bioRxiv preprint first posted online May. 2, 2016; doi: http://dx.doi.org/10.1101/051367. The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It is made available under aCC-BY 4.0 International license. Description ofanew genus of primitive ants fromCanadian amber, with the study ofrelationships between crownants andstem ants (Hymenoptera: Formicidae) LeonidH.Borysenko Canadian National Collection ofInsects, Arachnids andNematodes, AAFC,K.W. Neatby Building 960Carling Ave., Ottawa, K1A0C6, Canada [email protected] 1 bioRxiv preprint first posted online May. 2, 2016; doi: http://dx.doi.org/10.1101/051367. The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It is made available under aCC-BY 4.0 International license. Abstract A detailed study of the holotype of Sphecomyrma canadensis Wilson 1985 from Canadian amber led to the conclusion that the specimen belongs to a genus here named Boltonimecia gen.nov. not previously described. The taxonomy of stem ants is not clearly understood; thus, in order to find the taxonomic position of the new genus, it was necessary to review the classification of stem ants in a study of their relation to crown ants. Lacking morphological data for traditional taxonomic approaches, a statistical study was done based on a morphometric analysis of antennae. The elongation of the scape is believed to play an important role in the evolution of eusociality in ants. However, this hypothesis has never been confirmed statistically. Our statistical analysis lends support to the view that the morphology of antennae reliably distinguishes stem ants from crown ants to determine whether a species belongs to one or the other group. This in turn may indicate the relationship exists between eusociality and the elongation of the scape. A review of Cretaceous records of ants is made and the higher classification of Formicidae with definitions of stem and crown groups is proposed. The newly obtained data are discussed focusing particularly on the origin, evolution and diversity ofants. Introduction Mammals and birds immediately come to mind when thinking of groups of animals that began to flourish after the Cretaceous–Paleogene extinction event. However, ants also started to flourish at this time. Ants were rare in the Late Cretaceous but in the Cenozoic these insects underwent an impressive radiation to become one ofthe largest and most widespread of terrestrial animals (Hölldobler &Wilson 1990). There are poor quality impressions of Cretaceous ants found in sedimentary rocks, and rare specimens are also preserved in amber (reviewed in LaPolla et al.2013). Among only a handful of amber sites known to contain ants, Canadian amber is of a special interest: dating from a part of the Campanian, 78–79 million years (Ma) old (McKellar & Wolfe 2010), it contains traces of one of the most late Cretaceous ecosystems flourishing inNorth America only10Mabefore the Cretaceous–Paleogene (K-Pg) extinction event. It is noteworthy that both crown ants (i.e. the descendants of the most recent common ancestor of all extant ants) and stem ants (i.e. all extinct taxa outside the crown clade but most closely related to it than to other Aculeata) are found in Canadian amber. The two groups lived together for at least 10 Ma but only in Canadian amber are they found in almost equal numbers. Of these ants, four species represent recent subfamilies Dolichoderinae (Eotapinoma macalpini, Chronomyrmex medicinehatensis) (Dlussky 1999a; McKellar et al. 2013a), Ectatomminae (Canapone dentata) (Dlussky 1999a), Aneuretinae (Cananeuretus occidentalis) (Engel & Grimaldi 2005) and at least three species represent the extinct subfamily Sphecomyrminae: Sphecomyrma canadensis (Wilson 1985), Haidoterminus cippus (McKellar et al. 2013b), Sphecomyrma (?)sp.(Grimaldi etal. 1997). The present paper focuses on the holotype of Sphecomyrma canadensis held in the Canadian National Collection of Insects, Arachnids and Nematodes (CNC). Some authors noted that the description of S. canadensis is unsatisfactory, no character of Sphecomyrma is visible (Dlussky 1996; Grimaldi et al. 1997; Engel &Grimaldi 2005) andthe holotype and theparatype are irrelevant toone another (Dlussky 1999a). Examining the holotype, we found the head, mandibles and the antennae's distal parts almost invisible and so establishing a link to Sphecomyrma impossible. After the amber was polished, visibility improved and some details of the ants' morphology could be seen. What had seemed to be a black inclusion hiding the head, could be identified as a thick raised platform, a unique morphological structure. The head also have lateral protrusions posterior to the insertion points of the antennae. Because of these morphological characteristics,we decidedtotreat the specimen asbelonging toa genus notpreviously described. The next step was to find a taxonomic position for the new genus, which because of the unsatisfactory 2 bioRxiv preprint first posted