the apex of tibiae. The configuration of spurs on the middle and hind tibiae can be useful in identification of doryline genera. There can be two spurs on both middle and hind tibiae, which is the condition seen in Aenictus, Chrysapace, Cylindromyrmex, and Yunodorylus, as well as at least one Leptanilloides species. More commonly, there is one spur on both middle and hind tibiae. In a few genera there are no spurs on middle tibiae but one spur is present on the apex of hind tibia. These include Simopone, Tanipone, and Vicinopone. The shape of the spurs may also vary, from spurs that have a well-defined comb-like or pectinate margin, through barbulate surface, to simple seta- or spike-like spurs.

Form of hind basitarsus. In most dorylines the first segment of hind tarsus, the basitarsus, is circular in cross-section and as wide basally as distally. Syscia is an exception where the basitarsus is oval in cross-section, gradually widening towards the apex (Figure 53A).

Posterior flange of hind coxa. The joint between the coxa and femur is marked by a pronounced concavity in the former. Just posterior of that concavity, a thin lamella can be found in Lioponera (Figure 2). The lamella is usually broadest distally, toward the apex of coxa. This character is variously developed but will serve to distinguish most Lioponera species from other dorylines.

Metatibial gland. See the discussion under diagnostic character 15 above.

Metabasitarsal gland. See discussion under diagnostic character 15 above.

Hind pretarsal claws. The pretarsal claws can be simple or armed with a tooth in certain lineages. This feature is generally consistent within a genus and thus a reliable character for identification. Pretarsal claws are armed with a tooth at least on the hind leg in some Cerapachys, all Chrysapace, Simopone, Tanipone, Vicinopone, and the Eciton genus-group species except Neivamyrmex.

Because several features of male morphology are similar to those of the worker, I focus on the characters and character systems unique to the male, including flight sclerites, genitalia, and wing venation.

Number of antennal segments. This character varies as in the worker caste, but the plesiomorphic condition is 13 segments (Figure 52A) and a reduction to 12 is less common, occurring only in Eusphinctus (Figure 27A), Ooceraea, Simopone, and Syscia. Many Ooceraea have 11-segmented antennae. Some Acanthostichus and Zasphinctus also show a reduction in the number of antennal segments to 12.

Notauli. The notauli are grooves on the mesoscutum, or the anterior plate of the male mesonotum. When present, they are usually well-developed as V- or Y-shaped grooves converging towards the posterior (Figures 3, 27A). The notauli appear to be consistently absent from all army ant genera (Figure 8A) as well as Tanipone and Yunodorylus. Genera that are polymorphic with regard to the presence or absence of notauli include Acanthostichus, Cylindromyrmex, Eburopone, Lioponera, Parasyscia, and Zasphinctus.

Metapleural gland opening. Unlike doryline workers, where the orifice of the metapleural gland is always present, many males have lost this feature. Even in the case of males possessing a concavity or orifice in the cuticle where the gland would be located, it is not clear whether this structure is connected to functioning glandular tissue. Because this character appears to be of some diagnostic value, however, I coded its presence and absence in the doryline males without any assumptions on gland activity.

Propodeal lobes. See discussion of this character under worker morphology above. The lobes are well-developed in males of non-army ant dorylines, where they seem to nearly always project beyond the dorsal margin of propodeal foramen. They are somewhat better developed in many Eciton genus-group males relative to the worker but in these genera the dorsal margin of propodeal foramen projects about as far posteriorly as the propodeal lobes.

Shape of abdominal sternite VII. The abdominal sternite VII is often a simple sclerite with a flat surface and no protrusions. In most Ooceraea, however, this sternite is modified to be notched, often with extensions on either side of the sclerite supporting thick setation, sometimes forming a brush.

Shape of abdominal sternite IX. See discussion under male diagnostic character 1.

Male genitalia. In the descriptions I provide a very general account of the genital morphology. The terminology I use here follows Boudinot (2015).

