A Revision of the Philoliche aethiopica Species Complex (Diptera: Tabanidae)

ABSTRACT The Philoliche aethiopica (Thunberg) species complex has a history of being taxonomically difficult. This group of pollinating flies is of particular interest to botanists and pollination biologists. This paper reviews its taxonomy and redescribes its species with the aid of new morphological and molecular data. Molecular evidence supports the group's monophyly, with two sister-clades, the “aethiopica” clade (= P. aethiopica) and the “rondani” clade (= P. elegans (Bigot), P. rondani (Bertoloni, P. rubiginosa Dias, and P. umbratipennis (Ricardo) stat. rev.). These clades correspond to two species groups defined by male morphology (“aethiopica” group = P. aethiopica and P. formosa (Austen); “rondani” group = P. elegans, P. rondani, P. rubiginosa, and P. umbratipennis stat. rev.). Molecular evidence also reveals a new character, the index of frons divergence, which is diagnostic for the two clades, and is thus a tool for separating previously indistinguishable females of P. aethiopica from P. rondani. Additionally, I show that some of the longest proboscid specimens, previously confused with P. rondani and P. aethiopica, are actually P. umbratipennis (Ricardo) stat. rev. P. rubiginosa, previously known only from Mozambique, is shown to occur in South Africa, and the male of this species is described for the first time. Furthermore, I show that the distribution of morphotypes within species with differing proboscis lengths is geographically structured, with long-proboscid species occurring only in the Albany Centre of Endemism.


INTRODUCTION
Philoliche Wiedemann, 1820 (often cited as 1828) is an Old World genus of horse flies (Diptera: Tabanidae) including nine subgenera and 119 species, most of which are found in Africa (Oldroyd 1957;Chainey & Oldroyd 1980). This genus represents almost all horse flies in the subfamily Pangoniinae found in sub-Saharan Africa, and constitutes all but one species of the tribe Philolichini. About a dozen of the species in the subgenera Philoliche (24 species) and Stenophara Enderlein, 1922 (22 species) have proboscis 1-3 times as long as their body. These proboscides are specialised for extracting nectar from flowers (Oldroyd 1957;Jobling 1976), and although questioned by some (McKeever & French 1999), females of these species feed on blood (Mitter 1918;Tetley 1918;Sen 1931;Usher 1972;Morita pers. observ.). In southern Africa, several species groups in the subgenus Philoliche are involved in pollinator-mediated speciation of flowering plants (Goldblatt & Manning 1996;Johnson & Steiner 1997). In particular, the P. (P.) aethiopica species complex exhibits proboscis length variation that is assumed to be due to local adaptation to floral resources (Oldroyd 1957;Johnson & Steiner 1995;Johnson 2000;Johnson et al. 2002;Johnson & Morita 2006;Morita 2006, in press b).
The P. (P.) aethiopica species complex is taxonomically difficult. Previously, the group included five nominal species, P. aethiopica (Thunberg, 1789), P. rondani (Bertoloni, 1861(Bertoloni, [1862), P. elegans (Bigot, 1892), P. formosa (Austen, 1920), and P. rubiginosa Dias, 1966. The complex is distributed along the Great Escarpment and Lowveld of south-eastern Africa, from southern South Africa to the Eastern Highlands of Zimbabwe. P. rubiginosa was not previously recognised as occurring in South Africa, and has been systematically misidentified as the more common P. elegans. This is because they share the only diagnostic character previously available in any key, a grey hind margin on the mesothorax (Fig. 1). The two most common species, P. aethiopica and P. rondani, appear to have the longest proboscides, and are frequently observed visiting flowers. Historically, these two species have been difficult to separate (Oldroyd 1957;Usher 1972). Prior to this work, P. rondani was distinguishable from P. aethiopica based on three characters found only in males (J. Chainey, pers. comm.). I now provide characters to distinguish female P. rondani from P. aethiopica, and I show that the longest proboscis specimens previously assigned to P. rondani, and confused with P. aethiopica, are actually P. umbratipennis (Ricardo) stat. rev.
The species in this complex were placed in Philoliche by Oldroyd (1957) following Mackerras (1954). Mackerras (1954) revolutionised tabanid systematics when he used genital morphology to reorganise Tabanidae into three currently accepted subfamilies and their constituent tribes. This had the consequence of placing all but a few African pangoniines into the tribe Philolichini. When Oldroyd (1957) revised the Afrotropical tabanids, he found that all except questionable species records of Pangoniinae in sub-Saharan Africa belonged to Philolichini. Unfortunately, genital characteristics in tabanids are not distinctive below the tribal level (Cole 1927;Mackerras 1954;Burger 2002). Oldroyd (1957) proclaimed the status of African pangoniine genera as chaotic, due to a lack of consistent characteristics. Subsequently, he reclassified all recognised Afrotropica Philolichini as Philoliche, placing the species among nine subgenera. This brought all the members of the P. aethiopica species complex from various generic placements into the subgenus Philoliche (Philoliche). The purpose of this investigation is to clarify species and species groups within the P. aethiopica species complex using novel molecular (Morita, in press a) and new morphological data in a phylogenetic framework. Redescriptions are provided, as well as notes on geographic variants. Molecular evidence places all members of this complex into one clade, with two constituent sister-clades, hereinafter referred to as the "aethiopica" clade (= P. aethiopica and hypothetically the rare P. formosa), and the "rondani" clade (= P. elegans, P. rondani, P. rubiginosa, and P. umbratipennis stat. rev.). These clades correspond to two morphologically defined species groupsaethiopica group (= P. aethiopica and P. formosa) and rondani group (= P. elegans, P. rondani, P. rubiginosa, and P. umbratipennis stat. rev.)-that have differences in male morphology, previously used to separate P. aethiopica from P. rondani. This organisation also facilitated the discovery of a morphological character in females. This character, the index of frons divergence (IFD; sensu Usher 1965), reliably separates these two species groups. Both molecular, and morphological data support the species status of P. umbratipennis (Ricardo, 1900) stat. rev. Finally, I contribute to the known distribution of P. rubiginosa, and describe the male of the species for the first time.