online May. 2, 2016; doi: http://dx.doi.org/10.1101/051367. The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It is made available under aCC-BY 4.0 International license. state of stem ant taxonomy is not an easy task. Classification of stem ants is still in its infancy because there is very limited morphological data, and molecular data are impossible to obtain. Stem ants have never been a subject of a general revision, and only twice were they included in morphological cladistic analyses (Baroni Urbani et al. 1992; Grimaldi et al. 1997). At the moment, they are assigned to two poorly supported subfamilies, Sphecomyrminae and Armaniinae (Bolton 2003), but some species, such as Gerontoformica cretacica Nel & Perrault, and Myanmyrma gracilis Engel & Grimaldi, are so bizzare that they cannot be assigned even to these subfamilies and are left as insertae sedis (Nel et al. 2004; Engel & Grimaldi 2005). There is also a long-standing debate regarding the taxonomic status of Armaniinae, which represent either the most basic stem ants (Dlussky 1983) or are the sexuals of Sphecomyrminae preserved only in rock impressions due to their large size (Wilson 1987). The new genera described recently (Zigrasimecia) are causing uncertainty as well (Barden & Grimaldi 2013). The absence of sufficient morphological data creates a need to invent new methods of taxonomic analysis based on principal differences between stem ants and crown ants. In an attempt to fill this void, we chose Dlussky’s (1983) approach suggesting that the antennal morphology can be a diagnostic tool to distinguish stem ants from crown ants, carrying out statistical analysis of antennal indexes. This idea was first expressed by Wilson, Carpenter and Brown (1967) in their diagnosis of Sphecomyrminae, and later explained in terms of evolutionary history and expanded by means of comparative analysis by Dlussky (1983). Since then, antennal characteristics have been used in diagnoses of stem-ant subfamilies, including Bolton’s system (2003), the most comprehensive for the time being. That they have never been tested by means of statistics, is surprising considering the highly interesting biological background of Dlyssky’s hypothesis: the elongation of the scape was necessary for the emergence of eusociality inants (Dlussky 1983). The final logical step of such a study was to develop the higher classification of ants including both stem and crown branches. Developing the crown group/stem group distinction, affects our thinking about ant origins; the importance of taking some definitive position has been overlooked in previous discussions on this topic (Ward 2007). Material and methods Study of amber inclusion. Photographs were taken with a Nikon D1X digital camera attached to the microscope Leica Z6 APO. Photographs were used to make drawings, which were then computer generated and adjusted using Adobe Photoshop. All measurements were made using an ocular micrometer and are in millimeters (mm). The following measurements were recorded: HL - head length (measured in full-face view as a straight line from the anterior-most point of median clypeal margin to the mid-point of the posterior margin of the head), HW - head width (maximum head width in full-face view), SL - scape length (maximum length without the condyle and the neck) F1L-F9L - length of flagellomeres (from 1st to 9th), AL - antenna length, ML - mandible length (maximum length of horizontal part of a mandible), WL - Weber’s length (the distance from the anterodorsal margin of the pronotum to the posteroventral margin of the propodeum), TL - total body length. Taxon sampling and morphometry. We took all morphometric data on antennae and heads of stem ants available from the literature; also we either made measurements or took published data on antennae and heads of crown ants representing all extant subfamilies (Тable S1). Only species of crown ants with a satisfactory description in the literature and high resolution images with good visibility of all antennal segments (available from AntWeb) were selected. Also, species were selected in such a way that they represent broad phylogenetic diversity. If a subfamily has high diversity, we tried to choose representatives with SL between the extremes (so that SLs are near to the SL mean for that given subfamily). For example, in our data the minimum value of SL is in Pseudomyrmex pallidus; SLs in representatives of Myrmicinae and 3 bioRxiv preprint first posted online May. 2, 2016; doi: http://dx.doi.org/10.1101/051367. The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It is made available under aCC-BY 4.0 International license. Formicinae 2-3 times higher, although the minimum values in those subfamilies may be even smaller than in P. pallidus. All subfamilies are represented by one species, except for the largest subfamilies (Ponerinae, Dolichoderinae, Formicinae, Myrmicinae) represented by two species. Recently, six