A morphometric comparison of three closely related species of Myrmica (Formicidae), including a new species from England

In the course of a population study on Myrmica sabuleti Meinert a miniature queen was found in some colonies. This paper describes the form as a new parasitic species, Myrmica hirsuta sp.n., and shows how an analogous male was identified. Twelve measurements were then made on a sample of males and females of this new species and these were compared with M. sabuleti and Myrmica scabrinodis Nyl. using a multivariate discriminating technique. The status of the new species is discussed.


Introduction
This paper is divided into three parts. In the first part a new species, belonging t o the scabrinodis-complex of the genus Myrmica, is described and in the second part a morphometric comparison is made between the new species with two other species belonging t o the scabrinodis-complex. In the third part the new species is compared with similar Myrmica and its status is discussed.
In spring 1973, thirty-one colonies of Myrmica sabuleti Meinert were. collected from an area of limestone grassland at Durlston, Purbeck, Dorset, U.K., for a population study that has been published by Elmes (1 974);eight of these colonies contained miniature queens that at that time were called microgynes of M.sabuleti. Apart from their small size these queens could be distinguished easily from normal M.sabuleti queens by possession of much more body hair and an enlarged post-petiole. Subsequently, males that possess characteristics comparable t o those of the small queens were found in some of these M.sabuZeti colonies. This has led me t o consider the small queens and analogous males as a separate species that is described below. Since the first discovery of the new species I have collected specimens from three other sites in Dorset. These were chalk grassland in the Purbeck Hills, 10 miles from the Durlston site; wet heathland o n Hartland Moor, National Nature Reserve, 10 miles from either of the two other sites; and from Lyscombe Down, an area of chalk downland in Central Dorset that is a considerable distance from any of the Purbeck sites. This shows that the form is widespread in Dorset and may well be common throughout Britain.
M.sabuZeti is very close taxonomically t o the species Myrmica scabrinodis Nylander so that for many years M.sabrtleti was treated as a variety of M.scabronodis rather than as a separate species. The worker forms are very hard t o separate although males are easily separated by relative scape length. The purpose of the second part of this paper is t o investigate these three species by morphometry in an attempt to show that they are well separated by their morphology. Initially, a large sample of queens was investigated using four body measurements which were compared directly; later, a larger number of measurements were made upon a smaller sample of queens and males and these were compared by a discriminant analysis.
Myrmica hirsuta spn.  Elmes (1974). Forty-seven of these queens were mounted and placed in my collection, one of these has been designated as holotype while the others form paratypes. The holotype came from a nest containing three queens and 2641 workers of M.sabuleti along with forty-six other M. hirsuta. Twenty-eight Myrmica hirsuta males were collected from three of these colonies in 1974. These also form paratypes. The holotype, two female paratypes and three male paratypes have been deposited in the British Museum (Natural History).
(ii) Holotype: author's reference number D1 -B total length 5.50 mm. A female without wings.
Head: light, ginger brown, moderately sculptured and covered with erect bristles that are conspicuous at the rear side of the head. The mandibles are normal for Myrmica, having a large apical tooth and others reducing in size. The eyes appear t o be large but in fact tend t o be isometric with headwidth compared with Myrmica sabuleti. The antennae have twelve segments and a distinct bend of the scape near t o the base that is typical for the Myrmica scabrinodis group; there are lateral projections at the bend but these are reduced and could be confused with the scape of M.scabrinodis. (Headwidth 1.02 mm, head length 1.12 mm, frons width 0.40 mm, eye length 0.28 mm, length of a typical bristle 0.15 mm; see Fig. 1 for details of measurement.) Thorax: the thorax is moderately sculptured and slightly darker than the head, the parapsidal furrows are distinct. The spines are relatively short and blunt, curved inwards. Bristles are obvious,especially on the pronotum and scutellum. Abdomen: the gaster is lighter in colour than the head. The petiole is wide and the post-petiole is very wide. Bristles are obvious especially on the petiole region (Fig. 2). The post-petiole is wider than it is high and there is a tendency for the surface of the petiole node (seen in lateral profile) t o slope backwards when compared with that of a Myrmica sabuleti queen (Fig. 2 ) . A sting is present. (iii) Paratype: a winged male; author's reference number D1-10, total length 5.30 mm.