Specimen collection
Seventy-three specimens of the P. aethiopica species complex were collected from across its range over a two-year period (2002)(2003)(2004) in South Africa using an aerial net or an H-trap (Kappmeier 2000). Additionally, six specimens suitable for DNA extraction were provided by Craig Peter and Donald DeKlerk. Collected specimens were stored in 95-100% ethanol for future DNA extraction (see protocol in Morita 2006, in press a). Three hundred and eleven specimens were borrowed from museums and other collections, and examined as detailed below.

Location of specimens examined
The four letter codens used for each collection follow Arnett et al. (1993). Other personal collections have abbreviations that are based on the first and last initial of the owner of the collection followed by "PC" for "personal collection," where they do not conflict with Arnett et al. (1993)

Geographic areas covered
All species discussed in this paper occur in South Africa, Mozambique, Zimbabwe and Zambia. Specific regions of South Africa are given after Van Wyk and Smith (2001). The Albany Centre of Endemism is located at the transition between winter and summer rainfall area in the Eastern Cape of South Africa. The Maputaland Centre of Endemism covers the northeast section of KwaZulu-Natal into the southern portion of Mozambique.

Molecular data and phylogenetic relationships used for this study
Phylogenetic relationships amongst the taxa described here were inferred from the results of previous analyses of molecular data described by Morita (Morita 2006, in press a). This work was based on the combined analysis of 907 bp of the protein-coding nuclear gene CAD (carbomoylphosphate synthase (CPS) domain of carbamoylphosphate synthetase 2, aspartate transcarbamylase, and dihydroorotase, Moulton & Wiegmann 2004) and 886 bp of the mitochondrial gene COI (Brady et al. 2000;Hasegawa et al. 2002). Other South African members of the subgenus P. (Philoliche), exemplars from all other subgenera of Philoliche, all tribes of the horse fly subfamilies Chrysopsinae and Tabaninae, and to a sample of the family Pelecorhynchidae, serve as outgroups to the P. aethiopica species complex (Morita 2006, in press a).

Morphological characters
I searched for morphological characters that had state differences capable of separating species, based on clades suggested by the molecular data. In particular, I sought diagnostic characters for females of these species. Genitalia of male and female tabanids are not diagnostic below tribal level, because of their uniformity (Mackerras 1954;Burger & Chainey 2000), so external skeletal characters were examined. All morphological characters were measured to the nearest 0.1-millimeter using a binocular microscope with an ocular micrometer, and documented photographically using an EntoVision Imaging Suite (GT Vision LLC, MD, USA), with an LED illumination dome. Body length was not measured, due to its pliable nature. Proboscis length was not considered as a diagnostic character because of potential plasticity related to its adaptive value or allometry. However, because of its ecological significance, proboscis lengths were mapped over each species range.

Terminology
Morphological terminology used in this paper follows McAlpine et al. (1981). This terminology differs specifically from that of previous workers on this group in the following way: the mesonotum includes the prescutum, scutum and scutellum (not just the prescutum and scutum); the fifth radial cell = first posterior cell auct; the third medial cell = fourth posterior cell auct; tergites 1 and 2 = segment 1 (they are fused and have been considered as "tergite 1" by some authors), with tergite 3 = segment 2, etc.

Summary tree based on previous phylogenetic analyses
A summary tree, showing the relationships of taxa in this species complex and statistical support of all reconstruction methods, is shown in Fig. 2. All species were sampled, with the exception of P. formosa, which was not collected during this study. Also discussed, but not sampled in the molecular analysis, is the Zimbabwean form of P. aethiopica.
The relationships of taxa in the P. aethiopica species complex were consistent between Maximum Parsimony, Maximum Likelihood and Bayesian phylogenetic inference methods (Morita 2006: figs 1.4-1.6; Morita in press a: figs 4-6). All reconstructions Major clades are marked with letters in circles: A -P. aethiopica species complex, Baethiopica species group, Crondani species group, D -P. rubiginosa, E -P. umbratipennis.