Cupula. The basalmost sclerites form the cupula, also known as the basal ring. In most general terms, the doryline cupula can be short or long relative to the length of the genital capsule. The cupula is best developed in the Eciton genus-group, where it is characteristically nearing or exceeding the length of the rest of the genital capsule. In most non-army ant dorylines the cupula is shorter than half the length of the rest of genital capsule but nevertheless conspicuous. A few genera have a cupula that is very short, a narrow ring of cuticle at the base of genital capsule. These include Aenictus, Dorylus, Leptanilloides, and Yunodorylus. Apparent cupula length can also vary depending on whether viewed from above or ventrally.

Basimere and telomere. The outermost valve of the genital capsule is the paramere. The paramere can be divided into the basal portion called the basimere and the distal portion the telomere. In most dorylines these two portions are broadly connected but the New World army ants are an exception where the telomeres are very narrowly connected to dome-like basimeres (Figure 15D). In other dorylines that connection is broad and either marked by a sulcus or not. Furthermore, the left and right basimere arms are most often abutting ventrally although occasionally they can be separated, most conspicuously in some Aenictus (Figure 10D). The distal portion of the basimere, the telomere, can be variously developed, straight or hooked, gradually tapering to a point at the apex or broad distally.

Volsella. The second outermost valve is the volsella. Similarly to the telomere, volsella can be of variable shape.

Penisvalvae. The innermost valves, the penisvalvae, aedeagal valves, or collectively the aedeagus, are similarly variously shaped and can be apically rounded, grossly expanded, and straight or hooked. The setation of the apex of penisvalvae is a reliable character distinguishing otherwise similar males of New World army ant genera Cheliomyrmex, Labidus, and Nomamyrmex, which possess hairs, from Eciton and Neivamyrmex with hairless penisvalvae. Given much intrageneric variability of the genital capsule, and especially its inner valves, it is likely that this revision does not provide a complete description of its morphological diversity. Statements about male genitalia will certainly be revised and refined when additional males are examined.

Tegula. The tegula is a small, dome- or strap-shaped sclerite that covers the base of the fore wing. In the Dorylinae the tegula varies in shape from broadly oval to thin and strap-shaped but it is generally conspicuous (Figure 19A–B) except for Leptanilloides in which it is extremely reduced or completely absent (Figure 31B), i.e. not easily discernable except under high magnification in the largest of species.

Fore wing venation. Wing venation often varies considerably within a genus. However, because it is obvious and because a certain pattern can be characteristic for the vast majority of species within a given genus despite occasional reductions, this is a character system valuable for identification (Figure 4). The fore wing venation of a doryline ant can include one to eight closed cells. When referring to presence or absence of veins in the descriptions, a vein is considered present regardless of whether it is tubular, nebulous, or spectral (Mason 1986).

Costal vein (C). The vein on the leading margin of the fore wing anterior to the pterostigma is called the costal vein. This vein is an important character for genus identification although its presence may be challenging to ascertain when the leading wing margin is folded onto itself. The costal vein is often present (Figure 6F) but conspicuously absent from Lioponera (Figure 33A), Lividopone, Parasyscia, Syscia, and Zasphinctus. Ooceraea appears to be exceptional in being polymorphic for the presence of the costal vein. The pterostigma is a pigmented area that in the Dorylinae can be either a conspicuous oval at the leading edge of the fore wing or only a narrow extension in line with the longitudinal veins running along that edge. Most dorylines possess the former state where pterostigma is well-differentiated from surrounding venation. Dorylus and Eciton genus-group are exceptions as these genera have a more or less narrow pterostigma.