Head: is shiny black, finely punctuated with erect hairs that are easily visible at the back of the head. The clypeus is smooth and shiny and the mandibles have a large apical tooth and several equally small teeth. The eyes are isometric with head width but appear t o be large and protruding. The antennae has thirteen segments, the scape is short with a bend near t o the base. (Headwidth 0.82 mm, head length 0.90 mm, frons width 0.24 mm, eye length 0.28 mm, scape length 0.50 mm, length of a typical bristle 0.14 mm; see Fig. 1

for measurement.)
Thorax: is the same colour as the head having fine striations. The wings are as Myrmica with a closed discoidal and partially dissected cubital cell. The legs are yellowish brown with large pectinate spurs on the tibia of the first pair and reduced pectinate spurs on the other two pairs. The spines are rudimentary and have some bristles. (Thorax width 0.8 mm, thorax length 1.80 mm; see (iv) Position in current keys. It was first thought that these specimens were microgynes of M m b u l e t i for they possess many obvious Myrmica characters and appear t o be very microgyne like in size and range compared t o M.sabuleti queens (Elmes, 1974). However, the larger petiole size and the obvious hairiness are definite parasitic characters that can be seen in many inquilines and parasites of Myrmica species (Kutter, 1973). It will be shown later (Fig. 4) that although hairiness is one of the most obvious characters of this species, hence the name, it is a very viable character. No worker form has yet been discovered and evidence presented later will suggest that no worker form exists for this species. If it is attempted t o key the species out using the key given by Collingwood (1958) the queens would key out at couplet six as Myrmica scabrinodis. They could then be split at this stage by inserting a division which would read 'specimens small with many upright hairs easily visible, having relatively large post-petioles and a wide frons, frontal index FI about 40', giving Myrmica hirsuta then: 'specimens with relatively few hairs and a normal width petiole and postpetiole', go t o 7. Couplet 7 would then read as couplet 6 of Collingwood (1958). Similarly males would key out at couplet 5 as Myrmica sabuleti which separates Myrmica scabrinodis from Myrmica schencki Emery; M.hirsuta could be separated at couplet 5 from these others by the width of the post-petiole and their hairiness. Using the more recent key by Bolton & Collingwood (1 975), M. hirsuta 'keys out' as M.scabrin0di.v or M.sabuleti; it then can be easily separated from these in the manner outlined above. If Kutter's classification (Kutter, 1973) is used the female of the species can be coded as A l , B3, C5, D9, E12, F16, G19, H22, 126, K29, M33, N38, 0 4 1 , 448?, host Mmbuleti.
(v) The separation of the male form of M.hirsuta. In 1974 21 1 males were collected from five of the colonies at Durlston, three of these colonies were known t o contain M. hirsuta queens. All the specimens were weighed and a small number from each colony was mounted. Fig. 3 shows the distribution of fresh weights of these males; this distribution is best described by two normal curves that are shown superimposed, giving a fit of 0.90 by x2. The means are 2.9 f 0.4 mg and 4.1 f 0.6 mg. Examination of the mounted sample showed that a proportion of the specimens were much more hairy than the others and that these all come from the colonies known t o contain M.hirsuta queens; the histogram of weights of the hairy specimens (dotted) and the less hairy specimens (vertical hatching) are shown superimposed on Fig. 3. A glance at Fig. 3 shows that the hairy specimens are associated with the normal curve having the smaller mean and comparison with specimens from other sites shows that the less hairy individuals are the typical form of M.sabu2eti. It seems reasonable t o conclude that the smaller hairy form of the male which is associated with colonies that contain the miniature females of M. hirsuta is the male of that species.
(vi) Evidence for a worker f o r m of M. hirsuta. The data in this section have all come from the thirty-one colonies that were collected from Durlston (Elmes, 1974). There is no positive evidence that shows whether M.hirsuta has a worker form or not but all the indirect evidence points t o the absence of a worker form. First, the frequency distribution of headwidths of 196 workers from M m b u l e t i colonies containing M. hirsuta and 575 workers from normal M. sabuleti colonies were examined and it was found that both distributions were best described by a single normal curve and that the means of the two distributions were not significantly different. This showed that if M.hirsuta produces workers there is no significant difference in size between these and M.sabuleti workers. Secondly, a sample of 100 workers from each type of colony was examined carefully and no obvious M.hirsuta character, such as hairiness or large postpetiole width, could be detected. Finally, laboratory tests have not yet resulted in any workers being produced although eggs have been reared to the third instar diapause larval stage; these were most likely t o have become sexual individuals after a low temperature period. It is hoped that details of the laboratory experiments will be published elsewhere.

