A SPIDER AND OTHER ARACHNIDS FROM THE DEVONIAN OF NEW YORK, AND REINTERPRETATIONS OF DEVONIAN ARANEAE by PAUL A. SELDEN, WILLIAM A. SHEAR and PATRICIA M. BONAMO Abstract. The oldest known spider, from the Devonian (Givetian) of Gilboa, New York, is Attercopus fimbriunguis (Shear, Selden and Rolfe), parts ofwhich were originally described as a trigonotarbid, possibly of the genus Gelasinotarbus. Previous reports of Devonian spider fossils, from the Lower Emsian ofAlken- an-der-Mosel, Germany, and the Pragian of Rhynie, Scotland, are shown to be erroneous identifications. Attercopus is placed as sister-taxon to all living spiders, on the basis of characters of the spinneret and the arrangement ofthe patella-tibia joint ofthe walking legs. A cladogram ofthe relationships ofall pulmonate arachnids is presented. A pulmonate arachnid from Gilboa, related to Araneae and Amblypygi, is described as Ecchosispulchribothrium Selden and Shear, gen. et sp. nov., and additional arachnid material is described. A devonian age for the oldest known fossil spider was set by Hirst when he described Palaeocteniza crassipes Hirst, 1923, from the Pragian Rhynie Chert ofAberdeenshire, Scotland. The description ofanother fossil assigned to the Araneae, Archaeometal devonica Stormer, 1976, from the Emsian ofAlken-an-der-Mosel, Germany, added more evidence for the antiquity ofthe order. The find ofa spider spinneret (Shear, Palmer et al. 1989) from the Givetian ofGilboa, New York, provided conclusive evidence for the validity ofthe Devonian as the earliest period in which spider fossils are known to occur. In this paper, results ofa re-examination ofthe Rhynie and Aiken spider fossils are presented: the fossils are not spiders, and are reinterpreted as a probable juvenile trigonotarbid and an indeterminate fossil, respectively. The Gilboa spider is placed in a new genus, Attercopus described here. The new genus includes only the animal previously called Gelasino- , tarbus?fimbriunguis (Shear et al. 1987), which we now regard as the only known Devonian spider, and the oldest known fossil ofthe Araneae. In addition, podomeres originally placed in Arachnida incertae sedis by Shear et al. (1987) are redescribed here, with the addition of new material, as Ecchosis pulchribothrium gen. et sp. nov., and placed in Pulmonata incertae sedis (it may be an amblypygid), and other arachnid remains from Gilboa are described. RHYNIE PALAEOCTENIZA In 1923, Hirst described Palaeocteniza crassipes as a spider from the Pragian Rhynie Chert of Scotland. James Locke and W.A.S. carried out a detailed photographic study of the specimen (British Museum (Natural History) (BM(NH)) In 24670) in 1987 and 1988. The fossil is in a small chip ofchert mounted on a microscope slide. Even ifthe fossil were to be removed from the slide, no additional views could be obtained, owing to the opacity ofthe chert behind the specimen. The specimen itselfis highly three-dimensional, as are many ofthe arthropod remains from Rhynie, and thus difficult to photograph. Adding to the problems are the cloudiness of the matrix, opaque mm inclusions, and the very small size of the specimen, about 0-85 long. In addition to photographs ofthe whole specimen at low magnifications (Text-fig. 1), a series of about thirty-five optical sections was made at higher magnification, using the very shallow depth-of- field characteristic of Nomarski Differential Interference Contrast (NDIC - see below. Methods). These photographs were printed at a large size and each was carefully examined for evidence of IPalaeontologv, Vol. 34, Part 2, 1991, pp. 241-281,7 pls.| © The Palaeontological Association 242 PALAEONTOLOGY, VOLUME 34 text-fig. 1. Palaeocteniza crassipes Hirst. 1923. a, b, two views, at different planes of focus, ofthe holotype (and only known) specimen (BM(NH) In 24670), seen from the left side, anterior to the left, x 130. spider autapomorphies. In addition, each photograph was traced seriatim on agraphics pad andnJie resultant digitized images were stacked and reconstituted asa rotatable virtual solid using theJandel computer program PC3D™ (see below. Methods). We had hoped that James Locke’s efforts to reconstruct the specimen using this program would allow us to examine further details, but this was not to be. The level ofresolution attainable was too low, and there were considerable difficulties in digitizing the images, since shallow as the depth-of-field was, at the necessary magnifications subjective judgement was still required as to what was in the plane of focus and what was not, resulting in further blurring of the lines. A careful examination of the specimen itself and of the serial photographs proved to give the most