Morphological characters
The index of frontal divergence (as used by Usher 1965) is diagnostic in separating females between the aethiopica and rondani species groups (Fig. 3, see Clades B and C in Fig. 2). Usher (1965: 43) defines the index of frontal divergence as: index of frons divergence = frons width at callus base (at clypeal suture) frons width at vertex This may differ from the "frons index" or "divergence index" of other authors, as the formulae are not always defined. The measurements for each species and morphotype are presented in Table 1. The index of frons divergence differs significantly between the rondani and aethiopica species groups (Wilcoxon/Kruskal-Wallis one-way ranksums test for non-normal distributions: χ 2 =33.7333, p<0.0001). These species groups correspond to the aethiopica and rondani clades (Fig. 2,Clades B and C). The division between the species groups occurs at about a ratio of 2.1 (Fig. 4).   Images of all characters discussed in this paper have been deposited in Morphbank and can be viewed at http: //www.morphbank.net/myCollection/?id=195765.

Species redescriptions
The following are redescriptions of species in the Philoliche aethiopica species complex, with clarification of the limits of each. Literature citing species names in which no nomenclatural action was taken have been omitted from the synonymies. TAXONOMY Philoliche (Philoliche) aethiopica (Thunberg, 1789) Figs 3a, 5a, 6, 7 Diagnosis: P. aethiopica is distinguished from all species in the rondani species group as follows: _ eye facets of similar size, longer lappets with the dorsal process of tarsus 2 extending beyond proximal margin of tarsus 4; ^index of frons divergence greater than 2.1 (Fig. 3). P. aethiopica differs from P. formosa by lacking distinct, narrow, grey sublateral stripes on the scutum. Redescription: Female. Head: Eyes dichoptic; facets of similar size throughout. Frons strongly divergent towards antennae, eye margin with slightly angular curve; covered with yellow or golden pruinescence (pollinosity); slightly convex. Face conically produced forward a distance just greater than sagittal depth of eye. Antennae with first two segments brown or redbrown with black hairs; flagellum dark orange-brown to bright orange, sometimes with dark apex. Palps slender; distal segment wider with black hairs and tapering to a point just after a sensory groove. Proboscis equal to or almost twice body length. Thorax: Mesonotum dark brown with faint or absent sublateral stripes; if sublateral stripes are distinct, they are not narrow and grey, but instead, broad and brown or brownish red. Supra-alar areas of scutum sharply contrasting with white-grey pruinescence beginning at postpronotal lobes and ending at post-alar calli. Pleura light grey with numerous white to yellow hairs.
Wings: Slightly smoky, becoming clearer posteriorly; darkened slightly at cross-veins. Fifth radial cell usually open although closed at margin, or even closed and stalked in some specimens; third medial cell always open. Legs: Coxae grey, paler dorsally and with white hairs; trochanters black; femora orangish red. Tibia of foreleg and midleg paler dorsally than femora; returning to colour of femora distally. Hind tibia same colour as femora, but with black hairs dorsally, so appearing darker. Tarsi concolourous with tibia of same leg. Tarsomeres 3-5 with black hairs on foreleg; tarsomeres 2-4 and tip of tarsomere 1 with black hairs on mid and hind legs. Abdomen: Bright reddish orange dorsally with black medial spot on segments 1 and 2; segments 3 and 4 also darkened medially. Central posterior margin of segment 2 sometimes with white or yellow hairs. Posterior margin of segment 4 with dark hairs interrupted medially with white hairs forming a spot, or entirely white or black haired. Posterolateral margin of segments 3-7 with tufts of black or white hairs as follows: 3black, 4 -white, 5 -black, 6 -white, 7 -white. Male as female except: Head: Eyes holoptic; facets of similar size throughout. Frons small and triangular, pruinescent, covered with long, silky yellow, golden or white hairs; eye margins beginning to diverge a distance from vertex similar to length of first two segments of antennae. Face conically produced, but not as prominent as in female, clypeus highly reduced. Palps reduced compared to female. Clypeus reduced to frontoclypeal membrane. Thorax: Mesonotum with longer, more dense, golden or yellow hairs. Legs: Distally projecting dorsal process on tarsomeres 1 and 2 (lappet). Lappet of tarsomere 2 extending beyond proximal margin of tarsomere 4. Abdomen: Narrower and often more brightly coloured. Variation: This species is divisible into three morphotypes that overlap in distribution ( Fig. 6) and differ distinctly in proboscis length (Fig. 7). The Albany form is larger and consistently has a spot of white hairs in the center of the hind margin of the fourth segment of the abdomen (Fig. 5a). The Natal form is slightly smaller and may have a white fringe along the entire hind margin of the fourth segment of the abdomen, or a spot (like the Albany form) or no white hairs at all. The Zimbabwe form is similar in size to P. rondani, and consistently has a white fringe along the entire hind margin of the fourth abdominal segment.    Oldroyd (1957). However, Ricardo (1900) was uncertain as to where to place Mycteromyia Philippi, 1865, which included M. ensata as described by Bigot (1892); therefore the affiliation of African Mycteromyia with other African pangoniines was not recognised until Oldroyd (1957). After Diatomineura was synonymised by Brethes (1914) with Osca Walker, 1850, Austen (1920)  Diagnosis: P. rondani is distinguished from P. aethiopica as follows: _ eye facets distinctly larger dorsally than ventrally; shorter lappets on foretarsi with the dorsal process of tarsomere 2 never extending beyond proximal margin of tarsomere 4; ^index of frons divergence less than 2.1. Its lack of whitish grey pruinescence on the posterior margin of the scutum separates P. rondani from P. elegans and P. rubiginosa. In contrast to its close relative, P. umbratipennis, P. rondani has all coxae similar in colour and both black and white tufts of hair posterolaterally on segment 5 instead of segment 4.