subfamilies of the dorylomorph group have been subsumed into a single subfamily, Dorylinae (Brady et al. 2014), but here in order tocoverlarger variation we sampled all six subfamilies as valid groups. Measurements were made on mounted specimens using a Wild M10 stereomicroscope with an accuracy ±1 µm. For all stem and crown ants we calculated nine indexes showing size of antenna's parts compared to the head length (indexes SL/HL, FL/HL, PL/HL, F1L/HL, F2L/HL) and compared to the rest of antenna (іndexes SL/FL, PL/(AL-PL), F1L/(AL-F1L), F2L/(AL-F2L)) (Тables S2, S3). In those indexes, HL – head length (see definition above), SL – scape length (see definition above), FL – length of flagellum (including pedicel), PL – length of pedicel, F1L and F2L – length of 1stand 2nd flagellomeres respectively. Although F1 is in fact the second flagellar segment, the pedicel being the true first, we call it "first" throughout the paper to make a clear distinction from the pedicel; thus our nomenclature follows that of Barden and Grimaldi (2012, 2013). For indexes we always used HL, not HW, as HL is available for more fossil species; in addition, using HL, our data can be compared with Dlussky's data on antennal morphometry (Dlussky 1983, Dlussky and Fedoseeva 1988). Two additional indexes AL/HL and SL/AL (where AL – antenna length, SL – scape length) were used only to compare obtained data with Dlussky’s data on Vespoidea andApoidea (Dlussky andFedoseeva 1988). For general observations on the shape of pedicels, the morphology of the middle and terminal antennal segments, we used high resolution images available from AntWeb. To study the width of the petiolar attachment ofArmaniinae, we calculated the index PG/PH (where PG - the width ofthe petiole in the broadest point ofits attachment to the gaster; PH-the maximal height ofthepetiole). Statistical analysis. Two statistical tests on equality of means were performed using the SPSS 17.0 statistical package: the Student's t-test and one-way ANOVA with planned comparison. The log transformation was applied in those cases where substantial heteroscedasticity and non-normal distribution were observed. Also, a correlation and regression analysis as well as a canonical discriminant analyses were performed using the SPSS 17.0. The power analysis was performed using the G* Power 3 program (Faul et al. 2007). Despite the lack of sufficient data on extinct taxa, in all cases where the t test showed statistically significant results, the statistical power was also high enough. However an unbalanced design in an ANOVA could be be a problem (McDonald 2014). Indeed, the results showed that the unbalanced design in which extinct ants are underrepresented, had low statistical power. In such a case, the only way to confirm the reliability of the obtained results is to reduce big groups to the size of the smallest group, and run the power analysis and ANOVA again. Doing this, we obtained very high values of the statistical power with the results of ANOVA almost identical tothat ofthe unbalanced design. The next concern about the reliability of the obtained statistical data is a measurement error. Since the measurements of extinct taxa were taken independently by different persons, with different material and calibration, those slight differences might presumably alter the obtained conclusions. To check how strongly our statistical results are sensitive to fluctuations, we added/ subtracted 10% (a considerable measurement error) to/from the indexes and to/from the measurements. To insert these modifications randomly, we used a random number generator. Then the modified data were again used in ANOVAs and t tests. In all cases no noticeable effect was observed, the statistical model proved to be robust and not sensitive to fluctuations. Caution, asalways, should be taken about data onthe verge ofstatistical significance. Systematic palaeontology Family Formicidae Latreille, 1809 4 bioRxiv preprint first posted online May. 2, 2016; doi: http://dx.doi.org/10.1101/051367. The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It is made available under aCC-BY 4.0 International license. Subfamily Sphecomyrminae Wilson & Brown, 1967 Tribe Boltonimeciini trib.n. Typegenus Boltonimecia gen.n. Diagnosis (workers): see chapter"Higher classification ofants". Genus Boltonimecia gen.n. Type species:Sphecomyrma canadensis Wilson 1985 Diagnosis. As for the tribe. Etymology. This genus isdedicated tothe renowned English myrmecologist Barry Bolton. Boltonimecia canadensis (Wilson 1985)comb.nov. Figs 1,2 Sphecomyrma canadensis Wilson, 1985: 206,figs. 1,2(w.) Material examined. Holotype. Worker is preserved in a clear orange small (8×3×2 mm) piece of Canadian amber (chemawinite/cedarite) from Medicine Hat, Alberta (J.F.McAlpine, CAS 330), held in the CNC, Ottawa, Canada. Visibility is excellent from all directions, but for the ventral side which is clouded. Preservation of the ant is