information. The general condition of the specimen, much crumpled and folded, suggests that it may be a moult. Hirst (1923) noticed a small, thin, scarcely visible object dorsal to the abdomen, which he supposed to be the detached carapace. Since the carapace detaches when arachnids moult, if this identification is correct, its presence and position are further evidence for the specimen being a cast exoskeleton. The prosoma is almost entirely concealed behind the dorsally flexed legs and palps. While the palps appear to be complete, all of the legs on the left side of the specimen (facing the viewer) lack their distal portions. The abdomen is complexly crushed and folded. Hirst (1923) provided a detailed drawing, which, however, incorporates some errors. The proportions of the right palp are not correct in comparison with the left, to which a segment has been added. In ‘restoring’ the loose piece of cuticle to its supposed position as carapace, the SELDEN ET AL.. DEVONIAN ARACHNIDS 243 mass ofwrinkles and folds above the leg coxae (perhaps the true carapace) has been omitted, and some ofthe folds in this structure appear to have been confused with parts ofthe palps. The second or third left leg has the tibia omitted. In the region ofthe supposed abdomen. Hirst noted that what had been made in the drawing to resemble spinnerets might be folds ofcuticle. This is definitely so; the apparent internal structures of the abdomen are also cuticular folds on the right side of the specimen, seen through the left side. In attempting to determine the affinity of this fossil, a process ofelimination was followed. The general appearance and structure ofthe body (a prosoma with five pairs ofleg-like appendages, and an abdomen attached by a narrowed portion) establishes that it is an arachnid, and that it may belong to the known orders Araneae, Amblypygi, Uropygi, Schizomida, or Trigonotarbida. The presence ofleg-like (not raptorial) palps rules out Amblypygi, Uropygi, and Schizomida, at least as they are presently known. Devonian trigonotarbids differ from potentially contemporaneous spiders in a number of ways. While both groups may have segmented abdomens, trigonotarbids have three tergal plates per segment and lack spinnerets. The eyes of any contemporaneous spiders were likely to have been grouped on a centrally located tubercle, as in the modern mesothele spiders, while those of Devonian palaeocharinid trigonotarbids are dispersed in three groups: a median group oftwo, and two lateral groups which may consist ofseveral minor and major lenses each (Shear et al. 1987). All the Devonian trigonotarbids we have examined have a simple bicondylar hinge joint between the patella and tibia, and spiders have a monocondylar rocking joint in this position. Close examination of the abdomen of the specimen failed to reveal any evidence for or against segmentation (despite the clear segmental lines in his illustration. Hirst (1923, p. 460) wrote: ‘...it is impossible to be quite certain whether this [the abdomen] is segmented ornot.’). Thus the number of tergites that might be present for each segment cannot be ascertained. The ‘spinnerets' have already been alluded to; as Hirst inferred, this is in fact a fold ofthe abdominal cuticle that can be traced continuously until it merges with other folds ofthe structure. The entire abdomen was also carefully examined for spinnerets, because we suspected that it might have been twisted through 180°, and because in living mesothele spiders the spinnerets are located about in the middle ofthe ventral surface of the abdomen, which is supposedly their primitive position. We found no indication whatsoever of spinnerets. Careful focusing revealed that among the crushed mass of the prosoma was an object that resembles an eye tubercle and seems to bear at least two hemispherical lens-like protrusions. Unfortunately this evidence is inconclusive, because at least two eye lenses would be present on a median tubercle both in trigonotarbids and spiders. The complicated folding and distortion ofthe carapace and its concealment behind the legs made it impossible for us to find any indication of lateral eye groups. The patella-tibia articulation can be seen on just one of the legs, probably the left third leg. It may be possible to make out two dorsally situated articular condyles on the distal end ofthe patella, but at the level of magnification required to see them, the optical properties of the chert interfere significantly. In summary, the fossil carries none of the autapomorphies of spiders that could be seen on a specimen ofthis size and level ofpreservation, but its identity as a trigonotarbid is only suggested (by the possible pattern ofpatella-tibia articulation). It should be pointed out, however, that scores of trigonotarbids have been seen in the Rhynie chert, and that this specimen is the only one for which a spider identity has been suggested. Our hypothesis is that Palaeocteniza crassipes Hirst is a moulted exoskeleton from an early instar trigonotarbid. ALKEN ARCHAEOMETA One ofonly four fossil sites with Devonian terrestrial animals, Alken-an-der-Mosel, Germany, has yielded impression fossils of lower Emsian age, including trigonotarbids, scorpions, eurypterids, and arthropleurids (Stormer 1976; Brauckmann 1987). One fossil from this deposit, Archaeometa? 