Radial vein (R). In ants, this vein is considered to be fused with subcostal vein and radial sector (Sc+R+Rs) proximally, bifurcating into Sc+R and radial sector (R) before reaching pterostigma. The free abscissae R·f1–2 are fused with the pterostigma and R·f3 projects distally of the pterostigma on the leading edge of the fore wing. In the Dorylinae, R·f3 is generally associated with overall well-developed venation and is found in all New World army ant genera. Among non-army dorylines, it can be found in Acanthostichus, Cerapachys, Chrysapace, Cylindromyrmex, most Eburopone, Neocerapachys, Procerapachys, and Yunodorylus.

Radial sector (Rs). Past the separation from Sc+R, the radial sector continues posterior to the pterostigma, first as a usually short free abscissa Rs·f1, then merging with median vein (M) and continuing fused (Rs+M) for some time, followed by the free abscissae Rs·f2–5 that, when present, constitute the next longitudinal vein system posterior to the pterostigma. Rs·f5 may distally connect with R·f3 to close a marginal cell. When Rs is present distally to Rs+M, it connects to the pterostigma via the second radial-radial sector cross-vein (2r-rs). Various levels of reduction of the radial sector are found within the Dorylinae. The most complete development is found in Eciton genus-group, Cerapachys, Chrysapace, Cylindromyrmex, Procerapachys, Sphinctomyrmex, and some Neocerapachys and Yunodorylus. In these ants the free abscissae of the radial sector continue to close the marginal vein by joining R·f3 at the wing margin, in some species interrupted only at the connection with Rs+M. In other genera such as Parasyscia, Lividopone, or Zasphinctus radial sector abscissae Rs·f2–3 connect to Rs+M but radial abscissa R·f3 is absent and Rs·f4–5 do not close the marginal cell. In Aenictogiton and Simopone the radial sector is further reduced, interrupted near Rs+M junction and not reaching wing margin. In Lioponera, Eburopone, and some Ooceraea and Syscia the abscissae Rs·f2–3 are absent and Rs·f4–5 together with 2r-rs form a ‘free stigmal vein’ that does not reach wing margin. More reduction is found in various genera where smaller species sometime lost all radial sector veins past Rs+M.

Median vein (M). Further away from the leading wing margin is the median vein, proximally fused with cubital vein (M+Cu), following separation continuing as a free abscissa M·f1 before joining with radial sector to form Rs+M. In the Dorylinae the next free abscissa (M·f2) may be separated from Rs+M if Rs·f2–3 or continuous with Rs+M in the absence of radial sector. If median vein is present past the junction with the radial sector, further free abscissae M·f3 and M·f4 can be differentiated in the presence of second radial sector-median cross-vein (2rs-m) and M·f4 may extend all the way to the distal wing margin. Various reductions are possible from this basic pattern. The median vein is highly variable, from the best developed in the Eciton genus-group, where it almost always reaches wing margin as a tubular vein, through a state where nebulous or spectral free abscissae M·f3–4 are disconnected from other veins, to entirely absent past Rs+M as in certain Leptanilloides.

Cubital vein (Cu). Proximally the cubital vein is fused with median vein (M+Cu) and can have up to three free abscissae, Cu·f1 through Cu·f3. The first median-cubital cross-vein (1m-cu) may connect the cubital vein to the median vein between Cu·f2 and Cu·f3. Cu·f3 can further distally branch into as many as three branches, Cu1–3. Cu1 is often present, even in species with venation otherwise reduced in the radial sector and the median vein. Cu1 is the long branch running towards the distal wing margin. Cu2 is short and sometimes connects to the anal vein (A). When present, Cu3 is always a short stub directed towards the posterior wing margin. In dorylines it is found only in the largest of males, in the New World army ants and some Dorylus.

Anal vein (A). The anal vein is the longitudinal vein running near the posterior wing margin. In dorylines it consists of at least one free abscissa fused to or terminating near cubital-anal cross-vein (cu-a), a connection to the cubital vein, and often two abscissae are present (A·f1–2) if continuing past cu-a. The position of cu-a relative to the branching of M·f1 can help distinguish a male of the Eciton genus-group from an Old World army ant: in the former M·f1 arises much closer to the base of the wing than cu-a while in Aenictogiton, Aenictus, or Dorylus M·f1 arises either distally to cu-a, directly below it, or only slightly proximally.