excellent, although mesosoma, left flagellum and right side of the head are somewhat distorted (flattened) due to amber compression, and junctions of both scapes with the pedicels are gone. The specimen known as the paratype of Sphecomyrma canadensis (Wilson 1985) (CAS 205 held in CNC, Ottawa, Canada) represent body fragments of very poor preservation, with no taxonomic characters visible, and thusshould not beconsidered astheparatype and left asanunidentifiable ant. Diagnosis. As for genus. Description.HL 0.73 mm; HW 0.8 mm; SL0.5mm;ML 0.48mm; WL 1.2mm; TL3.4mm. The head is small compared to the body length (1/5 of its length), subglobular, prognathous, slightly broader than longer, seems to be triangular when seen from the top, and formed into a shield: its dorsal part thick, raised, curved in profile (Fig. 2B, C, D). Such shield-like head is likely formed by the expanded frontal lobes. Under posterolateral edges of the expanded frontal lobes, on both side of the head, there are two sticklike processes directed anterolaterally; they limit the mobility of the antenna posteriorly, so that when a scape pressed to the head and lifted, it touches a process. Eyes and ocelli absent. The wierd processes are hardly deformed eyes, since there is no trace of facets visible, plus processes are completely covered by appressed pubescence similar to that on the front of the head. The clypeus is not large, in profile strongly convex. The convex medial part of the clypeus is highest, it gradually declines towards the lateral margins. Lateral margins of the clypeus bear semitransparent semicircular projections, covering the insertions of mandibles. Clypeus posteriorly not inserted between antennal sockets. The anterior clypeal margin bears 25 peg-like setae 0.01 mm long. The width of the clypeus 0.5 mm (without the lateral semicircles), length - 0.15 mm. Mandibles linear, two toothed, curved at almost 90°. The apical tooth is longer than preapical - 0.15 5 bioRxiv preprint first posted online May. 2, 2016; doi: http://dx.doi.org/10.1101/051367. The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It is made available under aCC-BY 4.0 International license. mm and 0.06 mm respectively. On the inner side of the apical tooth there is a longitudinal impression, which probably fits the other mandible when mandibles are closed. When fully closed,the mandibles overlap at half the length of their horizontal parts. The length of the horizontal part of mandibles is 0.48 mm, vertical - 0.17 mm. Antennae 11-segmented. The scape hardly reaches the occipital margin. The measurements of antenna (mm): SL - 0.5, PL - 0.2, F1L - 0.15, F2L - 0.2, F3L-F8L - 0.17, F9L - 0.25, AL - 2.32. Insertions of antennae are not far from each other (0.17 mm), partly exposed under the anterior side of the expanded frontal lobes, touching the posterior margin of the clypeus. Toruluses not fused to the frontal lobes. Antenna moves in a special fossa bounded laterally by the expanded frontal lobes, and posteriorly by the sticklike process. The metanotal groove visible. The propodeum located a little bit lower than the promesonotum, without teeth or spines, its basal surface slightly shorter than the declivity. Propodeal spiracles located high, in the upper half ofthepropodeum.The metapleural gland orifice covered byguard setae. Petiole pedunculate, 0.4 mm long; peduncle 0,1 mm long, node 0.2 mm high. The anterior part of the peduncle is narrow (the width of its attachment to the propodeum is about half the height of the node); posterior foramen more broad. Gaster subglobular, 1.1 mm long. Helcium projecting below the midheight of the anterior face of abdominal segment III, which has vertical and relatively high anterior face. Abdominal segment IV without presclerites. Stingpresent; visible length 0.05mm. Legs long - 3.3 mm forelegs (shortest), 5.07 mm hind legs (longest, 1.5 of body length). Measurements of leg segments shown in Table 1. Pretarsal claws with one preapical tooth (Fig.1D). Each tarsal segment has two stiff setae on both sides of its apex. Basitarsus with notch and comb for cleaning the antennae. Protibia with one pectinate and two simple spurs, while meso- and metatibia with one pectinate and one simple spur(Fig. 1E). The dorsal surface of the head as well as antennal fossae covered with dense short appressed pubescence. Sticklike processes covered with very short and hardly visible appressed pubescence. The lateral margins of the head covered with erect and suberect hairs. Antennae covered with dense short appressed pubescence. Long standing sensory hairs project anteriorly from anterior margin of the clypeus, from its middle part, and from anterior margin of the expanded frontal lobes; some of those hairs are so long, that they go beyond the closed mandibles. The external margin of each mandible covered with suberect hairs. Short appressed pubescence completely cover mesosoma, legs and gaster. In addition, middle and hind