244 PALAEONTOLOGY, VOLUME 34 devonica Stormer, 1976, was identified as a spider (Stormer 1976). A policy against type-specimen loans at the Senckenberg Museum, which houses this specimen, meant that we were unable to examine the original. However, we were able to study a plaster cast, and the photograph and drawing published by Stormer. The specimen consists ofan elongate blob with a few transverse lines at one end and a vaguely indicated region at the other which may be part of some plant remains (Stormer 1976, figs 48 and 49; pi. 5, fig. 2a,b). Stormer indicated that he had before him Petrunkevitch’s drawing ofArchaeometa nephilina Pocock, 1911, from the Upper Carboniferous of Britain. This drawing (Petrunkevitch 1949, fig. 159) shows a featureless carapace with seven legs radiating from it, and an elongate abdomen with two longitudinal lines and four or five terminal segments. There are two similar specimens ofA. nephilina in the British Museum (Natural History) which were examined in 1986 by W.A.S., and subsequently by P.A.S. Specimen In 15863 is the more complete and was the specimen figured by Petrunkevitch. It is relatively poorly preserved and little can be added to the diagrammatic illustration and brief description. Specimen In 31259, the holotype, does not show the transverse ‘segmental' lines seen in In 15863. The cuticle is tuberculate and the abdomen bears longitudinal folds; neither ofthese features are found in contemporaneous spider fossils (e.g. Eocteniza silivicola, figured on Pocock’s pi. II, fig. 4), but are more reminiscent of other Carboniferous arachnid groups. There are other details visible on this specimen which would reward a detailed restudy. Nevertheless, there are no features which would distinguish either of these specimens as a spider rather than any other arachnid. In any case, the resemblance of Archaeometa? devonica to these two specimens is vague and probably coincidental. There seems to be no reason to consider Archaeometa? devonica as a spider or a fossil arachnid of any sort. THE GILBOA ARACHNIDS Early reports on the Gilboa fauna (Shear et al. 1984) raised the possibility ofspiders being among the animals present. The tip of an arachnid walking leg tarsus was illustrated, and diagnosed as being from a spider largely on the basis ofserrate ventral setae similar to the silk-handling accessory claws found in some living araneoid spiders. However, in later studies, the possibility of spiders beingpresent receded asit becameclearthat anotherrelatedgroupofarachnids, theTrigonotarbida, dominated the fauna. We were also unable to demonstrate conclusively in the fossils any autapomorphies ofspiders. Shear et al. (1987), in a detailed study ofthe trigonotarbids, assigned all pulmonate arachnid fossils from Gilboa to thisextinct order, which was placed as the plesiomorphic sister group to the other pulmonate orders. One animal represented only by legs was assigned with some doubt to the trigonotarbid genus Gelasinotarbus, and given the species epithetfimbriunguis. This name referred to the characteristic claws, set with ventral cuticular fimbriae, not found in any other trigonotarbids. Other characters in these legs, present but undetected in 1987, we now recognize as conclusive evidence ofa spider. A single femur with a patch ofacute spinules near its base was called Arachnida Incertae sedis B; its cuticle is similar to that offimbriunguis, and other similar femora have now been found in direct connection with pieces ofundoubtedfimbriunguis. A third group of specimens, consisting of podomeres and cuticular fragments, was referred to Arachnida Incertae sedis A. Re-examination ofthese specimens and ofnew material with the same distinctive cuticle has produced evidence that they belong to a pulmonate arachnid, close to Amblypygi and Araneae. To complicate matters further, the tarsus illustrated as a possible spider in Shear et al. (1984, fig. 1 b) is undoubtedly trigonotarbid ; it has smooth claws and lacks a tarsal organ. Late in 1988, conclusive evidence for spiders finally turned up in the Gilboa material: a spinneret (Shear, Palmer et al. 1989). This discovery triggered a search for other possible spider parts, and it was soon realized that the spinneret belonged with the legs described in 1987 as Gelasinotarbus? fimbriunguis. In addition, some previously unassigned cheliceraeand some pieces ofcarapace belong to this animal. SELDEN ET AL.