(i) Material examined and measurements used
The queens that have been measured for this paper are all preserved in the author's collection. Specimens were obtained in two ways: first, from collection of whole colonies for the population studies that have been published elsewhere (Elmes, 1974) and secondly, from among the contents of pitfall traps that have been used for studies on spider populations. Consequently I have a total of 170 M.sabuleti queens, 106 of which were obtained from pitfall traps at three Dorset sites and sixty-four from the nests that were collected from Durlston, Dorset. There are fifty-four M.hirsuta queens in my collection, forty-seven were collected from M.sabuleti nests at Durlston and seven from the pitfall traps. M.scabrinodis is represented by 143 queens, 101 collected from the pitfall traps in Dorset and forty-two from population studies on colonies from Bodmin Moor, Cornwall, and Dartmoor, Devon. I have many fewer males in my collection than queens. The M.scabrinodis males consist of nine collected from pitfall traps in Dorset and fourteen collected from Studland National Nature Reserve in the 1930s by Capt C. Diver, whose collection is held at Furzebrook Research Station. A further four were collected

G . W . Elmes
from a colony in Denmark. Nine Msabuleti males were available from pitfall and one from the Diver Collection, the remaining eighteen being taken from colonies at Durlston. Twenty-eight males of M.hirsuta were collected and mounted from the Durlston colonies.
The following characters were measured o n a sample of queens and all the males from each of the three species (see Fig. 1 for way in which measurements were made). Head characters were: headwidth, head length, frons width, eye length and scape length (the last was measured only for males because the curvature of the female scape makes measurement difficult). Frons ratio has been used by other authors t o separate M.sabuleti from M.scabrinodis (Collingwood, 1958) and this is calculated here as frons widthlheadwidth x 100. Thorax width and epinotal spine length were the thoracic characters measured (spines are essentially a female character and were not measured for males). The measurements of the petiolar region were petiole width, post-petiole width and post-petiole height (petiole height was not measured because it is difficult t o standardize between species).
An attempt was made to assess the hairiness of an individual in two ways. First, the length of a typical bristle o n the back of the head was measured (although the bristles o n an individual are quite variable a distinct difference between the species was hoped for). Secondly, the number of hairs o n the petiole were counted (the petiole was chosen because it is an easily definable region and has sufficient hairs t o make a count worthwhile but not enough t o make counting too difficult). The variability in the number of hairs o n the petiole for specimens of the three species is illustrated in Fig. 2 and

(ii) Examination of all the queens
Initially four measurements were made o n all the queens available t o see whether the three species could be distinguished by morphometrics. The measurements used were headwidth, head length, frons width and post-petiole width. The means and standard deviations of these four measurements are given in Table 1 ; comparisons between the species using t h e 't' test shows that the mean measurements of headwidth, head length and post-petiole width are significantly different (P <0.001). There is no significant difference between the mean frons width of M.scabrinodis and M.sabuleti but the frons width of M.hirsuta differs slightly from the other two (P = 0.01 5). Comparison of subsets of the data for the different species collected from the different sites shows that specimens from every site tend t o conform t o the following observations. In order to save space the tables of measurements are not given here but if head length is plotted against headwidth, head length appears isometric between the three species. There is an indication that  M.scabrinodis has a relatively slightly longer head than the other two species. Overall a moderate separation of the three can be made on head size alone. Fig. 5 shows the frequency histograms of headwidth of the three species with superimposed normal curves that are described by the means and standard deviations given in (iii) Analysis based on twelve measurements In the previous section it was demonstrated that a few measurements were sufficient t o illustrate that the three species differ morphologically. In this section more measurements, made on smaller samples, are investigated for both males and females. The mean values for these measurements are given in Table 2. Comparisons of Tables 1 and 2 show that the measurements based o n the smaller samples agree well with those based o n all the specimens and that the small samples are thus representative.
These measurements can be examined in either of two ways. First, the mean values for the measurements (Table 2) could be compared between the species, but, as we have already shown that there is a distinct size difference and because most of the measurements are correlated with size, relative differences will tend t o be masked. One way of overcoming this problem is t o standardize all the measurements for size; and this is done by considering headwidth t o be a good measure of size and adjusting each specimen t o a headwidth of 1 mm. The standardized size is then a ratio which makes strict statistical comparison difficult; however, the distributions of the standardized measurements, or ratios, are approximately normal, thus means and the associated variances can   within the groups t o give a set of canonical vectors (cv). There will at most be one less vector than the number of groups, thus, in this case, three species will produce two cv's.