Hind wing venation. Veins in the hind wing follow a similar pattern to that found in the fore wing but there is no pterostigma and the venation is simplified (Figure 4). In the dorylinae the hind wing venation can from zero up to three closed cells. At least vein Sc+R+Rs appears to be present in all species.

Because the morphology of the majority of doryline gynes is much like the worker, this revision does not describe gynes in detail. Instead, a general morphology is indicated, including how gynes differ from the worker, whether the known forms are alate, ergatoid (wingless and worker-like), or dichthadiigyne (see below), and references to more detailed descriptions are provided where available. The gyne morphology can be variable within a genus or even within species where intercastes, or individuals with morphology intermediate between workers and gynes, are known in addition to fully-developed gynes.

As mentioned above, the doryline gynes can be classified as alate, worker-like, or ‘dichthadiigyne’ or ‘subdichthadiigyne’, although a gradation of intermediates between all these morphologies is also observed among the doryline species. Fully alate gynes possess the usual complement of flight-associated sclerites, relatively large eyes and ocelli, and are apparently capable of flight. Alate gyne material available is often scarce and inference about the presence of wings from dealated gynes is difficult because obvious mesosomal sutures and wing scar-like structures are not always indicative of presence of fully developed wings in the virgin gynes. At least one species is known to have brachypterous (short-winged) gynes, bridging the gap between fully alate and worker-like morphologies. Wingless, ergatoid gynes are very common in dorylines. They exhibit variation in how different they are from the workers, ranging from gynes essentially indistinguishable from the worker to ones that have enlarged gasters, large eyes and ocelli, and wing scar-like structures on the mesosoma. The term ‘dichthadiigyne’ (Wheeler 1908) has been used to describe the highly specialized gyne morphology characteristic of the ‘true army ants’. Dichthadiigynes are wingless but, unlike ergatoids, much different from the worker caste. They usually have falcate mandibles, are often completely blind or possess only small eyes, have enlarged gasters capable of significant distension, and possess one-segmented waist, even if the corresponding worker caste has a well-differentiated abdominal segment III (postpetiole). Gynes intermediate between ‘simple’ ergatoids and highly derived dichthadiigynes are also known and these have been sometimes termed ‘subdichthadiigynes’.

Alate or apparently alate (known only from dealated specimens) gynes are so far known in Acanthostichus, Cerapachys, Chrysapace, Cylindromyrmex, Eburopone, Lioponera, Lividopone, Neocerapachys, Parasyscia, Simopone, Syscia, Vicinopone, and Zasphinctus. Ergatoid gynes are found in Cerapachys, Eburopone, Eusphinctus, Lioponera, Ooceraea, Parasyscia, Simopone, Sphinctomyrmex, Tanipone, and Zasphinctus. Subdichthadiigynes or dichthadiigynes are found in Acanthostichus, Leptanilloides, Ooceraea, Zasphinctus, and ‘true army ants’ in the genera Aenictus, Dorylus, Eciton, Labidus, Neivamyrmex, and Nomamyrmex. A description of a Yunodorylus subdichthadiigyne is currently awaiting publication (Eguchi et al. in press). No gynes are known so far for Aenictogiton and Cheliomyrmex. Because of the apparent repeated evolution of derived wingless gynes, the dorylines could become a good system for studying gyne evolution once a more complete picture of the diversity of gyne morphologies is available.

Certain couplets build upon keys in Gotwald (1982), Bolton (1994), and Bolton and Fisher (2012). Figure pointers refer to plate following couplet.

Keys to the true army ants are modified from Gotwald (1982). Males of Vicinopone are unknown. This key is preliminary and should be used in conjunction with generic diagnoses and descriptions. Figure pointers refer to plate following couplet.