\ DEVONIAN ARACHNIDS 245 The ‘clasp-knife’ form ofthe chilecera, places it in the Pulmonata (= Arachnidea sensu van der Hammen 1977; made up of the orders Trigonotarbida, Uropygi, Schizomida, Amblypygi, and Araneae). Illustrated here for comparison are chelicerae ofthe uropygid Mastigoproctus giganteus (PI. 7, fig. 5), and the amblypygid Heterophrynus elaphus (PI. 7, fig. 6), and see Shear et al. (1987, figs 7, 67, 68) for photographs of trigonotarbid chelicerae. A number ofcharacters unequivocally place the chelicera in Araneae (see discussion under phylogenetic relationships). A cheliceral gland, found only in spiders, is present. The cheliceral fang ofA.fimbriunguis lacks setae, which are present in all other pulmonates. In all other orders ofPulmonata, the largest cheliceral teeth are at the end of the tooth row opposing the tip of the fang (subchelate condition), while in A. fimbriunguis, as in the vast majority ofspiders, the largest teeth occur part-way along the row and nearer to the fang articulation than to the fang tip (the subchelate condition occurs in a small number ofspiders, but the described arrangement is found only in spiders, among the pulmonates). On the basis ofoutgroup comparison with, for example, scorpions, the subchelate state is primitive. Thus there are three definite spider synapomorphies present in the chelicera. A significant apomorphy ofspiders is the presence ofcheliceral venom glands. Whilst the evidence is not entirely certain, in at least two specimens ofA.fimbriunguis chelicerae there may be a subterminal venom pore near the fang tip (PI. 1, fig. 7). In addition, as discussed in the detailed descriptions, the articulations present make it clear that the A.fimbriunguis chelicera must have been orthognath. The legs ofA.fimbriunguis bear numerous lyriform organs; only in spiders are lyriform organs found on podomeres other than the metatarsi. The pieces ofcarapace are referred to A.fimbriunguis on the basis oftheir similarity ofcuticular patterning. The evidence that the spinneret, chelicera, legs, and carapace fragments all come from the same morphospecies is overwhelming. All the chelicerae are identical, except for some size differences, and all ofthe podomere types (trochanter, femur, etc.) are identical within each type. All specimens, including the spinneret and carapace fragments, have the same distinctive cuticular ornamentation, a pattern which appears in no other Gilboa specimens except those that can be unequivocally assigned to the spider on the grounds given above. Finally, the chelicerae and basal leg podomores occur in organicconnection on a number ofslides. Therefore these Gilboa specimens areconsidered to belong to the same species, Attercopusfimbriunguis. There are numerous fragments ofcuticle among the Gilboa slides which resemble the cuticle of A.fimbriunguis at first sight, and which we at first thought could belong to the body ofthe spider. Some ofthese were figured by Shearetal. (1987) and referred to as Arachnida Incertae sedis A. This animal is characterized by: generally large size; scale-like ornament rather than reticulation; setal sockets which range from small to very large; striated macrosetae and thick, striated, bifid spines (PI. 7, figs 4 and 8); groups ofslit sensilla and lyriform organs; ornamented trichobothrial base on the patella. Minute, c. 0005 mm, circular organs occur on the cuticle surface and appear, at low magnification, similar to thecharacteristic little slit sensilla ofAttercopus, but examination at higher magnifications reveals acircular hole rather than acentral slit, so they are not the same organ. None of these minute pores bears a seta, and their function is unknown; nevertheless, the difference in morphology from the little slit organs of Attercopus gives a useful criterion for distinguishing the two cuticle types. New information on Arachnida Incertae sedis A has been discovered during the present study, and the animal is named Ecchosis pulchribothrium gen. et sp. nov., below. The presence of lyriform organs suggests that E. pulchribothrium could be a spider, but the distinctive ornamented trichobothrial socket on the patella is puzzling. Virtually identical trichobothrial sockets are found on the living amblypygid Heterophrynus elaphus (PI. 7, fig. 2), but this animal has a qi ite dilferent leg articulation pattern to that in E. pulchribothrium and a lyriform organ only on , the metatarsus. The identity of E. pulchribothrium thus remains unclear, but we suggest that it is either an aberrant amblypygid or a member of an extinct, undiagnosed arachnid order. PALAEONTOLOGY, VOLUME 246 34 GEOLOGICAL SETTING Stratigraphy The fossils occur in a grey shale in the upper part ofthe Panther Mountain Formation at a locality on Brown Mountain, Gilboa, Schoharie Co., New York (7§' quadrangle sheet 6168 IV NW 1945, m N m approx. 