Head: Eyes dichoptic; facets of similar size throughout. Frons slightly divergent towards antennae, almost parallel; covered with yellow-grey or golden pruinescence (pollinosity), very slightly convex. Face conically produced forward, distance no longer than length of antennae. Antennae with first two segments light brown to brown with black hairs; flagellum orange-brown to bright orange, sometimes with dark apex. Palps slender; distal segment only slightly wider, with black hairs and shallow sensory groove; tip not ending in sharp point. Proboscis length varies from length of head and thorax combined to almost twice body length.
Thorax: Mesonotum yellow-brown to dark brown with faint or no sublateral stripes. If sublateral stripes are more distinct, they are not narrow and grey, but broad and brown or brownish red. Supra-alar areas of scutum sharply contrasting with white-grey pruinescence beginning at postpronotal lobes and ending at post-alar calli. White-grey pruinescence may continue in a very thin band just anterior to scutoscutellar suture, but never more anterior than pre-scutellum. Pleura light grey with numerous white to yellow hairs. Wings: Slightly smoky, becoming clearer posteriorly; rarely darkened slightly at crossveins. Fifth radial cell usually open although closed at margin or even closed and stalked in some specimens; third medial cell always open. Legs: Coxae grey, paler dorsally and with white hairs; trochanters black; femora orange. Tibia of foreleg and midleg paler dorsally than femora. Hind tibia same colour as femora with a few to many black hairs. Tarsi concolourous with tibia of same leg; darkened at apex of fifth tarsomere. Tarsomeres 3-5 with black hairs on foreleg; tarsomeres 2-4 and tip of tarsomere 1 with black hairs on mid and hind legs. Abdomen: Bright orange-yellow with black medial spot on segments 1 and 2; segments 3 and 4 may also be slightly darkened medially, but are most often darker throughout. Central posterior margin of segment 2 sometimes with white or yellow hairs. Posterior margin of segment 4 with black hairs (as rest of segment), except on posterolateral margin. Posterolateral margin of segments 3-7 with tufts of black or white hairs as follows: 3 -black, 4 -white, 5 -black and white, 6 -white, 7 -white.
Male as female except: Head: Eyes holoptic; facets increasing in size from ventral to dorsal very abruptly. Frons extremely small and triangular, pruinescent, covered with long, silky yellow, golden or white hairs; eye margins beginning to diverge a distance from vertex similar to length of antennae. Face conically produced, but not as prominent as in female, clypeus highly reduced. Palps reduced compared to female. Thorax: Mesonotum with longer, more dense, golden or yellow hairs. Legs: With short distally projecting dorsal process on tarsomeres 1 and 2 (lappet). Lappet of tarsomere 2 never extending beyond proximal margin of tarsomere 4. Abdomen: Abdomen narrower and often more brightly coloured. Variation: This widely collected species shows some variation in body size and proboscis length.
Remarks: This species is closely related to P. umbratipennis with which it overlaps in distribution (Figs 8,9), and has distinct differences in proboscis length (Fig. 9). Molecular evidence supports the monophyly of the long proboscid P. umbratipennis, but recognition of P. umbratipennis may render P. rondani paraphyletic based on preliminary molecular analysis (data not shown, see discussion).
Pangonia rondani was described first by Bertoloni (1861) based on one female specimen from Mozambique. When Surcouf (1921) reclassified it as Pangonius Latrielle, he changed the spelling to Pangonius rondanii to create Latin gender agreement with the new genus. Since the epithet was honorific for Prof. Camillio Rondani, this correction was unjustified (IZCN articles 31.2.1, 33.4, and 34.2.1). The incorrect subsequent spelling continued until its correction in the Afrotropical catalog of Diptera (Chainey & Oldroyd 1980). Ricardo (1900) first described Diatomineura (Corizoneura) pallidipennis under Rondani's (1864) subgenus based on a female from Maputo Bay, Mozambique. Later it was placed within Siridorhina Enderlein, and then in Nuceria Walker (see Remarks for P. aethiopica). Oldroyd (1957: 339) synonymised N. pallidipennis with P. (P.) rondani and made P. (P.) rondani a subspecies of P. (P.) aethiopica stating: "This common fly of south-eastern Africa is distinguished from the typical aethiopica by its shorter face in both sexes, by the reduced abdominal markings, and by the absence of dark brown spots on the wings. I have found that in Natal these two forms merge into one another, and I conclude that they should be regarded as subspecies." Note that all taxa overlap in distribution in the Eastern Cape, with the exception of Philoliche rubiginosa. Dias (1966) disagreed and resurrected P. (P.) rondani, based on the overlapping distribution of the two forms, and what he considered to be differences in their ecology. Due to taxonomic uncertainty, Usher (1972) chose to ignore P. (P.) rondani as a separate entity in her review of the South African horsefly fauna, and did not mention that Dias had revised its status. She discussed Oldroyd's decision to separate P. (P.) aethiopica into "two geographically separable subspecies", and stated, "This is a highly complex problem badly in need of further study. For the present, I have treated aethiopica as a single, very variable species" (Usher 1972: 467).