Headwidth
If each cv is normalized t o give unit withingroup SD, then for any set of measurements the cv can be thought of the sum of those measurements each adjusted by a weighting value: or cv = Z y ( Q~X~) where Xi is the ith measurement of N measurements and ai is the weighting value applied t o that measurement.  Table 5).
The values of the cv's for each individual of the three species have been calculated from part 1 of Table 5 and Figs. 6 and 7 show the values for the first variate (cvl) plotted against those for the second (cvz) for queens and males respectively. A glance at these figures shows that for both queens and males the three species are a11 well separated by the two variates; each set of individuals forming a distinct cluster for each of the species. The centroid, calculated from the mean values for the original measurements, is marked for each species by a star and a circle that corresponds t o the 99% confidence limit around this centroid is also shown. The numbers of individuals in each species group is relatively large compared t o the number of measurements made therefore the confidence regions in the two dimensions defined by cvl and cv2 can be approximated t o x2 with two degrees of freedom. Thus, a 99% confidence interval for each species would be a circle of radius r = dx' (2; 0.01) = d 9 . 2 1 = 3.03. Fig. 7 shows that the 99% confidence circles for males d o not overlap Compare this for the separation of M. scabrinodis from M.sabuleti based o n the ratio of two measurements, headwidth and frons width (section 2 and Fig. 5 ) . There, up t o 40% of each of t h e species could not be assigned to either group with any degree of confidence. Even using graphical interpolation up t o 10% can be misclassified, compare this t o the very high levels of confidence of correct classification using the canonical variates.
Having shown that the three species can be separated with a high degree of confidence by canonical variates it is usual t o attempt  t o interpret the variates in terms of the original measurements. If part 2 of Table  5 is examined it can be seen that high values of cv, are expected from queens that have long heads, narrow post-petioles, large eyes, wide thorax, narrow frons and long spines, in decreasing order of importance; and from males that have long scapes, narrow thorax and small eyes, in decreasing order of importance. Similarly, high values of cv2 are expected from queens that have narrow heads, long spines, wide frons and narrow petioles, and from males that have wide post-petioles, low post-petioles and many long hairs. Thus as was seen in Fig. 6  If the above interpretation of the canonical variates given in Table 5 are compared with  the results of the simple comparison between  species, Tables 3 and 4, it can be seen that M.hirsuta queens are shown t o differ from the others in much the same way whichever method is used; except that hairiness seems t o be more important in Table 4 than it does in Table 5 . M.scabrinodis queens are shown to differ from M.sabuleti queens in much the same way using either type of analysis in Table 3 or Table 5. In Tables 3, 4 and 5 M.scabrinodis males differ from the other two species, mainly by scape length, and in both Tables 4 and 5 M.hirsuta differs from the others, principally by hairiness and postpetiole width. It will be remembered from part (v) of the descriptive section that scape length, hairiness and post-petiole size were important in separating the males in the first instance, hence it is hardly surprising that these are shown t o be important in the subsequent analysis; this illustrates the caution that should be taken against possible circularity of argument when using this type of analysis. However, in the case of queens initial classification was mainly on ecological considerations and scape shape, which have not been quantified, and hence the morphometric characters that are shown t o be important in their separation are much more independent of initial prejudice than in the case of the males.
It might be asked what is the value of the canonical analysis if it shows what can be demonstrated from a simple two-way comparison (Tables 3 and 4) and if it has dangers of circularity. The answer is that it enables an unknown individual t o be assigned to a group with a known confidence based upon many characters taken simultaneously. Here, it also demonstrates a very good separation of three closely related species in a way that cannot be done otherwise and yet is easily interpreted biologically. In due course it is hoped that this type of analysis can be applied t o a similar set of data for many more Myrmica species, for then multivariate analysis becomes essential as simple two-dimensional comparisons become an almost impossibly cumbersome task.