271272 by 142951 E; Banks et al. 1985). Further locality details can be found in Banksetal. (1972). The original site has now been destroyed to make way fora pump-storage power plant associated with Schoharie Reservoir, but much ofthe fossil-bearing shale was removed to the Department of Biology, State University of New York at Binghamton, for later processing. The Panther Mountain Formation is part ofthe Hamilton Group, upper Middle Devonian Erian Series, and is equivalent to the middle Givetian of Europe. Palaeoecologv Detailed discussion ofthe taphonomy and palaeoecology ofthe biota is given in Shear (1986), Shear et al. (1987) and Shear and Bonamo (1988). The Gilboa lithology is a dark grey mudstone. The fauna occurs in close association with mats of interlocking spiny stems of the lycopod Leclercqia. Consideration ofthe manner ofpreservation ofthe plants suggested to Banks et al. (1985) that they were buried in situ by low-energy flood waters. Shear etal. (1984) suggested that the animals, which were living at the site ormay have been carried in by the flow, came to rest by the localized reduction of velocity created by the mesh of Leclercqia. the ‘natural sieve’ effect would exclude large pieces of arthropod cuticle, while the most minute particles could have passed through. Almost all the arthropods recovered from the Gilboa site were undoubtedly terrestrial. The only exception to this is the occurrence ofeurypterid fragments. In the Devonian, these animals lived in both marine and freshwater aquatic habitats, and some were amphibious (Selden 1984, 1985), so their presence in the Gilboa mudstones is not problematical. In addition to the external evidence of sedimentology and associated land flora for the habitat of the arthropods, palaeophysiology provides further proofof their terrestriality (Selden and Jeram 1989). Trichobothria are fine hairs sensitive to high-frequency vibrations, and could only function in air. They occur on the Gilboa pulmonates Gelasinotarbus bonamoae G. bifidus (Shear et al. 1987, figs 105-120), and Ecchosis , pulchribothrium (see below), and the pseudoscorpion (Shear, Schawaller and Bonamo 1989). Book- lungs for air breathing occur in the trigonotarbids ofGilboa (Shear et al. 1987). While we have no evidence of trichobothria or book-lungs in the Gilboa spider Attercopus all living spiders are , terrestrial apart from the secondarily aquatic Argyroneta aquatica found in fresh waters ofEurope, , and the littoral, southern hemisphere Desidae. The phylogenetic discussion (below) indicates that if Attercopus were aquatic, it would also have been secondarily so, since all other Pulmonata are primarily terrestrial. MATERIAL AND METHODS Preservation The animal fossils are preserved as minute, undistinguished, brown to black flakes, which are unrecognizable as animals when in the rock and under incident light microscopy, but transmitted light reveals their zoological nature. The cuticle appears brown in transmitted light, and the depth of colouration is directly correlated with the thickness of the cuticle (or the number of layers of cuticle superimposed in the specimen). The chemical composition ofthe cuticle is not known; the brown colouration suggests it is organic, but the reduction ofmuch ofthe plant material in the same beds to carbon indicates the likelihood that the arthropod cuticle has also been altered, probably by repolymerization of the organic molecules, during diagenesis. The arthropods are strongly compressed, necessitating the use of special techniques, such as NDIC, to separate overlapping layers ofcuticle. For the same reason, scanning electron microscopy (SEM) is virtually useless for the study of these fossils, revealing only surface features: both original structures and diagenetic effects. SELDEN ET AL.: DEVONIAN ARACHNIDS 247 The fossils are fragmentary; only rarely are podomeres and other parts found in organic connection with others. However, the occurrence of such specimens is vital for the correct identification of loose podomeres and reconstruction of the animals. The dearth of pieces of carapace and abdomen of the arachnids can be explained by the fact that podomeres have two