Philoliche (Philoliche) umbratipennis (Ricardo, 1900) Diagnosis: P. umbratipennis can be distinguished from P. rondani by its relatively paler hind coxa and hind femur, by the third abdominal segment which is relatively paler than the other segments, and by having both black and white tufts of hair posterolaterally on segment 4. This species appears to occur only in the Albany Centre of endemism. It usually has a proboscis 2 times greater in length than P. rondani. Description: Female. Head: Eyes dichoptic; facets of similar size throughout. Frons only slightly divergent towards antennae, almost parallel; covered with yellow-grey or golden pruinescence (pollinosity) with shiny, reddish surface visible underneath; convex. Face conically produced forward, distance just greater than sagittal depth of the eye. Antennae with first two segments light brown to brown with black hairs; flagellum orange-brown to bright orange, sometimes with dark apex. Palps slender; distal segment wider with black hairs and sensory groove leading to an aperture; tip ending in a point. Proboscis length usually 1.5-2× body length. Thorax: Mesonotum dark brown with faint or no sublateral stripes. If sublateral stripes are evident, they are not narrow and grey but instead broad and brown or brownish red. Supra-alar areas of scutum sharply contrasting with white-grey pruinescence beginning with the postpronotal lobes and ending at the post-alar calli. Pleura light grey to pinkish grey with numerous white to yellow hairs, also with very few dark hairs. Wings: Smoky, becoming clearer posteriorly; darkened at cross-veins. Fifth radial cell usually open but closed at margin or even closed and stalked in some specimens; third medial cell always open. Legs: Fore and mid coxae grey to pinkish grey, paler dorsally and with white and a few black hairs; hind coxa yellow, trochanters orange; fore and mid femora orange; hind femur yellow, same as hind coxa. Tibia of foreleg and midleg paler than femora. Hind tibia darker than hind femur with black hairs. Tarsi concolourous with tibia of same leg; darkened at apex of fifth tarsomere in fore and midtarsi. Tarsomeres with black hairs. Abdomen: Orange-yellow with black medial spot on segments 1 and 2; segments 3 orange-yellow with scattered spots of pigment, but no true spot; segments 4-7 dark yellow-brown. Central posterior margin of segment 2 sometimes with white or yellow hairs. Posterior margin of segment 4 with black hairs (as rest of segment), except on posterolateral margin. Posterolateral margin of segments 3-7 with tufts of black or white hairs as follows: 3 -black, 4 -black and white, 5 -black, 6 -white, 7 -white. Male as female except: Head: Eyes holoptic; facets increasing in size from ventral to dorsal very abruptly. Frons extremely small and triangular, pruinescent, covered with long, silky yellow, golden or white hairs; eye margins beginning to diverge a distance from vertex similar to length of antennae. Face conically produced, but not as prominent as in female, clypeus highly reduced. Palps reduced compared to female but with sensory pit. Clypeus reduced to frontoclypeal membrane. Thorax: Mesonotum with longer, more dense, golden or yellow hairs. Legs: With short, distally projecting dorsal process on tarsomeres 1 and 2 (lappet). Lappet of tarsomere 2 never extending beyond proximal margin of tarsomere 4. Abdomen: Abdomen narrower and often more brightly coloured.  Remarks: Ricardo (1900) described D. (C.) umbratipennis based on a series of specimens collected by Rev. Barrett at the Annshaw Mission in the former Ciskei (currently Eastern Cape). It was later reclassified as Corizoneura by Austen (1920), and then synonymised under P. aethiopica ssp. aethiopica by Oldroyd (1957). It was left in synonymy with P. aethiopica when Dias separated P. rondani as a species (1966). The molecular (Morita 2006, in press a) and morphological evidence presented here show that Corizoneura umbratipennis (Ricardo) was incorrectly synonymised with P. aethiopica. Examination of the type material reveals clear dimorphism in eye facets from dorsal to ventral, only short lappets on the fore tarsi of the male, and the index of frons divergence less than 2.1 in the female. These characters place P. umbratipennis in the rondani species group. Additionally, molecular data resolve P. umbratipennis as a monophyletic clade within the rondani clade (Fig. 2). However, the sister clade of P. umbratipennis is still unclear. It forms a strongly supported clade with both P. rondani and P. rubiginosa, which is itself a sister one to P. elegans. Strong support for monophyly of P. umbratipennis and its morphological distinctness, even where it overlaps geographically with P. rondani, testifies to this entity's species status (see discussion). Diagnosis: P. elegans is the only species of Philoliche with a band of white-grey pruinescence which forms a sharply contrasting border along the sides and posterior margin of the scutum. It is separated from P. rubiginosa by this sharp contrast in colour, whereas in P. rubiginosa the white-grey pruinescence on the posterior margin of the scutum blends gradually into the brown of the scutum. Also, the frons in females of P. elegans is yellow-grey pruinose, versus the shining, dark red frons of P. rubiginosa.
Head: Eyes dichoptic; facets of similar size throughout. Frons parallel-sided from vertex towards antennae, covered with grey to yellow-grey pruinescence (pollinosity), slightly convex. Face conically produced forward a distance not more than length of antennae. Antennae with first two segments light brown to brown with black hairs; flagellum orange-brown to bright orange sometimes with dark apex. Palps slender; distal segment wider, with black hairs and sensory groove leading to an aperture; apex blunt. Proboscis length rarely longer than length of head and thorax combined.