tibia covered with sparse suberect hairs; small bundles of erect hairs project from the apex of the middle and hind femurs; ventral surfaces of coxae covered with long erect hairs; pronotum and propodeum covered with long white erect hairs tapering to sharp points (especially long on pronotum); abdomen covered with sparse suberect hairs,longer on the ventral surface. Sculpture invisible. The entire body is transparent, colored as surrounding amber, but the ventral surfaces of coxae, abdomen, proximal halves of tibiae, lower half of propodeum are brown to black. The dorsal surface of the head and pronotum black, opaque. Discussion. Some authors noticed that Sphecomyrma canadensis has no synapomorphies of Sphecomyrma - namely a short pedicel and 2nd flagellomere two or more times longer than other flagellomeres (Dlussky 1996; Grimaldi et al. 1997; Engel & Grimaldi 2005). Dlussky & Fedoseeva (1988) even suggested that this species has "formicoid" antenna (or at least antenna transitional towards formicoid type), unlike Sphecomyrma which has "sphecoid" antennae, so S. canadensis has to be assigned to its own subfamily. After polishing the amber of the holotype, I uncovered even more differences of the specimen from all known Cretaceous ants. B.canadensis has 11-segmented antennae, a unique character of Cretaceous ants (the statement of McKellar et al. (2013b) that Haidomyrmex and Haidomyrmodes have 11-segmented antennae is erroneous). 6 bioRxiv preprint first posted online May. 2, 2016; doi: http://dx.doi.org/10.1101/051367. The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It is made available under aCC-BY 4.0 International license. A small proximal part of both scapes of B.canadensis is missing because of amber destruction, but the articulation of the scape and pedicel of the left antenna is intact and thus one can confidently infer about the scape length. The visible articulation is not an articulation between pedicel and F1 because (1) pedicel’s curved base (distinct character of a pedicel only and a point of its attachment to the scape) is clearly visible, (2) localization of the scape and flagellum infers that there is no room left for one more antennomer between them. In general, the relative size of the antennomeres are not diferent from the Wilson's drawings (Wilson 1985). The only exception is the distal parts of the antennae, which are curled back under the head and were almost invisible before polishing; this probably led to Wilson’s assumption about 12-segmented antenna as in Sphecomyrma. The most distinctive character of B.canadensis, which immediately sets it apart from all fossil and recent genera is a morphologically unique head: thick shield-like head, sticklike head processes, long sensory hairs. Although the left side of the head is flattened due to amber compression, the processes are clearly visible onboth sides ofthe head. Some Myrmicinae, Aulacopone relicta (Heteroponerinae) and Agroecomyrmecinae, have shield-like heads formed by the expanded frontal lobes. In Myrmicinae, the clypeus usually inserted between antennal sockets, unlike B.canadensis. In A. relicta the clypeus is shallow, but antennal insertions close together, and the fronto-clypeal part of the head is extended forwards and hangs over the mandibles (Taylor 1980), unlike B.canadensis. Species of enigmatic subfamily Agroecomyrmecinae have shield-like heads; in addition, in arboreal Ankylomyrma coronacantha the occipital margin of the head bears weird spiniform processes projecting especially strongly at the occipital corners (Bolton 1973), that resembles specialized head of Boltonimecia. However in Agroecomyrmecinae the clypeus is large and broadly inserted between the frontal lobes. Most Ponerini have a clypeus not inserted between antennal sockets, but their fused frontal lobes form only a small triangle or linear strip partially covering the sockets, never occupying the entire dorsal part of the head. In Proceratiinae, clypeus reduced, and antennae inserted close to the anterior margin of head. Some Proceratiinae (Dyscothyrea) have frontal lobes fused together and forming small raised platform behind the level of antennal sockets, the sides of which are strongly convergent anteriorly (Bolton 2003). Also in Proceratiinae as well as in A.relicta, and in Agroecomyrmecinae the frontal lobes take part in the formation of antennal scrobes, absent in B.canadensis. In any case, it is important toemphasize that such morphological similarities between crown ants and stem ant B.canadensis are likely acquired by a parallel evolutionary process. Any assumption about lifestyle of B. canadensis reflected in its bizzare head morphology, can be very speculative. The lack of eyes and ocelli, as well as multiple sensor hairs on the anterior margin of the head may imply cryptic lifestyle. However, extant blind ants are not always exclusively subterranean (e.g. Dorylus). Long legs of Boltonimecia speak in favor of