Size seems t o be relatively unimportant in the canonical analysis unlike a principal components analysis which would be expected t o take out a size component first. This was checked by comparing the weightings given to the canonical vectors with similar weightings of the major components of a principal component analysis. Further, in the case of M.hirsuta queens an analysis of headwidths suggested two size classes of queen (Fig. 5 ) yet detailed examination of the cluster in Fig. 6 shows that the individuals from the two size classes are mixed together. This is important when true microgynes are considered; these should be perfect reductions of the normal queen and thus, if size is ignored, should be classified with the normal queens. It is hoped that a canonical analysis would only separate microgynes from the normal queens if they differed relatively in other respects than size.

Taxonomic position of Mhirsuta
There seems little doubt that M.hirsuta is a workerless parasite that has evolved from its host species, M.sabuleti. There are several routes that can lead t o social parasitism; one is through a process of slave-making that has led t o a complete absence of worker forms in the slave-maker. Another is by a degeneration of a symbiosis or a temporary parasitism to obligate-temporary parasitism then further t o obligate-permanent parasitism. Yet another method is by a process of mutation within a colony to produce a form that achieves breeding isolation within that colony and eventually becomes a parasite of the normal form; this has been discussed by Buschinger (1970). I believe that M.hirsuta has evolved by this last method.
If the idea of rapid mutation within a colony is examined in more detail, several aspects of the colony structure would seem t o be important in order that speciation can be achieved. The original host species should by polygynous and tolerate the recruitment of young queens into its colony who can be considered t o act as a sort of parasite (Elmes, 1973); this faculative polygyny seems essential, for it is hard to see how a monogynous society that will not normally tolerate other fertilized females, even originating from their own colony, will tolerate the recruitment of many parasitic-like queens that are derived from itself. Monogynous colonies are more likely t o be parasitized by species evolved through the faculative temporary parasitic route, for example the parasitism of Te tramorium by Strongylognath us. Faculative polygynous species are most likely t o produce microgyne forms and the mutation that enables the establishment of a microgyne variety is probably not far removed from the mutation required t o produce a truly parasitic variety. In the case of M.rubra microgynes are known t o have a slightly aberrant breeding biology (Elmes, 1976), and if such a species produced a microgyne form that never contributed to the worker population of the colony, it could be considered as parasitic and it is easy t o imagine how breeding isolation of the new form could occur within the original colony. This would be especially so if microgyne forms tend t o mate within the nest and do not need the stimulus of a nuptial flight. Brian & Brian (1955) suggested that the spread of the microgyne form of Myrmica ruginodis Nyl. is chiefly by colony division; therefore if microgyny occurs spontaneously in different geographically isolated populations of the parent species it could lead t o simultaneous evolution of parasitic forms peculiar t o the local host population. If, once a parasitic form has been established for a long time, it still retains the capacity t o spread by flight, then it could become widespread in the host population. I consider that Sifolinia, the satellite genus of Myrmica (Kutter, 1973), has probably evolved in this way from a microgyne form that has retained its proclivity for nuptial flight or a parasitic form such as M.hirsuta. M.hirsuta may indulge in nuptial flights as is suggested by the occurrence of queens in pitfall traps, although an alternative explanation might be that they mate in the nest and then later wander off on foot in search of a new host colony.
I have unpublished morphometric evidence that suggests that M.scabrinodis is very typical for the Myrrnica genus as represented in Western Europe; and if this is compared with a Sifolinia species such as Sifolinia karavajevi Arnoldi, which I consider t o be fairly representative of Sifolinia it is seen that: Sifolinia typically has no worker caste and is a parasite of a Myrmica host, there being very little chance of interbreeding with the host; the males have only twelve antenna1 joints compared with Myrmica which has thirteen; the spurs on the tibia of legs I1 and I11 are absent or simple compared with the reduced pectinate spurs of Myrmica; the petiole and post-petiole are relatively wide, this is a character that can only be recognized after a degree of familiarization with Myrmica; Sifolinia have wings that have a partially open discoidal cell and no division of the cubital cell as opposed t o Myrmica where the cells are closed and partially divided respectively; finally, again a more comparative character, Sifoolinia tend to have longer more dense hairs o n the body compared t o Myrmica.