Thorax: Mesonotum dark brown with faint or no sublateral stripes. If sublateral stripes are more evident, they are not narrow and grey but instead broad and brown or brownish red. Supra-alar areas of scutum sharply contrasting with white-grey pruinescence beginning at postpronotal lobes and continuing along the sides and in band across posterior margin of scutum. Scutellum colour ranging from white-grey to slightly pink. Pleura light grey to pinkish grey with numerous white to yellow hairs. Wings: Slightly smoky, becoming clearer posteriorly; can be slightly darkened at crossveins. Fifth radial cell usually open although closed at margin or even closed and stalked in some specimens; third medial cell always open. Legs: Coxae pinkish grey with white hairs; trochanters orange to brown; fore and mid femora orange; hind femur yellow. Tibia of foreleg and midleg paler than femora. Hind tibia with black hairs; colour darker than hind femur towards apex. Tarsi concolourous with apex of tibia of same leg; darkened at apex of fifth tarsomere in all tarsi.
Abdomen: Bright orange-yellow with black medial spot on segments 1 and 2; segments 3 and 4 may also be slightly darkened medially, but are most often darker throughout. Central posterior margin of segment 2 sometimes with white or yellow hairs. Posterior margin of segment 4 with black hairs (as rest of segment), except on posterolateral margin. Posterolateral margin of segments 3-7 with tufts of black or white hairs as follows: 3 -black, 4 -white, 5 -some black but mostly white, 6 -white, 7white.
Male similar to female except: Head: Eyes holoptic; facets increasing in size from ventral to dorsal very abruptly. Frons extremely small and triangular, pruinescent, covered with long, silky yellow, golden or white hairs; eye margins beginning to diverge a distance from vertex similar to length of antennae. Face conically produced, but not as prominent as in female, clypeus highly reduced. Palps reduced compared to female, but with sensory pit. Clypeus reduced to frontoclypeal membrane. Thorax: Mesonotum with longer, more dense, golden or yellow hairs. Legs: With short distally projecting dorsal process on tarsomeres 1 and 2 (lappet). Lappet of tarsomere 2 never extending beyond proximal margin of tarsomere 4. Abdomen: Narrower and often more brightly coloured.  , iii.1920, 20.xii.1921, 20.i.1922, 3.i.1926. Distribution: Limited to the western portion of the Eastern Cape Province, South Africa (Fig. 8). Remarks: Bigot (1892) described both the male and female of Mycteromyia elegans. This species was later transferred to Philoliche (Philoliche) by Oldroyd (1957: 336), where he comments that "Apart from the grey thoracic border, elegans is closely related to rondanii [sic]". This is noteworthy, since on the next page Oldroyd (1957: 337) classifies P. (P.) rondani as a subspecies of P. (P.) aethiopica.
Philoliche (Philoliche) formosa (Austen, 1920) Corizoneura formosa Austen, 1920: 142 Distribution: South Africa: Northern Cape, Limpopo, KwaZulu-Natal, Eastern Cape. Remarks: C. formosa was described based on four male specimens with distinct longitudinal sublateral stripes on the thorax. Oldroyd (1957: 334) reclassified it as Philoliche (Philoliche) and surmised that it might be a "mere local race of aethiopica". However, he was not confident enough to synonymise these species, due to the small number of specimens available. Very few specimens of this species have ever been collected, and based on the variation in thoracic stripes present within P. (P.) aethiopica, P. (P.) formosa may be a variant of P. (P.) aethiopica. However, the sublateral stripes in the few existing specimens, that I have seen, are quite distinct. I contend that more collections are required to make a definitive decision on this issue (see discussion).
Philoliche (Philoliche) rubiginosa Dias, 1966 Figs 1a, 5b, 8, 9 Philoliche (Philoliche) rubiginosa Dias, 1966: 1189, Mozambique (MHNE). Type not available. nec Corizoneura rubiginosa Bigot, 1892: 617. "Australie." Diagnosis: P. rubiginosa is the only species of Philoliche with white-grey pruinescence that gradually blends from the scutellum, across the posterior margin of the scutum, into the brown colour of the scutum. It is separated from P. elegans by this lack of sharp contrast in colour, whereas in P. elegans the white-grey pruinescence on the posterior margin of the scutum is sharply demarcated from the brown of the scutum. Also, the frons in P. elegans is yellow-grey pruinose, versus P. rubiginosa, which is shining and dark red. Description: Female. Head: Eyes dichoptic; facets of similar size throughout. Frons very slightly divergent towards antennae, almost parallel; reddish brown and shining, if yellow-grey pruinescence present, then only on ridges along eye margin. Face conically produced forward a distance not more than length of the antennae. Antennae with first two segments orange to brown with black hairs; flagellum orange-brown to bright orange. Palps slender; distal segment wider and short with black hairs and sensory groove leading to an aperture; apex blunt. Proboscis length rarely much longer than length of head and thorax combined. Thorax: Mesonotum brown to reddish brown with faint, paler brown or red-brown sublateral stripes. Supra-alar areas of scutum gradually blending into white-grey pruinescence beginning with postpronotal lobes and continuing across the posterior margin of the scutum; pruinescence continues anteriorly beyond the pre-scutellum. Scutellum same white-grey colour that may be slightly pink. Pleura light grey to pinkish grey with numerous white to yellow hairs. Wings: Slightly smoky and slightly tinted yellow, usually without darkening at crossveins. Fifth radial cell usually open, although closed at margin or even closed and stalked in some specimens; third medial cell