above-ground or arboreal lifestyle. I also do not reject the possibility that the eyes of Boltonimecia could be reduced to a single facet, and thus simply invisible in amber inclusion. Determining the phylogenetic place of Boltonimecia, we cannot avoid the study of classification of stem ants and their relationship tocrown ants. That is adifficult task still absolutely unexplored. What is the difference between stem ants and crown ants? We realized the distinction between crown group and stem group with the launch of phylogenetic approach to taxonomy (Hennig 1966). Crown group is a clade that consists of the living species together with their ancestors back to their most recent common ancestor, while a pan group or total group is the crown group and all organisms more closely related to it than to any other extant organisms. From this definition, stem group is just the total group minus the crown group. It is important to emphasize that stem group, contrary tothe crown group, isnota taxa -itisan artificial assemblage, aparaphyletic group (Ax 1985). Here I treat ant subfamilies Sphecomyrminae and Armaniinae, as well as insertae sedis genera 7 bioRxiv preprint first posted online May. 2, 2016; doi: http://dx.doi.org/10.1101/051367. The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It is made available under aCC-BY 4.0 International license. Gerontoformica andMyanmyrma asstem ants.This idea will be expanded further indetails. Extant ants (i.e. crown ants) have four commonly accepted synapomorphies: elongated scape, petiole, metapleural gland, and the non-morphological trait of eusocial behavior, which is expressed morphologically in caste differences between queens and workers (LaPolla etal.2013). Out of characters thought to be plesiomorphic for Formicidae, and helpful in distinction between stem ants and crown ants, the following ones are mentioned most often: a wide attachment of the petiole and the gaster, pretarsal claws with preapical tooth, bidentate mandibles, two spurs on meso- and metatibiae, short scape, trochantellus, anterial clypeal peg-like setae, and ocelli. It would be useful to review briefly all of them. Petiole of non-ant Vespoidea is less pronounced and more broadly attached to the gaster than in most Formicidae, except for only one ant subfamily, Amblyoponinae. Thus a broad petiolar attachment can be viewed as a reliable plesiomorphic character. It follows from the direction of ant evolution, highlighted in an increase of the gaster mobility and as a result in an ease ofmesosoma-gaster articulation. If the ancestor of all ants possessed a wide attachment of the petiole and the gaster, all presently known stem ants should be viewed as quite advanced, because they all have narrow petiole-gaster attachment, as most crown ants (the issue ofpetiole ofArmaniinae ramains open andIwill come back to itbelow). Preapical tooth on pretarsal claws present in many crown ants: poneroids and primitive members of the formicoid clade (Myrmeciinae, Pseudomyrmecinae, Dorylinae) (Dlussky & Fedoseeva 1988). This character is common inother families ofVespoidea, and are present virtually inallstem ants. Bidentate mandibles is a very common character in Vespoidea and Apoidea and universal for stem ants. It is believed that in females of crown ants this character is a result of reduction of initial triangular mandibles (Dlussky & Fedoseeva 1988). In crown ants, this character is mainly present in poneroids and primitive formicoids, being quite rare in advanced formicoids - Formicinae, Myrmicinae, Dolichoderinae (Bolton 2003, Appendix 2). In males of crown ants, this character is usual, although it is not as pronounced as in other Vespoidea, and is again most likely the result of teeth reduction. Such a reduction in males may be linked to the fact that they do not feed outside a nest. As in females, in males of the formicoid clade this character is not as usual as in males of the poneroid clade: in Formicinae and Myrmicinae it present in 1/4 of genera; rarely present in males of Dolichoderinae, ectaheteromorphs, Pseudomyrmecinae; absolutely absent in males of Myrmicinae, but usual inDorylinae (Bolton 2003,Appendix 2). Two spurs on the meso- and metatibia (one of which may be pectinate) are usual in poneroids and primitive formicoids (Bolton 2003, Appendix 2), as well as in other families of Vespoidea, although pectinate spur of metatibia is found only in ants and Tiphiidae (Brothers 1975). This character present in all stem ants. Haidoterminus cippus was reported with a single metаtibial spur and two mesotibial spurs (McKellar et al. 2013b) that most probably is a result of poor preservation of legs, as such a condition is unknown forFormicidae. Trochantellus absent in crown ants, except putative Cretaceous crown group species Cananeuretus occidentalis (Engel & Grimaldi 2005). In stem ants, trochantellus is present in one species of Haidomyrmex (H. scimitarus) (Barden & Grimaldi 