The earliest known ants: an analysis of the Cretaceous species and an inference concerning their social organization

The known Cretaceous formicoids are better interpreted from morphological evidence as forming a single subfamily, the Sphecomyrminae, and even a single genus, Sphecomyrma, rather than multiple families and genera. The females appear to have been differentiated as queen and worker castes belonging to the same colonial species instead of winged and wingless solitary females belonging to different species. The former conclusion is supported by the fact that the abdomens of workers of modern ant species and extinct Miocene ant species are smaller relative to the rest of the body than is the case for modern wingless solitary wasps. The wingless Cretaceous formicoids conform to the proportions of ant workers rather than to those of wasps (Figs. 1–2) and hence are reasonably interpreted to have lived in colonies. The Cretaceous formicoids are nevertheless anatomically primitive with reference to modern ants and share some key traits with nonsocial aculeate wasps. They were distributed widely over Laurasia and appear to have been much less abundant than modern ants.


Introduction
This article attempts to resolve a problem in systematics that bears significantly on the origin and early evolution of the ants and hence the antiquity of advanced social behavior in insects generally. Studies of four mid-Cretaceous amber specimens have established the presence of the ant subfamily Sphecomyrminae across a wide portion of present-day North America: Spheco myrma freyi from New Jersey (Wilson et al. 1967a,b) and S. canadensis from Alberta (Wil son 1985a). Dlussky (1975Dlussky ( , 1983 has de scribed an important additional collection of ant like forms from the Upper Cretaceous of the Taymyr Peninsula (extreme north-central Sibe ria), southern Kazakh S.S.R., and the Magadan region of extreme eastern Siberia. He erected 10 new genera to accommodate this material. In his more recent article, he also created a new family, the Armaniidae, to accommodate some of the genera while elevating the Sphecomyr minae to family rank (hence, Sphecomyrmidae) to receive others. This classification is summarized in Table 1. Dlussky's taxonomic interpretation, which is based on careful and accurate descriptions of the new material, has sweeping consequences for our conception of the origin of the ants. First, it presents a picture of an extensive radiation of ants or ant-like forms by the early part of the Upper Cretaceous-in other words, by no later than 80 ma B.P. Second, Dlussky suggested that the Sphecomyrmidae are not true ants or even precursors of the Formicidae but a closely related side branch, while the Armaniidae are the true ancestors of the ants. He doubted that either the Sphecomyrmidae or the Armaniidae were eusocial; in other words he questioned whether they possessed the most advanced mode of social or ganization in which distinct queen and worker castes form overlapping adult generations and care for the developing young (see Wilson 1971).
Because of the light that the early fossils can shed on the origin of the ants and their distinc tive social systems, which are matters hitherto largely unexplored due to the exclusively eusocial status of modern ant species, I decided to reexamine closely the morphological and biogeographic evidence provided by the Cretaceous material. I have arrived at a wholly different conclusion from that of Dlussky. In essence, the differences among the fossils cannot support the separation of the two new families, the Sphe comyrmidae and Armaniidae, from the Formi cidae. It is difficult to justify even the recognition of any genus other than Sphecomyrma on the basis of the morphological evidence. The most parsimonious explanation of the data is that the winged females and males from the Soviet de posits are queens and males of eusocial colonies of which the wingless Sphecomyrma and wingless Soviet species are the workers.
In proposing this view, I wish to emphasize that the disagreement is not over the facts as sembled earlier by Dr. Dlussky and myself. There is only one exception, the supposed divided con dition of the hind trochanter in the Armaniidae, to be discussed later. Instead, the disparity is due to a difference of interpretation of the facts with reference to higher classification and phy logeny. This is all to the good. By such contrasts the basic issues can be better clarified and the gathering of new data stimulated.