always open. Darker yellow pterostigma, subcostal cell and costal cell. Legs: Fore coxae brown, mid and hind coxae yellow with white and some black hairs; trochanters orange to brown; fore and mid femora yellow-orange; hind femur paler yellow. Tibia of foreleg and midleg paler than femora. Hind tibia darker than hind femur towards apex; tarsi concolourous with apex of tibia of same leg; darkened at apex of fifth tarsomere in all tarsi. Abdomen: Orange-yellow with black medial spot on segments 1 and 2; segments 3-7 slightly darker than 1 and 2 and with coat of small, black hairs. Central posterior margin of segment 2 sometimes with white or yellow hairs. Posterior margin of segment 4 with black hairs (as rest of segment), except on posterolateral margin. Posterolateral margin of segments 3-7 with tufts of black or white hairs as follows: 3 -black, 4 -mostly white, 5 -mostly white, 6 -white, 7 -white. Male as female except: Head: Eyes holoptic; facets abruptly increasing in size from ventral to dorsal margin. Frons extremely small and triangular, pruinose, covered with long, silky yellow, golden or white hairs; eye margins beginning to diverge a distance from vertex similar to length of antennae. Face conically produced, but not as prominent as in female, clypeus highly reduced. Palps reduced compared to female but with sensory pit. Clypeus reduced to frontoclypeal membrane. Antennae and proboscis similar to female.
Thorax: Legs with short distally projecting dorsal process on tarsomeres 1 and 2 (lappet). Lappet of tarsomere 2 never extending beyond proximal margin of tarsomere 4. Abdomen: Abdomen narrower and often more brightly coloured.  -24.ii.2003-24.ii. , 1^3.iii.2003; 2^Mkhuze (27°37'S:32°02'E), xii.1960, P. de Moor (NMSA); St Lucia (28°23'S:32°25'E): 1^16.xii.1983, J. Manning (NMSA, det. as elegans by Manning, 19831^20.xii.1970, H.V. & B.B. Daly (CASC, det. as elegans by Chainey, 1987 Distribution: Southern Mozambique and north-eastern KwaZulu-Natal, South Africa (Maputaland Centre) (Figs 8,9). Remarks: P. (P.) rubiginosa was described by Dias (1966) based on one female specimen. This species has not been mentioned by any subsequent author other than its appearance in the Afrotropical catalogue of Diptera (Chainey & Oldroyd 1980). Dias (1966) gives its distribution as the southern parts of Mozambique. Based on my own collections and a few specimens in NMSA, CASC, AEIC, and a long series of females collected by J. Esterhuisen, I can extend the species range to the Maputaland region of South Africa near St Lucia, KZN. It is likely due to the previously understood distribution (Dias 1966), that Usher (1972) did not record this species in her review of the South African tabanids. The lack of recognition of this species by Usher (1972) probably also explains why no subsequent researchers have considered P. rubiginosa, and have therefore misidentified this species as P. (P.) elegans, based on the grey colouration of the scutellum in both species (Fig. 1).
Key to species in the Philoliche (Philoliche) aethiopica species complex  (Thunberg) 5 Dorsum of mesothorax with grey band on lateral margins and with grey on posterior margin; scutellum greyish pink (Fig. 1)  . 6 -Dorsum of mesothorax with grey band on lateral margins, but not on posterior margin; scutellum concolourous with centre of mesothorax, not greyish pink .... 7 6 Grey posterior margin of mesothorax sharply contrasting with brown at centre; frons generally light coloured and pollinose, not wholly shiny (Fig. 1b) .... elegans (Bigot) -Grey posterior margin of mesothorax gradually blending into brown anteriorly; frons generally dark and shiny. If frons pollinose, then only along eye margin (Fig. 1a)  The majority of workers interested in species of Philoliche are not exclusively insect systematists, but pollination biologists (Johnson & Steiner 1995, 1997, 2000Johnson 1997Johnson , 2000, 2000, 2006Manning et al. 1999;Johnson et al. 2002;Morita 2006, in press b). The previous inability to separate P. aethiopica and P. rondani females, with consistent misidentification of P. rubiginosa as P. elegans, highlights the importance of a clear redescription of this group. Ecological work has been hampered by taxonomic confusion stemming from combining two similarlooking species groups/clades and grouping taxa based on proboscis length. With researchers able to separate P. aethiopica from P. rondani in both males and females, the complexity of this species group problem shifts to understanding the relationships of taxa within the rondani and aethiopica species groups (clades B and C, Fig. 2). This study presents the additional data required to study the evolution of a recently diverged group of pollinating flies with interesting floral-associated morphology.

The aethiopica species group
Historically, all the species mentioned in this paper, with the exception of P. elegans, have been at one time synonymised with P. aethiopica. Only two of the three geographic forms of P. aethiopica discussed are clearly separate. The only consistent character, other than general geographic range, is the presence of a spot of white hairs in the centre of the hind margin of the fourth segment of the abdomen in the larger Albany form, and the presence of a white fringe along the entire hind margin of the fourth abdominal segment in the smaller, Zimbabwe form. The Natal form appears to be intermediate in the body size, in the proboscis length, and in having all states of white vestiture along the hind margin of the fourth abdominal segment. The Natal form may represent a hybrid zone between the Albany and Zimbabwe forms; however, it is important to note that where the three forms overlap at the edges of their ranges, they appear to be distinct (Figs 6, 7). These might represent zones of secondary contact following recent speciation.