2012), Haidoterminus cippus (McKellar et al. 2013b), both species of Zigrasimecia (Barden & Grimaldi 2013; Perrichot 2014), some species of Sphecomyrmodes (Barden & Grimaldi 2014), and also in males of Baikuris (Dlussky 1987; Grimaldi et al. 1997; Perrichot 2015). In Armaniinae, this character is unclear as a result of poor preservation of rock impressions: Armania and Pseudarmania have been reported either with or without a trochantellus (Dlussky 1983; Wilson 1987; Dlussky & Fedoseeva 1988). In general, trochantellus is not often present in Vespoidea (Rhopalostomatidae, some Vespidae). Clypeal peg-like setae are thought to be an important ant plesiomorphy (Engel & Grimaldi 2005). In stem ants, this character is present in Sphecomyrmodes, Boltonimecia, Gerontoformica, Myanmyrma, Zigrasimecia; in crown ants - in Amblyoponinae. Peg-like setae on anterior clypeal margin present in some Vespoidea, for example in Myzinum (Tiphidae), and on anterolateral margins - in Apterogyna 8 bioRxiv preprint first posted online May. 2, 2016; doi: http://dx.doi.org/10.1101/051367. The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It is made available under aCC-BY 4.0 International license. (Bradynobaenidae). Ocelli in workers are often considered as ant plesiomorphy (Engel & Grimaldi 2005). In crown ants, this character is well known in extant ants, mainly in the formicoid clade: Myrmeciinae, Pseudomyrmecinae, Dolichoderinae, Formicinae, some Dorylinae, but absent in some Cretaceous genera (Eotapinoma, Chronomyrmex,Kyromyrma).In stem ants,ocelli present insome Sphecomyrmini. Therefore all mentioned characters, which can be used to separate stem and crown ants, are not universal and reliable. As noted by Dlussky (1983), the most reliable character may be an elongated scape of crown ants, compared to a short scape of stem ants. Scape elongation in crown ants follows from biological feasibility: long scape favours eusociality, since it allows brood and food manipulation (Dlussky 1983). Below Iexamine this character, alongwith other morphometric characteristics of antennae, in details. The antennalstructure as ahallmark ofdistinction between stem ants and crown ants The scape length of more than 25% of the antennal length is thought to be a characteristic of extant (i.e. crown) ants (Dlussky & Fedoseeva 1988), although in the diagnosis of Sphecomyrminae Bolton (2003) stated that a "short scape" of Sphecomyrminae means “0.25 times length of flagellum”. The role of other antennal parts in distinguishing stem ants from crown ants may beno less important. The pedicel of all insects contains Johnston's organ - a mechanosensory chordotonal organ responsible for hearing, graviception (Kamikouchi et al. 2009) and electric field changes which may play a role in social communication (Greggers et al. 2013). According to Dlussky & Fedoseeva (1988), the pedicel in crown ants is shorter than in stem ants, it is narrowed and curved at the base. This enables close contact of the pedicel and scape resulting in greater freedom and accuracy of movement of the flagellum, which together with scape elongation ledtothe emergence of eusociality inants (Dlussky and Fedoseeva 1988). The first segment is the longest flagellar segment in stem ants, males of primitive crown ants (Dlussky 1983), and the Aculeata closely related to ants (Engel & Grimaldi 2005); so it is a a symplesiomorphic character (Engel & Grimaldi 2005). Bolton (2003) listed this characters as a synapomorphy of Sphecomyrmini (while the longest flagellar segment in Haidomyrmecini is the second one). In crown ants, the first and second flagellomeres are not different from the other flagellomeres, except for the elongated terminal one (Dlussky &Fedoseeva 1988). In stem ants, segments beyond the second flagellomere decrease in length towards the apex of the antenna, while in crown ants they often increase ending in a club-shaped expansion of the longest terminal segments (Dlussky & Fedoseeva 1988). In females of crown ants, a club is common in advanced taxa, except for formicines in which 3/4 of the genera lack it. Males of crown ants rarely have clubs too (Bolton 2003, Appendix 2). Finally, the entire flagellum instem ants is long andflexuous (Bolton 2003). Making his hypothesis from a comparison of ants with other Aculeata, Dlussky, however, has not provided any statistical support. This has resulted in criticism and even removal of a character "short scape" from the data matrix as it is “difficult to define” (Grimaldi et al. 1997). Below we check Dlussky’s hypothesis using a statistical analysis of antennal indexes as well as try to expand and generalize the aforementioned observations on antennal structure. Comparison of antennal indexes of crownants andstem ants. Although indexes of Cretaceous males1 are within the range of the indexes of crown males, in most cases they shifted from the crown males’ mean value (Table S3). The statistical tests showed that some differences between these indexes were significant (Table S15): (1) Scape. Indexes SL/HL did not show statistically significant differences, while for SL/FL such a difference existed.The latter canbe explained by aconsiderably longer flagellum of Cretaceous males. (2) Flagellum. The mean of FL/HL is noticeably greater in Cretaceous males than in crown males, 1SincenoneofCretaceousmaleshasyetbeenassociated withworkers orgynes,wecanonlyassumetheybelongtostemants.Forthisreason, wedonotcallthem"malesofstemants"throughoutthepaper. 