surfaces, so that when compressed together they remain coherent and are less likely to fragment than the body parts which consist ofa single sheet ofcuticle. The carapace and abdomen cuticle is represented by the many ‘scraps’ which occur on the slides. The nearly complete trigonotarbid carapaces and abdomens described by Shear et al. (1987) are rare, and mostly consist of both left and right (or dorsal and ventral) surfaces compressed together. Further discussion of the preservation of the Gilboa fauna is given in Shear et al. (1987). Methods The specimens were recovered from the rock matrix by digestion in concentrated hydrofluoric and hydrochloric acids (see Shear et al. 1987; Shear and Bonamo 1988, for details). After washing in distilled water, the animal fossils were separated from the abundant plant fragments, as far as CMC possible, and mounted in or Clearcol on plain microscope slides. The preparation was done in the laboratory of P.M.B. in Binghamton, and the prepared slides were then sent to Hampden- Sydney for study by P.A.S. and W.A.S. The slides were studied using an Olympus Vanox II biological microscope with a Nomarski Differential Interference Contrast (NDIC) facility. This illumination is particularly useful at high magnification and for the optical separation of closely adpressed layers ofcuticle. Use was made of an Olympus SZH stereomicroscope for low magnification work, particularly on comparative extant material; for photography, this was cleared ofmuscles by soaking overnight in a solution of potassium hydroxide. Camera lucida attachments to both microscopes facilitated accurate drawing ofthe specimens, and photographs were taken on 35 mm Kodak Technical Pan film at ASA 50 with Olympus PM10 cameras mounted on these instruments. On plates and text-figures, unless stated otherwise, all photographs were taken in transmitted light with NDIC on the Vanox. The computer program Jandel PC3D™ (available from Jandel Scientific, 2526 Bridgeway, Sausalito, California 94965, USA) was used for the three-dimensional reconstruction of Palaeocteniza crassipes and the program MacClade 2.1 (Maddison and Maddison 1987) was , extremely useful in the phylogenetic analysis. Abbreviations and conventions used in text-figures are as follows; a, anterior, antero-; ar, articulation; ch, chelicera(l); cl, claw; co cx, costa coxalis; cu. cuticle; Cx, coxa; d, dorsal; di, distal; e, edge; f, fold; Fe, femur; gl, gland; i, inferior, infero-; m, arthrodial membrane; ma, marginal; me, median; ms, macroseta; Mt, metatarsus; p, posterior, postero-; pa sp, palpal spinules; Pa, patella; pd, paired; po, poison duct opening; pr, proximal; ps, prosoma; r, ridge; s, superior, supero-; sc, sclerite; si, slit sensilla; sr, serrated; st, sternum, su, surface; t b, trichobothrial base; Ta, tarsus; ta or, tarsal organ; Ti, tibia; Tr, trochanter; tv, transverse; v, ventral; X, artefact. Unless stated otherwise in the legend tocamera lucidadrawings: dashed lines show linear features showing through cuticle from behind; finely dotted areas are internal surfaces; coarse dots show arthrodial membrane; setal sockets and slit sensilla (where shown) are infilled in black when on surfaces showing through from behind; prominent spores (where shown) are in black. Repository ami authorship Type and figured material is deposited in the Department of Invertebrates, American Museum of Natural History, New York (numbers prefixed AMNH). but are referred to in the text by their slide numbers. Most slide numbers consist ofa series number (the first two numbers, e.g. 411.7, or the first only ifonly two numbers are present, e.g. 329), followed by the number ofthe slide within the series. The last, slide, number is prefixed with the letters AR (or Ar) on the slide itself, and quoted thus in earlier publications; these letters are omitted here for brevity. The slide may include more than one specimen, commonly of a different arthropod, but quoting the slide number makes retrieval of specimens for future study easier, facilitates references to earlier papers on the Gilboa 1 467984 1 248 PALAEONTOLOGY. VOLUME 34 table 1. List of specimens mentioned in text. AMNH Slide No. No. Illustration Briefdescription Aitercopusfimbriunguis 329. 43162 PI. 3, fig. 4; Text-fig. 6d palpal femur+patella 329.3 43163 PI. 3, fig. 2; Text-fig. 6b femur 329.3 43163 PI. 4, fig. 1; Text-fig. 7a distal tibia 329.3 43163 PI. 4, fig. 10; Text-fig. 7f metatarsus 329.38 43168 PI. 4. fig. 8 metatarsus 329.39 43098 Text-fig. 12b patella 329.53 43099 PI. 4, fig. 9 tibia 329.57 43100 Text-fig. 12f metatarsus 329.58 43101 Shear et al. 1987, fig. 134 holotype, metatarsus, tarsus 329.59 43102 PI. 3, fig. 3; Text-fig. 6c distal femur+patella 329.59 43102 Text-fig. 12c trochanter 329.69 43106 PI. 2, fig. 5; Text-fig. 5e various; femur, patella, tibia 329.69 43106 PI. 6, fig. 5; Text-fig. 9d palpal tarsus 329 70 43107 Text-fig. 12a paratype, femur+patella . 