P. formosa is known only from the male sex. The species is rare; I encountered none during extensive collecting, and I know of only four specimens other than the type series. The scarcity of specimens diagnosable to P. formosa, and the fact that the dorsolateral stripes represent the end of a spectrum of variation found within P. aethiopica, led Oldroyd (1957) to designate P. formosa as a possible synonym of P. aethiopica. Males are the more rarely collected sex in Tabanidae, so it is unusual to find 8 males and no females. Also, males in tabanids are often more boldly coloured. Perhaps the distinct thoracic stripes of P. (P.) formosa are due this general trend in sexual dimorphism. Molecular analysis of the Zimbabwe form of P. aethiopica and of P. formosa will aid in understanding the validity and relationships of the morphotypes in this group.

The rondani species group
The closely related members of the rondani species group have both long and short proboscis forms. P. umbratipennis have proboscis lengths of 20-27 mm, whereas P. rondani and P. rubiginosa have proboscis lengths ranging 7-15 mm. The evolutionary implications of this morphological variation among such closely related species have been discussed by Morita (in press b). P. umbratipennis and P. rondani are sympatric. Unfortunately, the mating biology of these species is unknown and patterns of geneflow are not understood. Although, to the extent sampled, previous molecular work shows strong support for monophyly of Clade E (P. rubiginosa + P. rondani + P. umbratipennis, Fig. 2) there is little support for relationships between these taxa. With more exhaustive sampling (Morita, unpubl.), P. umbratipennis appears as a monophyletic group, arising from within a paraphyletic P. rondani, and the status of monophyly in P. rubiginosa is questionable, with only two sequenceable samples. This analysis was based on a 349 bp segment of the mitochondrial gene ND4 that was subsequently found to have multiple paralogues (data not shown). However, even if P. rondani is found to be paraphyletic, it does not necessarily preclude it from species status (De Queiroz 2005;see Sites & Marshall 2003, 2004 for reviews). The biogeography, ecology and paraphyletic nature of P. rondani with respect to P. umbratipennis are patterns consistent with recent, parapatric speciation (Funk et al. 1995a, b;Funk & Omland 2003). This is also supported by the inferred young age of these lineages (Fig. 2) and their current distribution. More directed molecular work on this problem is planned for future studies.
The male of the species was unknown at the time that Dias (1966) described P. rubiginosa. I now describe the male; all males of the rondani species group clearly share distinct eye facet dimorphism. This trait is not seen in other members of P. (Philoliche), and appears to be derived in the group. Interestingly, P. elegans, with a grey hind-margin on the scutum, is sister to the rest of the rondani clade, indicating that this patterning may be pleisiomorphic for the clade. P. rubiginosa has what might be considered an intermediate state in a transition towards the absence of this grey hindmargin in P. rondani. This, combined with the disjunct distribution of these two taxa (Figs 8,9), indicates that P. elegans may have had a wider historical distribution, and that P. rubiginosa and P. rondani may have speciated due to vicariance. Such a hypothesis implies that P. rondani has re-invaded the Albany region, where it overlaps with P. elegans and the longer proboscis form of P. aethiopica (Figs 6-9). This might explain the evolution of P. umbratipennis, which represents a second evolution of long-proboscis morphology in the Albany region of endemism (Morita 2006, in press b) and is coincident with a hypothesis of parapatric speciation for P. umbratipennis under divergent selection. The young age of these lineage divergences, and the suggestion that P. rondani has not yet achieved monophyly subsequent to divergence, support this. However, distinguishing between vicariant speciation followed by a range expansion, and parapatric speciation by divergent selection is extremely difficult since dispersal events occur on a much shorter time-scale than speciation (Coyne & Orr 2004).
The purpose of describing forms or morphospecies here is to identify possible incipient species and entities of potential ecological or evolutionary importance, in the absence of adequate data to delineate new species. Both the rondani and aethiopica species groups and corresponding clades are divisible into at least two geographic morphotypes that have discrete differences in proboscis length where they overlap in distribution (Figs 7,9). The most striking example of this is P. umbratipennis, which has a long proboscis and is restricted in distribution to the Albany Centre of Endemism, where it overlaps with the long-proboscis P. aethiopica Albany form. These species exhibit convergent morphology to a similar environment or habitat-namely the available floral community (Morita 2006, in press b). However, for other such examples, such as the aethiopica species group, consistent external morphological characters other than proboscis length are lacking. Phylogenetic monophyly is not the only or even a necessary criterion for species; lineages may be genetically isolated and yet not be reciprocally monophyletic, or populations within a species may show genetic structure (i.e. monophyly) but not be reproductively isolated (De Queiroz 2005;see Sites & Marshall 2003, 2004 for reviews). Although some discrete morphological differences exist where forms overlap (Fig. 7), implying genetic isolation, the data presented here are inadequate to understand the true nature of gene flow within the aethiopica species group.