9 bioRxiv preprint first posted online May. 2, 2016; doi: http://dx.doi.org/10.1101/051367. The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It is made available under aCC-BY 4.0 International license. although Pvalue is quite high. (3) Pedicel. For PL/HL, the difference was statistically insignificant, while for PL/(AL-PL), it was on the verge ofsignificance.The latter again resultesfrom alonger flagellum ofCretaceous males. (4) The first and second segments of flagellum. The means of F1L/HL and F2L/HL are noticeably greater in Cretaceous males, although P values are quite high. The differences between F1L/(AL-F1L) as well as F2L/(AL-F2L) are not well understood dueto low statistical power. The male’sregression lines look very similar for FL/HL, F1L/HL, and F2L/HL (Fig.3). Because of low sample size of Cretaceous males, it is difficult to make broad conclusions. However preliminary results show some differences exist between lengths of flagellum, F1, and F2 in that Cretaceous males havebigger mean values than crown males. The situation is completely different for females in which statistical analysis of indexes shows highly significant differences exist between crown ants and stem ants. ANOVA in 5 groups (extant crown, Cretaceous crown, Sphecomyrmini, Haidomyrmecini, Armaniinae) showed that the means for all indexes were significantly heterogeneous, except for the indexes of pedicel: SL/HL: F =18.38, P<0.0001; FL/HL: 4,79 F =23.29, P<0.0001; PL/HL: F =1.47, P=0.22; F1L/HL: F =34.36, P<0.0001; F2L/HL: F =18.01, 3,74 4,78 4,78 4,78 P<0.0001; SL/FL: F =87.78, P<0.0001; PL/(AL-PL): F =0.35, P=0.79; F1L/(AL-F1L): F =38.58, 3,74 3,74 3,74 P<0.0001; F2L/(AL-F2L): F =18.07, P<0.0001. Aplanned comparison revealed the following picture. 3,74 All indexes ofCretaceous crown ants are very close to the mean values ofthe indexes ofextant crown ants (Table S2); statistical analysis showed no differences between the two groups (Tables S4-S12). Relationship among othergroups are more complicated. Scape (indexes SL/FL, SL/HL): (1) For SL/FL, all groups of stem ants differed significantly from crown ants in that they have shorter scape length compared to flagellum length. Haidomyrmecini were significantly different from both crown ants and Sphecomyrmini, with anintermediate mean value (Table S9). (2) For SL/HL, Sphecomyrmini and Armaniinae have significantly lower mean values than crown ants. Haidomyrmecini have greater means, which are intermediate between crown ants and Sphecomyrmini, Armaniinae (Table S4);Haidomyrmecini’sindexes are widely represented in crown ants (Table S2). (3) For SL/HL, Armaniinae were not different from Sphecomyrminae and Sphecomyrmini (Table S4); for SL/FL, the only available index of Armaniinae is similar to the mean of Sphecomyrmini (Tables S2, S4). (4) Myanmyrma lies close to the regression line of Sphecomyrminae (Figs 4A, 4C). Myanmyrma's SL/HL is similar to the mean of Sphecomyrmini, Armaniinae, and the lowest value of crown ants obtained in this study, the index of Pseudomyrmex pallidus. Myanmyrma's SL/FL is the lowest one, but is quite close to the minimal value ofSphecomyrmini, the index ofSphecomyrmodes contegus (Table S2). (5) Gerontoformica's SL/HL is much greater than that of all stem ants (i.e. Gerontoformica has noticeably longer scape); on the bivariate plot Gerontoformica lies close to the regression line of crown ants (Fig. 4). Such high SL/HL was often found in crown ants, for example in Ponerinae, Aneuretinae, Dolichoderinae, Formicinae (Table S2). Gerontoformica's SL/FL is greater than the mean of Sphecomyrmini, lower than that ofcrown ants,and about equal tothat ofHaidomyrmecini (Table S2). (6) Boltonimecia’s SL/HL is close to the mean of crown ants (similar indexes were found in Dorylinae, Proceratiinae, Myrmecinae, Ponerinae, Agroecomyrmecinae), but greater than that of all Sphecomyrmini and Haidomyrmecini (except for Haidoterminus cippus). Boltonimecia’s SL/FL is greater than that ofmost Sphecomyrmini,but lower thanthat of several species of Haidomyrmecini(Table S2). Pedicel (indexes PL/(AL-PL), PL/HL): (1) For both indexes, there was no statistical difference between the groups studied (Tables S6, S10; Fig.4D). Such stability, as already noted, may be explained by important function of apedicel as a location of the Johnston's organ. (2) The greatest PL/HL is in Cananeuretus occidentalis, followed by Gerontoformica cretacica, 10