329.70 43107 Text-fig. 12d, e 2 metatarsi, proximal tarsus 329.16.34 43164 PI. 5, fig. 2 tarsus 329.22.9 43165 PI. 1, fig. 7; Text-fig. 4e chelicera 329.31a.Ml 43166 PI. 3, fig. 7; Text-fig. 6e various; femur+patella 329.31a.M2 43047 PI. 6, fig. 4 legs 334. la. 43170 PI. 5, figs 1 and 3; Text-figs 8a-c 2 legs, patella to tarsus 334.la. 43171 PI. 2, fig. 4; Text-fig. 5d femur 334.la. 43172 PI. 1, figs 6 and 8; Text-fig. 4c chelicera 334.la. 43173 PI. 4, figs 6; Text-fig. 7e tibia 334.la. 43174 PI. 2, fig. 1; Text-fig. 5a femur 334.16.12 43175 PI. 3, fig. 5; Text-fig. 6g distal femur+patella 334.16.34 43176 Text-figs 10. and 11a, b, c spinneret 334.16.38 43177 PI. 5, fig. 5; Text-fig. 8d tarsus 334.16.86 43178 PI. 3, fig. 6; Text-fig. 6f femur+patella 411.02.12M.6 43179 PI. 6, figs 1 and 2; Text-fig. 9a metatarsus+tarsus 411.7.19 43052 paratype, femur 411.7.33 43180 PI. 1, figs 4 and 5; Text-fig. 4d chelicera 411.7.45 43181 PI. 4, fig. 3; Text-fig. 7c distal tibia 411 . 19.83 43182 PI. 2, fig. 2; Text-fig. 5b coxa 411.19.98 43183 PI. 4, fig. 7; Text-fig. 7g distal tibia 411.19.102 43184 PI. 2, fig. 7; Text-fig. 5h 3 coxae, 1 trochanter 411.19.243 43185 PI. 3, fig. 8 proximal femur 411.19.248 43186 PI. 4, fig. 5; Text-fig. 7d patella 411 . 19.250 43187 PI. 2, fig. 8; Text-fig. 5g coxa 411 .19.251 43188 PI. 4, fig. 1 metatarsus 411.20.25 43189 PI. 4, fig. 2; Text-fig. 7b patella 2002.12.49 43190 PI. 4, fig. 4 tibia 2002.12.79 43191 PI. 3, fig. 1; Text-fig. 6a femur 2002.12.90 43192 PI. 1, figs 2 and 3; Text-fig. 4b cheliceral teeth 2002.12.102 43193 PI. 1, fig. 1; Text-fig. 4a anterior carapace Ecchosispulchribothrium 411.1.33 43194 PI. 7, fig. I paratype, distal femur 411.7.37 43195 PI. 6, fig. 6; Text-fig. 9b holotype, patella-Fprox. tibia 411.7.86 43111 Shear et al. 1987, figs 149 and 150 paratype, distal patella 411.19.96 43198 PI. 6, fig. 3; Text-fig. 9c patella 411. 19.137 43169 PI. 7, fig. 4 large, bifid spine 411 . 19.184 43195 PI. 7, fig. 3 lyriform organ 411 . 19.188 43196 PI. 7, fig. 8 paratype, probable tibia 411.19.206 43197 PI. 7, fig. 7 sheet ofcuticle 2002.9.13 43097 PI. 2, fig. 3; Text-fig. 5c coxa Arachnida incertae sedis 334.la. 43198 PI. 5, Fig. 3 flagelliform appendage 2002.9.20 43199 PI. 5, Fig. 4 flagelliform appendage SELDEN ET AL.: DEVONIAN ARACHNIDS 249 fauna in which slide numbers are used, and locates the specimen to the original rock sample. Thus it will be possible in the future to collate data on the whole Gilboa biota to a fine degree ofaccuracy. AMNH Table 1 lists the described specimens both by their accession number and the slide number. A complete list ofthe microscope slides which bear fragments ofAttercopusfimbriunguis, Ecchosis pulchribothrium and Arachnida incertae sedis is deposited as Supplementary Publication No. SUP , 14040, 5 pp., at the British Library, Boston Spa, Wetherby, Yorkshire LS23 7BQ, England. Copies of this can be obtained by writing to the British Library at the above address, enclosing prepaid coupons available from most libraries throughout the world. In addition to the fossils, the following material (both males and females, and from the W. A. Shear Collection, unless otherwise stated) of extant arachnids was studied for comparative purposes: Araneae: Liphistiussumatranus Thorell, Sumatra, American Museum ofNatural History collection; Amblypygi Heterophrynus elaphus Pocock, Ecuador; Uropygi Mastigoproctus : : giganteus (Lucas), Florida; Schizomida: species indet., Mexico. Following previous practice (Shearelal. 1987), authorship ofnew taxa is attributed to Selden and Shear. Bonamo discovered and supervised the preparation ofthe Gilboa material Selden and Shear ; are responsible for other information and ideas in this paper. RECONSTRUCTION OF THE GENERALIZED LEG OF ATTERCOPUS The reconstruction (Text-fig. 2) reflects a combination of the known morphology of various legs, some ofwhich are suspected to be leg by theirclose relationship with palpal femora and chelicerae, 1 but for most specimens the leg-to which they belong is not known. The reconstruction is to be used as a key to interpretation ofthe fossils, and for comparative purposes in a general sense. However, it must be remembered that no one leg of Attercopus fimbriunguis looked exactly like this reconstruction, and in particular, the relative proportions of the podomeres would have varied between legs. There are a number ofways in which the orientation ofpodomores can be inferred. Inferior and superior