To summarize, I believe that the genus Sifolinia has evolved by a fairly direct route from microgyne-like ancestors through para-siticMyrmica forms; the path of this evolution seems t o be associated with a degeneration of Myrmica characters until the end product is a form that resembles S. karavajevi, differing from Myrmica as outlined above. Thus the different species of Sifolinia could be phylogenetically closer t o their host Myrmica species than t o each other, the morphological similarities resulting from convergence due t o their pattern of evolution. If this is so, the generic name of Sifolinia is unnecessary and Sifolinia species would best be called Myrmica. However, until more convincing morphological proof or proof obtained, for example, by electrophoretic studies of Sifolinia species and their hosts is available, the generic status of Sifolinia is best left alone; although the concept of a satellite genus (Kutter, 1973) should be more fully appreciated. As long as the two genera are recognized as being separate there exists the difficulty of placing those species that are intermediate between the typical forms, M. hirsuta being such a species: this can only be resolved by placing them in the genera t o which they bear most resemblance. Although this is not satisfactory, if the close relationship of the genera is borne in mind it makes very little difference in practice.
There are only four species of parasitic Myrmica known at this time and M.hirsuta is the fifth, assuming that it is not a microgyne of M.sabuleti nor a species of Sifolinia; multivariate analysis of the morphometry rejects the former, and presence of pectinate tibial spurs and absence of Sifolinia like wing venation and post-petiolar lobes reject the latter supposition. The four species are Myrmica bibikoffi Kutter (1 973), Myrmica myrmicoxena Forel (1874), Myrmica lampra Francoeur (1 968) and Myrmica faniensis van Boven (1970). M.larnpra is a North American species that is parasitic on Myrmica kuschei Wheeler; the female has curved antennal scapes with no projections and the male has only twelve antennal joints, this is quite different from M.hirsuta. The female of M. myrmicoxena resembles the rubra-group (Bernard, 1968) whereas the female ot M. hirsuta falls in the scabrinodis-group; the male of M.myrmicoxena has scabrinodis-group characters but it differs from M.hirsuta in several ways according t o Kutter (1973). M. faniensis has been found with M.scabrinodis and is quite different from M.hirsuta especially in the possession of Sifolinia-type characters shown by the female (van Boven, 1970).

M. bibikoffi is the only other parasitic
Myrmica and this is the species that most resembles M.hirsuta; the females are slightly larger than M.hirsuta and rather less hairy, also their tibial spurs are either indistinctly or not pectinate (Kutter, 1973). In common they have broad post-petioles with no ventral projections, males that have thirteen antennal joints and a shared host, M.sabuleti. Another very similar form is Sifolinia lemasnei (= M. lemasnei Bernard, 1968) which is parasitic upon M.sabuleti and is very similar t o M. hirsuta in general description. However it is smaller than M.hirsuta and has sufficient Sifolinia characters, the most notable of which is a male that has only twelve antennal joints, to be considered a member of that genus. The parasites of M.sabuleti show a range as follows: M.sabuleti microgynes (unknown and hypothetical); M. bibikoffi which is microgynelike but has been found free-living with its own workers; M.hirsuta a parasitic form without workers; S.lemasnei an obligate parasite with many characters normally associated with parasitic ants; and Sifolinia laurae Emery a true and widespread parasite. These then illustrate a hypothetical evolutionary range from the widspread host species through locally distributed parasitic forms t o a widespread parasite. M.scabrinodisfMyrmica alboa Forel, M. faniensis, Sifolinia cabylica Cagniant ( 1 970) and Sifolinia karavajevi (Arnoldi) show a similar though less well defined range. At the present time, with the exception of the nearctic M.lampra, few parasites have been discovered from the M.rubra fruginodis group although microgynes are well known for these two species. It seems likely that as more nests are examined in detail more parasitic forms will be reported, for since the first discovery of Sifolinia in Britain myrmicologists have become more alert t o its likely occurrence and subsequently it has been reported on several other occasions (Barry Bolton, personal communication).
Earlier in this paper it was suggested that M.hirsuta queens might show a bimodal frequency distribution of size. Elmes (1976) discussed variations iu queen size for M.rubra and suggested that normal queens have a size range that reflects trophic variations during their larval stage. It was suggested that microgynes of M.rubra not only have a smaller size as a result of genetic variation but also show trophic variation in the same manner as the normal queens. Elmes ( 1 976) pointed out that the smaller microgynes of M.rubra tended towards intercastes and were frequently unfertilized; Mhirsuta could likewise show irophic variation and the smaller queens are likely t o be incapable of fertilization. Preliminary laboratory experiments have shown that the smallest M.hirsuta have no effect upon M.sabuleti workers whereas the larger M.hirsuta queens have a queen affect that is akin t o that of M.sabuleti queens. This suggests that there is some difference between large and small M.hirsuta queens, not of genetic origin, and that the bimodal size distribution is not an artifact.