Materials and Methods
All of the characters known to vary among the Cretaceous species were first broken into character states. Each genus and species was then redescribed character by character in matrix form to allow close comparison of taxa. The holotypes of two of the best preserved and taxonomically most important Soviet species, Armania robusta and Pseudarmania rasnitsyni, were studied closely in comparison with the descriptions. I also worked with all of the four known specimens of the North American formicoids, placed in Sphe comyrma (Wilson 1985a).
In addition, measurements were made of a wide range of contemporary queen and worker ants and the winged and wingless females of other, nonsocial aculeate families deposited in the Museum of Comparative Zoology, in order to detect consistent differences in body propor tions between social and nonsocial females. A single representative specimen was taken from the series available in each species, and the species in turn were selected to provide a large amount of phyletic diversity. This information was need ed to infer the level of social evolution of the earliest fossils. Further measurements were uti lized from the monograph of Miocene ants of North America by Carpenter (1930). All of these fossils belong to extinct species but surviving subfamilies and in some cases surviving genera. They were considered useful because as rock fos sils their shapes were likely to have been dis torted in the same manner as some of the Soviet specimens, which had been similarly preserved.   In particular, it was important to take into ac count the possible increase in length of the ab domen, which tends to occur when this softest of body parts is crushed laterally or dorsoventrally.
Head length was adopted as a reliable index of body size as a whole (see Wilson 1971). The size of the abdomen is critical as an indicator of reproductive as opposed to nonproductive sta tus in females, because it is the main part of the body containing the ovaries, organs that are pro portionately large in fully social hymenopterans. The abdomen in turn was denned in two ways for the present study. First, a "functional" def inition marks the abdomen as the posteriormost discrete body part, commonly called the gaster by ant specialists-the ovary-bearing portion be hind the one or two segments of the waist. In addition, a strictly homologous definition of the abdomen (or, more precisely, posteriormost ma jor body tagma) was used: all of the true ab dominal segments from II posteriad, including the one or two segments of the waist in ants and a few aculeate wasps. The first, functional defi nition is intuitively the better, because the waist has by definition been reduced to a relatively thin, largely musculated portion that increases the mobility of the gaster. However, both mea sures were employed in order to evaluate the situation as fully as possible.
The results were evaluated with reference to the following criteria derived from standard sys tematic practice on modern faunas of insects and other animals. Individual taxa, whether species, genera, or representative of higher taxa, should be distinguished and named only if they differ by character states. The states can be relatively minor in the case of species, but should be more substantial in the case of genera (e.g., in ants they include the number of antennal and palpal segments, presence or absence of clypeal teeth, and presence or absence of antennal scrobes; and they preferably should exist in multiples). In the case of families, traits should be truly major, as for example the presence or absence of the pet iole, presence or absence of principal exocrine glands, and the pattern of wing folding-again preferably occurring in multiples. Dlussky (1975Dlussky ( , 1983 appears to have used lighter criteria of the kind more commonly accepted in paleontol ogy, in which it is recognized that fossils sepa rated by large geographic distances and stretches of geological time are more likely to belong to different genera or higher categories. Add to this the fact that characters are often obscured due to imperfections in fossilization, and reliance is therefore placed on minor character states.
Because of the importance of the evolutionary issues involved, I believe the criteria employed in assessing the early ant fossils must be the stricter ones used in neontology. That is, it is preferable not to recognize taxa unless the char acter states separating them can be seen and are of approximately the same magnitude used in recognizing contemporary taxa of the same rank.

Results
The results of the character-state analysis for all of the Cretaceous formicoid genera are sum marized in Tables 2 and 3. An inspection shows that no single character state or combination of states can be used to separate a distinct family, the Armaniidae, if neontological standards are applied. Moreover, almost none of the genera can be unambiguously supported from the exist ing evidence. The only exception is Cretomyrma, possessing a hornlike protuberance on the propodeum (in C. unicornis), more compact gaster, and proportionately longer legs; a second, weak er possibility is Dolichomyrma, with an elongate head and slight constriction of the gaster (Dlus sky 1975(Dlus sky , 1983, and new details provided the author in litt.). It cannot be denied that still other genera might be represented by the Soviet fossils; we are only sure that few can be defined by neontological standards with existing data. Hence the synonymy suggested in Table 4 is for the moment the appropriate nomenclatural ar rangement. Figures 1 and 2 show that the measurable, relatively undistorted wingless females among the North American and Soviet fossils, in other words Sphecomyrma freyi and Dolichomyrma longiceps, have small abdomens in proportion to the re mainder of the body, by both the functional and strictly homologous definitions of the abdomen. In this respect they resemble the worker caste of modern ants more than they do the wingless females of modern nonsocial aculeate wasps. Moreover, the winged female type of Pseudar mania rasnitsyni, a relatively undistorted speci men, is within the range of the queens of modern ants (as well as the females of nonsocial aculeate wasps), as shown in Fig. 3. In sum, the wingless females among the Cretaceous fossils are best interpreted as worker ants rather than wingless reproductive aculeate wasps belonging to solitary maniids, and by direct examination I confirmed that it is indeed absent in the holotypes of Ar maria robusta (the type genus and species of the family) and Pseudarmania rasnitsyni. In the winged females of the Armaniidae the petioles are more broadly attached posteriorly to the gaster than in the sphecomyrmine wingless females. But this is not a subfamilial or even species-level character. It is a common difference between the queens and workers belonging to the same species among modern ants, and hence it cannot be reliably used as a taxonomic char acter to separate higher formicoid taxa.