are fairly straightforward comparison of the articulation points with those of living : spiders, togetherwith aconsideration ofthe way the leg has to work as a functional unit, is normally sufficient. Assessing which is anterior and which posterior is less easy. The trochanter can be oriented by observing its relationship to the coxa, the orientation ofwhich is known because ofthe asymmetry in the joint and comparison with extant arachnids. However, there are no trochanters connected to femora which are sufficiently well preserved to enable the following ofthe orientation down the leg. Since most joints beyond the coxa are symmetrical, their morphology is of little use in orientation, but there is an asymmetrical distribution ofslit sensilla and lyriform organs around the distal joints of podomeres. The palpal femur bears a patch ofspinules in an inferior position, to one side ofits sagittal plane. The function of these spinules is not known, but we are assuming that, whatever their function (see below), they are most likely to occur on the anterior side of the podomere. Therefore, the palpal femur can be oriented, and since this podomere is attached to a patella, this podomere can also, and so on down the leg. A further logical step is required in the assumption that the apparent similar distribution of slit sensilla on palpal podomeres and on the podomeres ofother legs reflects a real serial homology. These assumptions have only been made in order to provide an orientation forthe reconstructed generalized leg, and not for any other purpose. Should the orientation prove to be incorrect, then the references to anterior and posterior would simply require reversal. PHYLOGENETIC RELATIONSHIPS OF ATTERCOPUS F1MBRIUNGUS Cladistic analysis Characters and character states used in the analysis are listed in Table 2, the data matrix is given in Table 3, and the cladogram in Text-figure 3. The tree was rooted by arbitrarily including an ancestor plesiomorphic for all characters. PALAEONTOLOGY, VOLUME 250 34 a si text-fig. 2. Attercopusfimbriunguis(Shear, Selden and Rolfe, 1987). A, reconstruction ofa typical walking leg, posterior aspect, b, diagrammatic representation of walking leg joints, distalmost to the left; each joint is viewed from the distal direction with the anterior to the left, the inner circle representing the distal podomere, the outer the proximal podomere; solid circles are articulation points and straight lines are articulation axes, short lines represent slit sensilla. The body-coxa joint is highly diagrammatic; the lower articulation representing the coxosternal attachment, the upper triangle representing the attachment of the coxa to the prosomal marginal cuticle. The upper coxa-trochanter articulation is a movable sclerite set in the arthrodial membrane, which allows rocking. Slit sensilla omitted from coxal distal joint. The trochanter-femur joint is a horizontal pivot. The femur-patellajoint is a superior bicondylar hinge, and there is a sclerite embedded in the inferior arthrodial membrane. The patella-tibiajoint has a superior articulation, but acloseconnection of the podomeres inferiorly allows the joint to work as a loose vertical pivot. The tibia-metatarsus joint is a superior bicondylarpivot. Themetatarsus-tarsusjointbearsantero-and posterosuperiorarticulationsforming a superior bicondylar hinge, but thejoint may be uncoupled on relaxation ofthe muscles, allowing rocking. Shear et al. (1987) presented a cladistic analysis based on 23 ofthe same characters as used here. The additional characters accommodate the division of the Araneae into Attercopus Mesothelae, , Mygalomorphae, and Araneomorphae. Ifa character is not discussed below, the discussion will be found in the 1987 paper. Some ofthe previously used 23 characters have been re-evaluated; in the following discussion, the character number given is from Table 2, and the character number from Shear et al. (1987) is in brackets. Original characters. Character 8 [5] has been recoded. Further investigation of the patella-tibia articulation demonstrated that thejoint in living spiders has an additional specialization, compression zone Y (CZY, see later), not present in Attercopus. Further, while the joint is immobilized (fixed) in Amblypygi, considerable movement is possible at that articulation in legs 2-4 ofUropygi and Schizomida (in leg 1 the patella and tibia are entirely fused without trace ofa suture). We do not know if the condition on the more posterior legs of Uropygi and Schizomida representsa reversal orthe retention ofaprimitivecondition, but wedecided tocode it as a primitive retention on the grounds ofparsimony. Character 9 [16] has also been recoded, because an