Discussion
The most parsimonious interpretation of the Cretaceous formicoid fossils, neady joining the facts we know, is that they all belong to the subfamily Sphecomyrminae of the family Formicidae, or true ants. Furthermore, so long as contrary evidence is lacking, the Cretaceous fos sils should all be placed provisionally in the gen-era Sphecomyrma and Cretomyrma. It is entirely possible that other genera, and even taxa in ad ditional subfamilies or still higher categories, ex isted in Cenomanian to Santonian times and might be represented by the existing fossils, but until supporting evidence emerges, the conser vative taxonomic arrangement suggested here (Table 4) is both more accurate and heuristic.
This interpretation means that the three phas es represented among the Cretaceous fossils are most reasonably interpreted to be queen, worker, and male formicoids, respectively, in other words, what we would call ants as opposed to wasps. This hypothesis is more clearly depicted by jux taposing the best preserved representatives of the three phases as though they are members of the same colony (see Fig. 4). The hypothesis receives considerable support from the size differences between the best-preserved winged fossils and the best-preserved wingless ones, consistent with their being queens and workers. It receives ad ditional support from the fact that the propor tionate size of the abdomen in the Cretaceous ants is closer to modern ants than to modern aculeate wasps (Figs. 1, 2). Dlussky (1983) made two inferences con necting anatomy to behavior inclining him to the hypothesis that the Cretaceous formicoids were not eusocial. The first is that the tips of antennal funiculi are too far removed from the mandibles to allow the precise coordination re quired for social behavior: "The antennae in these insects did not permit them to control the ma nipulation of small objects, so that they could not have transported their brood or entered into trophallaxis with their larvae-that is, they could not have been true social insects." This suppo sition is surely incorrect. The eusocial vespid wasps have similarly proportioned antennae yet experience no difficulty in transporting all prey objects and placing them on the larvae. They also engage in trophallaxis, or liquid food ex change. The flexibility of the funiculi contribute to these skills, and the twists and curves of the Cretaceous ant antennae suggest that their fu niculi were likewise flexible. To this may be added the fact that some modern ants, such as the primitive Amblyopone pallipes (Traniello 1982) and more advanced Pogonomyrmex badius (Wilson 1971) do not engage in trophallaxis. Hence this form of food exchange was not es- sential for the evolution of eusocial behavior in ants.
Dlussky also inferred that the short, wasplike mandibles of the Cretaceous formicoids "indi cates that these insects did not build true nests, and could have used only pre-existing hollows." But this overlooks the fact that some primitively eusocial wasps and bees use similar mandibles to build quite elaborate nests, mostly from car ton and wax. A few, such as the halictid bees, excavate soil in a very antlike fashion.
To summarize, nothing in the observable anatomy of the Cretaceous formicoids precludes their having possessed a eusocial organization, characterized by brood care, overlap of adult generations, and division of labor between re productive and nonreproductive castes. Al though direct evidence either way is lacking on the matter, these insects could also have con structed nests in the soil, rotting wood, or ar boreal cavities.
The low accumulation rate of Cretaceous for micoids to date indicates that they occurred in low densities compared with modern ants. Only two individuals {Sphecomyrma canadensis) have been found so far among thousands of insects in amber from Alberta Province, Canada (Carpen ter et al. 1939;J. F. McAlpine, pers. comm.). Formicoids constituted just 13 of the 1,200 insea impressions in the Magadan collection and 5 of the 526 impressions in the Kazakhstan col lection, in other words about 1% in both cases (Dlussky 1983). These figures contrast sharply with Oligocene and Miocene deposits. In the Florissant and other shales of North America (Carpenter 1930), as well as the Baltic amber of northern Europe (Wheeler 1914) and amber of the Dominican Republic (Wilson 1985b), the ants are among the most abundant insects, mak ing up a large minority of all insect specimens. Thus the adaptive radiation that took place in the late Cretaceous or early Tertiary, yielding at least three of the dominant modern subfamilies (Myrmicinae, Dolichoderinae, Formicinae) by mid-Eocene times (Wilson 1985a), was accom panied by a marked increase in abundance.
The ants are seen to have paralleled the mam mals by achieving dramatic increases in diversity and abundance around the close of the Mesozoic Era. How they accomplished this breakthrough and managed to sustain a dominant position in the insect world to the present time is a matter of unusual interest, which additional paleontological studies should help to illuminate.